3
0
Mirror von https://github.com/PaperMC/Paper.git synchronisiert 2024-11-15 04:20:04 +01:00
Paper/patches/server/0009-MC-Utils.patch

7740 Zeilen
314 KiB
Diff

2021-06-11 14:02:28 +02:00
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Aikar <aikar@aikar.co>
Date: Mon, 28 Mar 2016 20:55:47 -0400
Subject: [PATCH] MC Utils
== AT ==
public net.minecraft.server.level.ServerChunkCache mainThread
public net.minecraft.server.level.ServerLevel chunkSource
public org.bukkit.craftbukkit.inventory.CraftItemStack handle
public net.minecraft.server.level.ChunkMap getVisibleChunkIfPresent(J)Lnet/minecraft/server/level/ChunkHolder;
public net.minecraft.server.level.ServerChunkCache mainThreadProcessor
public net.minecraft.server.level.ServerChunkCache$MainThreadExecutor
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/com/destroystokyo/paper/util/concurrent/WeakSeqLock.java b/src/main/java/com/destroystokyo/paper/util/concurrent/WeakSeqLock.java
new file mode 100644
index 0000000000000000000000000000000000000000..4029dc68cf35d63aa70c4a76c35bf65a7fc6358f
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/concurrent/WeakSeqLock.java
@@ -0,0 +1,68 @@
+package com.destroystokyo.paper.util.concurrent;
+
+import java.util.concurrent.atomic.AtomicLong;
+
+/**
+ * copied from https://github.com/Spottedleaf/ConcurrentUtil/blob/master/src/main/java/ca/spottedleaf/concurrentutil/lock/WeakSeqLock.java
+ * @author Spottedleaf
+ */
+public final class WeakSeqLock {
+ // TODO when the switch to J11 is made, nuke this class from orbit
+
+ protected final AtomicLong lock = new AtomicLong();
+
+ public WeakSeqLock() {
+ //VarHandle.storeStoreFence(); // warn: usages must be checked to ensure this behaviour isn't needed
+ }
+
+ public void acquireWrite() {
+ // must be release-type write
+ this.lock.lazySet(this.lock.get() + 1);
+ }
+
+ public boolean canRead(final long read) {
+ return (read & 1) == 0;
+ }
+
+ public boolean tryAcquireWrite() {
+ this.acquireWrite();
+ return true;
+ }
+
+ public void releaseWrite() {
+ // must be acquire-type write
+ final long lock = this.lock.get(); // volatile here acts as store-store
+ this.lock.lazySet(lock + 1);
+ }
+
+ public void abortWrite() {
+ // must be acquire-type write
+ final long lock = this.lock.get(); // volatile here acts as store-store
+ this.lock.lazySet(lock ^ 1);
+ }
+
+ public long acquireRead() {
+ int failures = 0;
+ long curr;
+
+ for (curr = this.lock.get(); !this.canRead(curr); curr = this.lock.get()) {
+ // without j11, our only backoff is the yield() call...
+
+ if (++failures > 5_000) { /* TODO determine a threshold */
+ Thread.yield();
+ }
+ /* Better waiting is beyond the scope of this lock; if it is needed the lock is being misused */
+ }
+
+ //VarHandle.loadLoadFence(); // volatile acts as the load-load barrier
+ return curr;
+ }
+
+ public boolean tryReleaseRead(final long read) {
+ return this.lock.get() == read; // volatile acts as the load-load barrier
+ }
+
+ public long getSequentialCounter() {
+ return this.lock.get();
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Int.java b/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Int.java
new file mode 100644
index 0000000000000000000000000000000000000000..59868f37d14bbc0ece0836095cdad148778995e6
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Int.java
@@ -0,0 +1,162 @@
+package com.destroystokyo.paper.util.map;
+
+import com.destroystokyo.paper.util.concurrent.WeakSeqLock;
+import it.unimi.dsi.fastutil.longs.Long2IntMap;
+import it.unimi.dsi.fastutil.longs.Long2IntOpenHashMap;
+import it.unimi.dsi.fastutil.longs.LongIterator;
+import it.unimi.dsi.fastutil.longs.LongOpenHashSet;
+import it.unimi.dsi.fastutil.objects.ObjectIterator;
+
+/**
+ * @author Spottedleaf
+ */
+public class QueuedChangesMapLong2Int {
+
+ protected final Long2IntOpenHashMap updatingMap;
+ protected final Long2IntOpenHashMap visibleMap;
+ protected final Long2IntOpenHashMap queuedPuts;
+ protected final LongOpenHashSet queuedRemove;
+
+ protected int queuedDefaultReturnValue;
+
+ // we use a seqlock as writes are not common.
+ protected final WeakSeqLock updatingMapSeqLock = new WeakSeqLock();
+
+ public QueuedChangesMapLong2Int() {
+ this(16, 0.75f);
+ }
+
+ public QueuedChangesMapLong2Int(final int capacity, final float loadFactor) {
+ this.updatingMap = new Long2IntOpenHashMap(capacity, loadFactor);
+ this.visibleMap = new Long2IntOpenHashMap(capacity, loadFactor);
+ this.queuedPuts = new Long2IntOpenHashMap();
+ this.queuedRemove = new LongOpenHashSet();
+ }
+
+ public void queueDefaultReturnValue(final int dfl) {
+ this.queuedDefaultReturnValue = dfl;
+ this.updatingMap.defaultReturnValue(dfl);
+ }
+
+ public int queueUpdate(final long k, final int v) {
+ this.queuedRemove.remove(k);
+ this.queuedPuts.put(k, v);
+
+ return this.updatingMap.put(k, v);
+ }
+
+ public int queueRemove(final long k) {
+ this.queuedPuts.remove(k);
+ this.queuedRemove.add(k);
+
+ return this.updatingMap.remove(k);
+ }
+
+ public int getUpdating(final long k) {
+ return this.updatingMap.get(k);
+ }
+
+ public int getVisible(final long k) {
+ return this.visibleMap.get(k);
+ }
+
+ public int getVisibleAsync(final long k) {
+ long readlock;
+ int ret = 0;
+
+ do {
+ readlock = this.updatingMapSeqLock.acquireRead();
+ try {
+ ret = this.visibleMap.get(k);
+ } catch (final Throwable thr) {
+ if (thr instanceof ThreadDeath) {
+ throw (ThreadDeath)thr;
+ }
+ // ignore...
+ continue;
+ }
+
+ } while (!this.updatingMapSeqLock.tryReleaseRead(readlock));
+
+ return ret;
+ }
+
+ public boolean performUpdates() {
+ this.updatingMapSeqLock.acquireWrite();
+ this.visibleMap.defaultReturnValue(this.queuedDefaultReturnValue);
+ this.updatingMapSeqLock.releaseWrite();
+
+ if (this.queuedPuts.isEmpty() && this.queuedRemove.isEmpty()) {
+ return false;
+ }
+
+ // update puts
+ final ObjectIterator<Long2IntMap.Entry> iterator0 = this.queuedPuts.long2IntEntrySet().fastIterator();
+ while (iterator0.hasNext()) {
+ final Long2IntMap.Entry entry = iterator0.next();
+ final long key = entry.getLongKey();
+ final int val = entry.getIntValue();
+
+ this.updatingMapSeqLock.acquireWrite();
+ try {
+ this.visibleMap.put(key, val);
+ } finally {
+ this.updatingMapSeqLock.releaseWrite();
+ }
+ }
+
+ this.queuedPuts.clear();
+
+ final LongIterator iterator1 = this.queuedRemove.iterator();
+ while (iterator1.hasNext()) {
+ final long key = iterator1.nextLong();
+
+ this.updatingMapSeqLock.acquireWrite();
+ try {
+ this.visibleMap.remove(key);
+ } finally {
+ this.updatingMapSeqLock.releaseWrite();
+ }
+ }
+
+ this.queuedRemove.clear();
+
+ return true;
+ }
+
+ public boolean performUpdatesLockMap() {
+ this.updatingMapSeqLock.acquireWrite();
+ try {
+ this.visibleMap.defaultReturnValue(this.queuedDefaultReturnValue);
+
+ if (this.queuedPuts.isEmpty() && this.queuedRemove.isEmpty()) {
+ return false;
+ }
+
+ // update puts
+ final ObjectIterator<Long2IntMap.Entry> iterator0 = this.queuedPuts.long2IntEntrySet().fastIterator();
+ while (iterator0.hasNext()) {
+ final Long2IntMap.Entry entry = iterator0.next();
+ final long key = entry.getLongKey();
+ final int val = entry.getIntValue();
+
+ this.visibleMap.put(key, val);
+ }
+
+ this.queuedPuts.clear();
+
+ final LongIterator iterator1 = this.queuedRemove.iterator();
+ while (iterator1.hasNext()) {
+ final long key = iterator1.nextLong();
+
+ this.visibleMap.remove(key);
+ }
+
+ this.queuedRemove.clear();
+
+ return true;
+ } finally {
+ this.updatingMapSeqLock.releaseWrite();
+ }
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Object.java b/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Object.java
new file mode 100644
index 0000000000000000000000000000000000000000..7bab31a312463cc963d9621cdc543a281459bd32
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/map/QueuedChangesMapLong2Object.java
@@ -0,0 +1,202 @@
+package com.destroystokyo.paper.util.map;
+
+import com.destroystokyo.paper.util.concurrent.WeakSeqLock;
+import it.unimi.dsi.fastutil.longs.Long2ObjectLinkedOpenHashMap;
+import it.unimi.dsi.fastutil.longs.Long2ObjectMap;
+import it.unimi.dsi.fastutil.objects.ObjectBidirectionalIterator;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.List;
+
+/**
+ * @author Spottedleaf
+ */
+public class QueuedChangesMapLong2Object<V> {
+
+ protected static final Object REMOVED = new Object();
+
+ protected final Long2ObjectLinkedOpenHashMap<V> updatingMap;
+ protected final Long2ObjectLinkedOpenHashMap<V> visibleMap;
+ protected final Long2ObjectLinkedOpenHashMap<Object> queuedChanges;
+
+ // we use a seqlock as writes are not common.
+ protected final WeakSeqLock updatingMapSeqLock = new WeakSeqLock();
+
+ public QueuedChangesMapLong2Object() {
+ this(16, 0.75f); // dfl for fastutil
+ }
+
+ public QueuedChangesMapLong2Object(final int capacity, final float loadFactor) {
+ this.updatingMap = new Long2ObjectLinkedOpenHashMap<>(capacity, loadFactor);
+ this.visibleMap = new Long2ObjectLinkedOpenHashMap<>(capacity, loadFactor);
+ this.queuedChanges = new Long2ObjectLinkedOpenHashMap<>();
+ }
+
+ public V queueUpdate(final long k, final V value) {
+ this.queuedChanges.put(k, value);
+ return this.updatingMap.put(k, value);
+ }
+
+ public V queueRemove(final long k) {
+ this.queuedChanges.put(k, REMOVED);
+ return this.updatingMap.remove(k);
+ }
+
+ public V getUpdating(final long k) {
+ return this.updatingMap.get(k);
+ }
+
+ public boolean updatingContainsKey(final long k) {
+ return this.updatingMap.containsKey(k);
+ }
+
+ public V getVisible(final long k) {
+ return this.visibleMap.get(k);
+ }
+
+ public boolean visibleContainsKey(final long k) {
+ return this.visibleMap.containsKey(k);
+ }
+
+ public V getVisibleAsync(final long k) {
+ long readlock;
+ V ret = null;
+
+ do {
+ readlock = this.updatingMapSeqLock.acquireRead();
+
+ try {
+ ret = this.visibleMap.get(k);
+ } catch (final Throwable thr) {
+ if (thr instanceof ThreadDeath) {
+ throw (ThreadDeath)thr;
+ }
+ // ignore...
+ continue;
+ }
+
+ } while (!this.updatingMapSeqLock.tryReleaseRead(readlock));
+
+ return ret;
+ }
+
+ public boolean visibleContainsKeyAsync(final long k) {
+ long readlock;
+ boolean ret = false;
+
+ do {
+ readlock = this.updatingMapSeqLock.acquireRead();
+
+ try {
+ ret = this.visibleMap.containsKey(k);
+ } catch (final Throwable thr) {
+ if (thr instanceof ThreadDeath) {
+ throw (ThreadDeath)thr;
+ }
+ // ignore...
+ continue;
+ }
+
+ } while (!this.updatingMapSeqLock.tryReleaseRead(readlock));
+
+ return ret;
+ }
+
+ public Long2ObjectLinkedOpenHashMap<V> getVisibleMap() {
+ return this.visibleMap;
+ }
+
+ public Long2ObjectLinkedOpenHashMap<V> getUpdatingMap() {
+ return this.updatingMap;
+ }
+
+ public int getVisibleSize() {
+ return this.visibleMap.size();
+ }
+
+ public int getVisibleSizeAsync() {
+ long readlock;
+ int ret;
+
+ do {
+ readlock = this.updatingMapSeqLock.acquireRead();
+ ret = this.visibleMap.size();
+ } while (!this.updatingMapSeqLock.tryReleaseRead(readlock));
+
+ return ret;
+ }
+
+ // unlike mojang's impl this cannot be used async since it's not a view of an immutable map
+ public Collection<V> getUpdatingValues() {
+ return this.updatingMap.values();
+ }
+
+ public List<V> getUpdatingValuesCopy() {
+ return new ArrayList<>(this.updatingMap.values());
+ }
+
+ // unlike mojang's impl this cannot be used async since it's not a view of an immutable map
+ public Collection<V> getVisibleValues() {
+ return this.visibleMap.values();
+ }
+
+ public List<V> getVisibleValuesCopy() {
+ return new ArrayList<>(this.visibleMap.values());
+ }
+
+ public boolean performUpdates() {
+ if (this.queuedChanges.isEmpty()) {
+ return false;
+ }
+
+ final ObjectBidirectionalIterator<Long2ObjectMap.Entry<Object>> iterator = this.queuedChanges.long2ObjectEntrySet().fastIterator();
+ while (iterator.hasNext()) {
+ final Long2ObjectMap.Entry<Object> entry = iterator.next();
+ final long key = entry.getLongKey();
+ final Object val = entry.getValue();
+
+ this.updatingMapSeqLock.acquireWrite();
+ try {
+ if (val == REMOVED) {
+ this.visibleMap.remove(key);
+ } else {
+ this.visibleMap.put(key, (V)val);
+ }
+ } finally {
+ this.updatingMapSeqLock.releaseWrite();
+ }
+ }
+
+ this.queuedChanges.clear();
+ return true;
+ }
+
+ public boolean performUpdatesLockMap() {
+ if (this.queuedChanges.isEmpty()) {
+ return false;
+ }
+
+ final ObjectBidirectionalIterator<Long2ObjectMap.Entry<Object>> iterator = this.queuedChanges.long2ObjectEntrySet().fastIterator();
+
+ try {
+ this.updatingMapSeqLock.acquireWrite();
+
+ while (iterator.hasNext()) {
+ final Long2ObjectMap.Entry<Object> entry = iterator.next();
+ final long key = entry.getLongKey();
+ final Object val = entry.getValue();
+
+ if (val == REMOVED) {
+ this.visibleMap.remove(key);
+ } else {
+ this.visibleMap.put(key, (V)val);
+ }
+ }
+ } finally {
+ this.updatingMapSeqLock.releaseWrite();
+ }
+
+ this.queuedChanges.clear();
+ return true;
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/maplist/ChunkList.java b/src/main/java/com/destroystokyo/paper/util/maplist/ChunkList.java
new file mode 100644
index 0000000000000000000000000000000000000000..554f4d4e63c1431721989e6f502a32ccc53a8807
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/maplist/ChunkList.java
@@ -0,0 +1,128 @@
+package com.destroystokyo.paper.util.maplist;
+
+import it.unimi.dsi.fastutil.longs.Long2IntOpenHashMap;
+import java.util.Arrays;
+import java.util.Iterator;
+import java.util.NoSuchElementException;
+import net.minecraft.world.level.chunk.LevelChunk;
+
+// list with O(1) remove & contains
+/**
+ * @author Spottedleaf
+ */
+public final class ChunkList implements Iterable<LevelChunk> {
+
+ protected final Long2IntOpenHashMap chunkToIndex = new Long2IntOpenHashMap(2, 0.8f);
+ {
+ this.chunkToIndex.defaultReturnValue(Integer.MIN_VALUE);
+ }
+
+ protected static final LevelChunk[] EMPTY_LIST = new LevelChunk[0];
+
+ protected LevelChunk[] chunks = EMPTY_LIST;
+ protected int count;
+
+ public int size() {
+ return this.count;
+ }
+
+ public boolean contains(final LevelChunk chunk) {
+ return this.chunkToIndex.containsKey(chunk.coordinateKey);
+ }
+
+ public boolean remove(final LevelChunk chunk) {
+ final int index = this.chunkToIndex.remove(chunk.coordinateKey);
+ if (index == Integer.MIN_VALUE) {
+ return false;
+ }
+
+ // move the entity at the end to this index
+ final int endIndex = --this.count;
+ final LevelChunk end = this.chunks[endIndex];
+ if (index != endIndex) {
+ // not empty after this call
+ this.chunkToIndex.put(end.coordinateKey, index); // update index
+ }
+ this.chunks[index] = end;
+ this.chunks[endIndex] = null;
+
+ return true;
+ }
+
+ public boolean add(final LevelChunk chunk) {
+ final int count = this.count;
+ final int currIndex = this.chunkToIndex.putIfAbsent(chunk.coordinateKey, count);
+
+ if (currIndex != Integer.MIN_VALUE) {
+ return false; // already in this list
+ }
+
+ LevelChunk[] list = this.chunks;
+
+ if (list.length == count) {
+ // resize required
+ list = this.chunks = Arrays.copyOf(list, (int)Math.max(4L, count * 2L)); // overflow results in negative
+ }
+
+ list[count] = chunk;
+ this.count = count + 1;
+
+ return true;
+ }
+
+ public LevelChunk getChecked(final int index) {
+ if (index < 0 || index >= this.count) {
+ throw new IndexOutOfBoundsException("Index: " + index + " is out of bounds, size: " + this.count);
+ }
+ return this.chunks[index];
+ }
+
+ public LevelChunk getUnchecked(final int index) {
+ return this.chunks[index];
+ }
+
+ public LevelChunk[] getRawData() {
+ return this.chunks;
+ }
+
+ public void clear() {
+ this.chunkToIndex.clear();
+ Arrays.fill(this.chunks, 0, this.count, null);
+ this.count = 0;
+ }
+
+ @Override
+ public Iterator<LevelChunk> iterator() {
+ return new Iterator<LevelChunk>() {
+
+ LevelChunk lastRet;
+ int current;
+
+ @Override
+ public boolean hasNext() {
+ return this.current < ChunkList.this.count;
+ }
+
+ @Override
+ public LevelChunk next() {
+ if (this.current >= ChunkList.this.count) {
+ throw new NoSuchElementException();
+ }
+ return this.lastRet = ChunkList.this.chunks[this.current++];
+ }
+
+ @Override
+ public void remove() {
+ final LevelChunk lastRet = this.lastRet;
+
+ if (lastRet == null) {
+ throw new IllegalStateException();
+ }
+ this.lastRet = null;
+
+ ChunkList.this.remove(lastRet);
+ --this.current;
+ }
+ };
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/maplist/EntityList.java b/src/main/java/com/destroystokyo/paper/util/maplist/EntityList.java
new file mode 100644
index 0000000000000000000000000000000000000000..0133ea6feb1ab88f021f66855669f58367e7420b
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/maplist/EntityList.java
@@ -0,0 +1,128 @@
+package com.destroystokyo.paper.util.maplist;
+
+import it.unimi.dsi.fastutil.ints.Int2IntOpenHashMap;
+import net.minecraft.world.entity.Entity;
+import java.util.Arrays;
+import java.util.Iterator;
+import java.util.NoSuchElementException;
+
+// list with O(1) remove & contains
+/**
+ * @author Spottedleaf
+ */
+public final class EntityList implements Iterable<Entity> {
+
+ protected final Int2IntOpenHashMap entityToIndex = new Int2IntOpenHashMap(2, 0.8f);
+ {
+ this.entityToIndex.defaultReturnValue(Integer.MIN_VALUE);
+ }
+
+ protected static final Entity[] EMPTY_LIST = new Entity[0];
+
+ protected Entity[] entities = EMPTY_LIST;
+ protected int count;
+
+ public int size() {
+ return this.count;
+ }
+
+ public boolean contains(final Entity entity) {
+ return this.entityToIndex.containsKey(entity.getId());
+ }
+
+ public boolean remove(final Entity entity) {
+ final int index = this.entityToIndex.remove(entity.getId());
+ if (index == Integer.MIN_VALUE) {
+ return false;
+ }
+
+ // move the entity at the end to this index
+ final int endIndex = --this.count;
+ final Entity end = this.entities[endIndex];
+ if (index != endIndex) {
+ // not empty after this call
+ this.entityToIndex.put(end.getId(), index); // update index
+ }
+ this.entities[index] = end;
+ this.entities[endIndex] = null;
+
+ return true;
+ }
+
+ public boolean add(final Entity entity) {
+ final int count = this.count;
+ final int currIndex = this.entityToIndex.putIfAbsent(entity.getId(), count);
+
+ if (currIndex != Integer.MIN_VALUE) {
+ return false; // already in this list
+ }
+
+ Entity[] list = this.entities;
+
+ if (list.length == count) {
+ // resize required
+ list = this.entities = Arrays.copyOf(list, (int)Math.max(4L, count * 2L)); // overflow results in negative
+ }
+
+ list[count] = entity;
+ this.count = count + 1;
+
+ return true;
+ }
+
+ public Entity getChecked(final int index) {
+ if (index < 0 || index >= this.count) {
+ throw new IndexOutOfBoundsException("Index: " + index + " is out of bounds, size: " + this.count);
+ }
+ return this.entities[index];
+ }
+
+ public Entity getUnchecked(final int index) {
+ return this.entities[index];
+ }
+
+ public Entity[] getRawData() {
+ return this.entities;
+ }
+
+ public void clear() {
+ this.entityToIndex.clear();
+ Arrays.fill(this.entities, 0, this.count, null);
+ this.count = 0;
+ }
+
+ @Override
+ public Iterator<Entity> iterator() {
+ return new Iterator<Entity>() {
+
+ Entity lastRet;
+ int current;
+
+ @Override
+ public boolean hasNext() {
+ return this.current < EntityList.this.count;
+ }
+
+ @Override
+ public Entity next() {
+ if (this.current >= EntityList.this.count) {
+ throw new NoSuchElementException();
+ }
+ return this.lastRet = EntityList.this.entities[this.current++];
+ }
+
+ @Override
+ public void remove() {
+ final Entity lastRet = this.lastRet;
+
+ if (lastRet == null) {
+ throw new IllegalStateException();
+ }
+ this.lastRet = null;
+
+ EntityList.this.remove(lastRet);
+ --this.current;
+ }
+ };
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/maplist/IBlockDataList.java b/src/main/java/com/destroystokyo/paper/util/maplist/IBlockDataList.java
new file mode 100644
2021-11-23 12:27:39 +01:00
index 0000000000000000000000000000000000000000..277cfd9d1e8fff5d9b5e534b75c3c5162d58b0b7
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/maplist/IBlockDataList.java
@@ -0,0 +1,128 @@
+package com.destroystokyo.paper.util.maplist;
+
+import it.unimi.dsi.fastutil.longs.LongIterator;
+import it.unimi.dsi.fastutil.shorts.Short2LongOpenHashMap;
+import java.util.Arrays;
2021-11-23 12:27:39 +01:00
+import net.minecraft.world.level.block.Block;
2021-06-11 14:02:28 +02:00
+import net.minecraft.world.level.block.state.BlockState;
+import net.minecraft.world.level.chunk.GlobalPalette;
+
+/**
+ * @author Spottedleaf
+ */
+public final class IBlockDataList {
+
2021-11-23 12:27:39 +01:00
+ static final GlobalPalette<BlockState> GLOBAL_PALETTE = new GlobalPalette<>(Block.BLOCK_STATE_REGISTRY);
2021-06-11 14:02:28 +02:00
+
+ // map of location -> (index | (location << 16) | (palette id << 32))
+ private final Short2LongOpenHashMap map = new Short2LongOpenHashMap(2, 0.8f);
+ {
+ this.map.defaultReturnValue(Long.MAX_VALUE);
+ }
+
+ private static final long[] EMPTY_LIST = new long[0];
+
+ private long[] byIndex = EMPTY_LIST;
+ private int size;
+
+ public static int getLocationKey(final int x, final int y, final int z) {
+ return (x & 15) | (((z & 15) << 4)) | ((y & 255) << (4 + 4));
+ }
+
+ public static BlockState getBlockDataFromRaw(final long raw) {
2021-06-17 23:39:36 +02:00
+ return GLOBAL_PALETTE.valueFor((int)(raw >>> 32));
2021-06-11 14:02:28 +02:00
+ }
+
+ public static int getIndexFromRaw(final long raw) {
+ return (int)(raw & 0xFFFF);
+ }
+
+ public static int getLocationFromRaw(final long raw) {
+ return (int)((raw >>> 16) & 0xFFFF);
+ }
+
+ public static long getRawFromValues(final int index, final int location, final BlockState data) {
2021-06-17 23:39:36 +02:00
+ return (long)index | ((long)location << 16) | (((long)GLOBAL_PALETTE.idFor(data)) << 32);
2021-06-11 14:02:28 +02:00
+ }
+
+ public static long setIndexRawValues(final long value, final int index) {
+ return value & ~(0xFFFF) | (index);
+ }
+
+ public long add(final int x, final int y, final int z, final BlockState data) {
+ return this.add(getLocationKey(x, y, z), data);
+ }
+
+ public long add(final int location, final BlockState data) {
+ final long curr = this.map.get((short)location);
+
+ if (curr == Long.MAX_VALUE) {
+ final int index = this.size++;
+ final long raw = getRawFromValues(index, location, data);
+ this.map.put((short)location, raw);
+
+ if (index >= this.byIndex.length) {
+ this.byIndex = Arrays.copyOf(this.byIndex, (int)Math.max(4L, this.byIndex.length * 2L));
+ }
+
+ this.byIndex[index] = raw;
+ return raw;
+ } else {
+ final int index = getIndexFromRaw(curr);
+ final long raw = this.byIndex[index] = getRawFromValues(index, location, data);
+
+ this.map.put((short)location, raw);
+
+ return raw;
+ }
+ }
+
+ public long remove(final int x, final int y, final int z) {
+ return this.remove(getLocationKey(x, y, z));
+ }
+
+ public long remove(final int location) {
+ final long ret = this.map.remove((short)location);
+ final int index = getIndexFromRaw(ret);
+ if (ret == Long.MAX_VALUE) {
+ return ret;
+ }
+
+ // move the entry at the end to this index
+ final int endIndex = --this.size;
+ final long end = this.byIndex[endIndex];
+ if (index != endIndex) {
+ // not empty after this call
+ this.map.put((short)getLocationFromRaw(end), setIndexRawValues(end, index));
+ }
+ this.byIndex[index] = end;
+ this.byIndex[endIndex] = 0L;
+
+ return ret;
+ }
+
+ public int size() {
+ return this.size;
+ }
+
+ public long getRaw(final int index) {
+ return this.byIndex[index];
+ }
+
+ public int getLocation(final int index) {
+ return getLocationFromRaw(this.getRaw(index));
+ }
+
+ public BlockState getData(final int index) {
+ return getBlockDataFromRaw(this.getRaw(index));
+ }
+
+ public void clear() {
+ this.size = 0;
+ this.map.clear();
+ }
+
+ public LongIterator getRawIterator() {
+ return this.map.values().iterator();
+ }
+}
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
diff --git a/src/main/java/com/destroystokyo/paper/util/maplist/ReferenceList.java b/src/main/java/com/destroystokyo/paper/util/maplist/ReferenceList.java
new file mode 100644
index 0000000000000000000000000000000000000000..190c5f0b02a3d99054704ae1afbffb3498ddffe1
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/maplist/ReferenceList.java
@@ -0,0 +1,125 @@
+package com.destroystokyo.paper.util.maplist;
+
+import it.unimi.dsi.fastutil.objects.Reference2IntOpenHashMap;
+import java.util.Arrays;
+import java.util.Iterator;
+import java.util.NoSuchElementException;
+
+/**
+ * @author Spottedleaf
+ */
+public final class ReferenceList<E> implements Iterable<E> {
+
+ protected final Reference2IntOpenHashMap<E> referenceToIndex = new Reference2IntOpenHashMap<>(2, 0.8f);
+ {
+ this.referenceToIndex.defaultReturnValue(Integer.MIN_VALUE);
+ }
+
+ protected static final Object[] EMPTY_LIST = new Object[0];
+
+ protected Object[] references = EMPTY_LIST;
+ protected int count;
+
+ public int size() {
+ return this.count;
+ }
+
+ public boolean contains(final E obj) {
+ return this.referenceToIndex.containsKey(obj);
+ }
+
+ public boolean remove(final E obj) {
+ final int index = this.referenceToIndex.removeInt(obj);
+ if (index == Integer.MIN_VALUE) {
+ return false;
+ }
+
+ // move the object at the end to this index
+ final int endIndex = --this.count;
+ final E end = (E)this.references[endIndex];
+ if (index != endIndex) {
+ // not empty after this call
+ this.referenceToIndex.put(end, index); // update index
+ }
+ this.references[index] = end;
+ this.references[endIndex] = null;
+
+ return true;
+ }
+
+ public boolean add(final E obj) {
+ final int count = this.count;
+ final int currIndex = this.referenceToIndex.putIfAbsent(obj, count);
+
+ if (currIndex != Integer.MIN_VALUE) {
+ return false; // already in this list
+ }
+
+ Object[] list = this.references;
+
+ if (list.length == count) {
+ // resize required
+ list = this.references = Arrays.copyOf(list, (int)Math.max(4L, count * 2L)); // overflow results in negative
+ }
+
+ list[count] = obj;
+ this.count = count + 1;
+
+ return true;
+ }
+
+ public E getChecked(final int index) {
+ if (index < 0 || index >= this.count) {
+ throw new IndexOutOfBoundsException("Index: " + index + " is out of bounds, size: " + this.count);
+ }
+ return (E)this.references[index];
+ }
+
+ public E getUnchecked(final int index) {
+ return (E)this.references[index];
+ }
+
+ public Object[] getRawData() {
+ return this.references;
+ }
+
+ public void clear() {
+ this.referenceToIndex.clear();
+ Arrays.fill(this.references, 0, this.count, null);
+ this.count = 0;
+ }
+
+ @Override
+ public Iterator<E> iterator() {
+ return new Iterator<>() {
+ private E lastRet;
+ private int current;
+
+ @Override
+ public boolean hasNext() {
+ return this.current < ReferenceList.this.count;
+ }
+
+ @Override
+ public E next() {
+ if (this.current >= ReferenceList.this.count) {
+ throw new NoSuchElementException();
+ }
+ return this.lastRet = (E)ReferenceList.this.references[this.current++];
+ }
+
+ @Override
+ public void remove() {
+ final E lastRet = this.lastRet;
+
+ if (lastRet == null) {
+ throw new IllegalStateException();
+ }
+ this.lastRet = null;
+
+ ReferenceList.this.remove(lastRet);
+ --this.current;
+ }
+ };
+ }
+}
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/com/destroystokyo/paper/util/misc/AreaMap.java b/src/main/java/com/destroystokyo/paper/util/misc/AreaMap.java
new file mode 100644
index 0000000000000000000000000000000000000000..41b9405d6759d865e0d14dd4f95163e9690e967d
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/misc/AreaMap.java
@@ -0,0 +1,453 @@
+package com.destroystokyo.paper.util.misc;
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+import io.papermc.paper.util.IntegerUtil;
2021-06-11 14:02:28 +02:00
+import it.unimi.dsi.fastutil.longs.Long2ObjectLinkedOpenHashMap;
+import it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap;
+import it.unimi.dsi.fastutil.objects.Object2IntOpenHashMap;
+import it.unimi.dsi.fastutil.objects.Object2LongOpenHashMap;
+import io.papermc.paper.util.MCUtil;
2021-06-11 14:02:28 +02:00
+import net.minecraft.server.MinecraftServer;
+import net.minecraft.world.level.ChunkPos;
+import javax.annotation.Nullable;
+import java.util.Iterator;
+
+/** @author Spottedleaf */
+public abstract class AreaMap<E> {
+
+ /* Tested via https://gist.github.com/Spottedleaf/520419c6f41ef348fe9926ce674b7217 */
+
+ protected final Object2LongOpenHashMap<E> objectToLastCoordinate = new Object2LongOpenHashMap<>();
+ protected final Object2IntOpenHashMap<E> objectToViewDistance = new Object2IntOpenHashMap<>();
+
+ {
+ this.objectToViewDistance.defaultReturnValue(-1);
+ this.objectToLastCoordinate.defaultReturnValue(Long.MIN_VALUE);
+ }
+
+ // we use linked for better iteration.
+ // map of: coordinate to set of objects in coordinate
+ protected final Long2ObjectOpenHashMap<PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E>> areaMap = new Long2ObjectOpenHashMap<>(1024, 0.7f);
+ protected final PooledLinkedHashSets<E> pooledHashSets;
+
+ protected final ChangeCallback<E> addCallback;
+ protected final ChangeCallback<E> removeCallback;
+ protected final ChangeSourceCallback<E> changeSourceCallback;
+
+ public AreaMap() {
+ this(new PooledLinkedHashSets<>());
+ }
+
+ // let users define a "global" or "shared" pooled sets if they wish
+ public AreaMap(final PooledLinkedHashSets<E> pooledHashSets) {
+ this(pooledHashSets, null, null);
+ }
+
+ public AreaMap(final PooledLinkedHashSets<E> pooledHashSets, final ChangeCallback<E> addCallback, final ChangeCallback<E> removeCallback) {
+ this(pooledHashSets, addCallback, removeCallback, null);
+ }
+ public AreaMap(final PooledLinkedHashSets<E> pooledHashSets, final ChangeCallback<E> addCallback, final ChangeCallback<E> removeCallback, final ChangeSourceCallback<E> changeSourceCallback) {
+ this.pooledHashSets = pooledHashSets;
+ this.addCallback = addCallback;
+ this.removeCallback = removeCallback;
+ this.changeSourceCallback = changeSourceCallback;
+ }
+
+ @Nullable
+ public final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> getObjectsInRange(final long key) {
+ return this.areaMap.get(key);
+ }
+
+ @Nullable
+ public final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> getObjectsInRange(final ChunkPos chunkPos) {
+ return this.areaMap.get(MCUtil.getCoordinateKey(chunkPos));
+ }
+
+ @Nullable
+ public final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> getObjectsInRange(final int chunkX, final int chunkZ) {
+ return this.areaMap.get(MCUtil.getCoordinateKey(chunkX, chunkZ));
+ }
+
+ // Long.MIN_VALUE indicates the object is not mapped
+ public final long getLastCoordinate(final E object) {
+ return this.objectToLastCoordinate.getOrDefault(object, Long.MIN_VALUE);
+ }
+
+ // -1 indicates the object is not mapped
+ public final int getLastViewDistance(final E object) {
+ return this.objectToViewDistance.getOrDefault(object, -1);
+ }
+
+ // returns the total number of mapped chunks
+ public final int size() {
+ return this.areaMap.size();
+ }
+
+ public final void addOrUpdate(final E object, final int chunkX, final int chunkZ, final int viewDistance) {
+ final int oldViewDistance = this.objectToViewDistance.put(object, viewDistance);
+ final long newPos = MCUtil.getCoordinateKey(chunkX, chunkZ);
+ final long oldPos = this.objectToLastCoordinate.put(object, newPos);
+
+ if (oldViewDistance == -1) {
+ this.addObject(object, chunkX, chunkZ, Integer.MIN_VALUE, Integer.MIN_VALUE, viewDistance);
+ this.addObjectCallback(object, chunkX, chunkZ, viewDistance);
+ } else {
+ this.updateObject(object, oldPos, newPos, oldViewDistance, viewDistance);
+ this.updateObjectCallback(object, oldPos, newPos, oldViewDistance, viewDistance);
+ }
+ //this.validate(object, viewDistance);
+ }
+
+ public final boolean update(final E object, final int chunkX, final int chunkZ, final int viewDistance) {
+ final int oldViewDistance = this.objectToViewDistance.replace(object, viewDistance);
+ if (oldViewDistance == -1) {
+ return false;
+ } else {
+ final long newPos = MCUtil.getCoordinateKey(chunkX, chunkZ);
+ final long oldPos = this.objectToLastCoordinate.put(object, newPos);
+ this.updateObject(object, oldPos, newPos, oldViewDistance, viewDistance);
+ this.updateObjectCallback(object, oldPos, newPos, oldViewDistance, viewDistance);
+ }
+ //this.validate(object, viewDistance);
+ return true;
+ }
+
+ // called after the distance map updates
+ protected void updateObjectCallback(final E Object, final long oldPosition, final long newPosition, final int oldViewDistance, final int newViewDistance) {
+ if (newPosition != oldPosition && this.changeSourceCallback != null) {
+ this.changeSourceCallback.accept(Object, oldPosition, newPosition);
+ }
+ }
+
+ public final boolean add(final E object, final int chunkX, final int chunkZ, final int viewDistance) {
+ final int oldViewDistance = this.objectToViewDistance.putIfAbsent(object, viewDistance);
+ if (oldViewDistance != -1) {
+ return false;
+ }
+
+ final long newPos = MCUtil.getCoordinateKey(chunkX, chunkZ);
+ this.objectToLastCoordinate.put(object, newPos);
+ this.addObject(object, chunkX, chunkZ, Integer.MIN_VALUE, Integer.MIN_VALUE, viewDistance);
+ this.addObjectCallback(object, chunkX, chunkZ, viewDistance);
+
+ //this.validate(object, viewDistance);
+
+ return true;
+ }
+
+ // called after the distance map updates
+ protected void addObjectCallback(final E object, final int chunkX, final int chunkZ, final int viewDistance) {}
+
+ public final boolean remove(final E object) {
+ final long position = this.objectToLastCoordinate.removeLong(object);
+ final int viewDistance = this.objectToViewDistance.removeInt(object);
+
+ if (viewDistance == -1) {
+ return false;
+ }
+
+ final int currentX = MCUtil.getCoordinateX(position);
+ final int currentZ = MCUtil.getCoordinateZ(position);
+
+ this.removeObject(object, currentX, currentZ, currentX, currentZ, viewDistance);
+ this.removeObjectCallback(object, currentX, currentZ, viewDistance);
+ //this.validate(object, -1);
+ return true;
+ }
+
+ // called after the distance map updates
+ protected void removeObjectCallback(final E object, final int chunkX, final int chunkZ, final int viewDistance) {}
+
+ protected abstract PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> getEmptySetFor(final E object);
+
+ // expensive op, only for debug
+ protected void validate(final E object, final int viewDistance) {
+ int entiesGot = 0;
+ int expectedEntries = (2 * viewDistance + 1);
+ expectedEntries *= expectedEntries;
+ if (viewDistance < 0) {
+ expectedEntries = 0;
+ }
+
+ final long currPosition = this.objectToLastCoordinate.getLong(object);
+
+ final int centerX = MCUtil.getCoordinateX(currPosition);
+ final int centerZ = MCUtil.getCoordinateZ(currPosition);
+
+ for (Iterator<Long2ObjectLinkedOpenHashMap.Entry<PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E>>> iterator = this.areaMap.long2ObjectEntrySet().fastIterator();
+ iterator.hasNext();) {
+
+ final Long2ObjectLinkedOpenHashMap.Entry<PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E>> entry = iterator.next();
+ final long key = entry.getLongKey();
+ final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> map = entry.getValue();
+
+ if (map.referenceCount == 0) {
+ throw new IllegalStateException("Invalid map");
+ }
+
+ if (map.contains(object)) {
+ ++entiesGot;
+
+ final int chunkX = MCUtil.getCoordinateX(key);
+ final int chunkZ = MCUtil.getCoordinateZ(key);
+
+ final int dist = Math.max(IntegerUtil.branchlessAbs(chunkX - centerX), IntegerUtil.branchlessAbs(chunkZ - centerZ));
+
+ if (dist > viewDistance) {
+ throw new IllegalStateException("Expected view distance " + viewDistance + ", got " + dist);
+ }
+ }
+ }
+
+ if (entiesGot != expectedEntries) {
+ throw new IllegalStateException("Expected " + expectedEntries + ", got " + entiesGot);
+ }
+ }
+
+ private void addObjectTo(final E object, final int chunkX, final int chunkZ, final int currChunkX,
+ final int currChunkZ, final int prevChunkX, final int prevChunkZ) {
+ final long key = MCUtil.getCoordinateKey(chunkX, chunkZ);
+
+ PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> empty = this.getEmptySetFor(object);
+ PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> current = this.areaMap.putIfAbsent(key, empty);
+
+ if (current != null) {
+ PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> next = this.pooledHashSets.findMapWith(current, object);
+ if (next == current) {
+ throw new IllegalStateException("Expected different map: got " + next.toString());
+ }
+ this.areaMap.put(key, next);
+
+ current = next;
+ // fall through to callback
+ } else {
+ current = empty;
+ }
+
+ if (this.addCallback != null) {
+ try {
+ this.addCallback.accept(object, chunkX, chunkZ, currChunkX, currChunkZ, prevChunkX, prevChunkZ, current);
+ } catch (final Throwable ex) {
+ if (ex instanceof ThreadDeath) {
+ throw (ThreadDeath)ex;
+ }
+ MinecraftServer.LOGGER.error("Add callback for map threw exception ", ex);
+ }
+ }
+ }
+
+ private void removeObjectFrom(final E object, final int chunkX, final int chunkZ, final int currChunkX,
+ final int currChunkZ, final int prevChunkX, final int prevChunkZ) {
+ final long key = MCUtil.getCoordinateKey(chunkX, chunkZ);
+
+ PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> current = this.areaMap.get(key);
+
+ if (current == null) {
+ throw new IllegalStateException("Current map may not be null for " + object + ", (" + chunkX + "," + chunkZ + ")");
+ }
+
+ PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> next = this.pooledHashSets.findMapWithout(current, object);
+
+ if (next == current) {
+ throw new IllegalStateException("Current map [" + next.toString() + "] should have contained " + object + ", (" + chunkX + "," + chunkZ + ")");
+ }
+
+ if (next != null) {
+ this.areaMap.put(key, next);
+ } else {
+ this.areaMap.remove(key);
+ }
+
+ if (this.removeCallback != null) {
+ try {
+ this.removeCallback.accept(object, chunkX, chunkZ, currChunkX, currChunkZ, prevChunkX, prevChunkZ, next);
+ } catch (final Throwable ex) {
+ if (ex instanceof ThreadDeath) {
+ throw (ThreadDeath)ex;
+ }
+ MinecraftServer.LOGGER.error("Remove callback for map threw exception ", ex);
+ }
+ }
+ }
+
+ private void addObject(final E object, final int chunkX, final int chunkZ, final int prevChunkX, final int prevChunkZ, final int viewDistance) {
+ final int maxX = chunkX + viewDistance;
+ final int maxZ = chunkZ + viewDistance;
+ final int minX = chunkX - viewDistance;
+ final int minZ = chunkZ - viewDistance;
+ for (int x = minX; x <= maxX; ++x) {
+ for (int z = minZ; z <= maxZ; ++z) {
+ this.addObjectTo(object, x, z, chunkX, chunkZ, prevChunkX, prevChunkZ);
+ }
+ }
+ }
+
+ private void removeObject(final E object, final int chunkX, final int chunkZ, final int currentChunkX, final int currentChunkZ, final int viewDistance) {
+ final int maxX = chunkX + viewDistance;
+ final int maxZ = chunkZ + viewDistance;
+ final int minX = chunkX - viewDistance;
+ final int minZ = chunkZ - viewDistance;
+ for (int x = minX; x <= maxX; ++x) {
+ for (int z = minZ; z <= maxZ; ++z) {
+ this.removeObjectFrom(object, x, z, currentChunkX, currentChunkZ, chunkX, chunkZ);
+ }
+ }
+ }
+
+ /* math sign function except 0 returns 1 */
+ protected static int sign(int val) {
+ return 1 | (val >> (Integer.SIZE - 1));
+ }
+
+ private void updateObject(final E object, final long oldPosition, final long newPosition, final int oldViewDistance, final int newViewDistance) {
+ final int toX = MCUtil.getCoordinateX(newPosition);
+ final int toZ = MCUtil.getCoordinateZ(newPosition);
+ final int fromX = MCUtil.getCoordinateX(oldPosition);
+ final int fromZ = MCUtil.getCoordinateZ(oldPosition);
+
+ final int dx = toX - fromX;
+ final int dz = toZ - fromZ;
+
+ final int totalX = IntegerUtil.branchlessAbs(fromX - toX);
+ final int totalZ = IntegerUtil.branchlessAbs(fromZ - toZ);
+
+ if (Math.max(totalX, totalZ) > (2 * Math.max(newViewDistance, oldViewDistance))) {
+ // teleported?
+ this.removeObject(object, fromX, fromZ, fromX, fromZ, oldViewDistance);
+ this.addObject(object, toX, toZ, fromX, fromZ, newViewDistance);
+ return;
+ }
+
+ if (oldViewDistance != newViewDistance) {
+ // remove loop
+
+ final int oldMinX = fromX - oldViewDistance;
+ final int oldMinZ = fromZ - oldViewDistance;
+ final int oldMaxX = fromX + oldViewDistance;
+ final int oldMaxZ = fromZ + oldViewDistance;
+ for (int currX = oldMinX; currX <= oldMaxX; ++currX) {
+ for (int currZ = oldMinZ; currZ <= oldMaxZ; ++currZ) {
+
+ // only remove if we're outside the new view distance...
+ if (Math.max(IntegerUtil.branchlessAbs(currX - toX), IntegerUtil.branchlessAbs(currZ - toZ)) > newViewDistance) {
+ this.removeObjectFrom(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+
+ // add loop
+
+ final int newMinX = toX - newViewDistance;
+ final int newMinZ = toZ - newViewDistance;
+ final int newMaxX = toX + newViewDistance;
+ final int newMaxZ = toZ + newViewDistance;
+ for (int currX = newMinX; currX <= newMaxX; ++currX) {
+ for (int currZ = newMinZ; currZ <= newMaxZ; ++currZ) {
+
+ // only add if we're outside the old view distance...
+ if (Math.max(IntegerUtil.branchlessAbs(currX - fromX), IntegerUtil.branchlessAbs(currZ - fromZ)) > oldViewDistance) {
+ this.addObjectTo(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+
+ return;
+ }
+
+ // x axis is width
+ // z axis is height
+ // right refers to the x axis of where we moved
+ // top refers to the z axis of where we moved
+
+ // same view distance
+
+ // used for relative positioning
+ final int up = sign(dz); // 1 if dz >= 0, -1 otherwise
+ final int right = sign(dx); // 1 if dx >= 0, -1 otherwise
+
+ // The area excluded by overlapping the two view distance squares creates four rectangles:
+ // Two on the left, and two on the right. The ones on the left we consider the "removed" section
+ // and on the right the "added" section.
+ // https://i.imgur.com/MrnOBgI.png is a reference image. Note that the outside border is not actually
+ // exclusive to the regions they surround.
+
+ // 4 points of the rectangle
+ int maxX; // exclusive
+ int minX; // inclusive
+ int maxZ; // exclusive
+ int minZ; // inclusive
+
+ if (dx != 0) {
+ // handle right addition
+
+ maxX = toX + (oldViewDistance * right) + right; // exclusive
+ minX = fromX + (oldViewDistance * right) + right; // inclusive
+ maxZ = fromZ + (oldViewDistance * up) + up; // exclusive
+ minZ = toZ - (oldViewDistance * up); // inclusive
+
+ for (int currX = minX; currX != maxX; currX += right) {
+ for (int currZ = minZ; currZ != maxZ; currZ += up) {
+ this.addObjectTo(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+
+ if (dz != 0) {
+ // handle up addition
+
+ maxX = toX + (oldViewDistance * right) + right; // exclusive
+ minX = toX - (oldViewDistance * right); // inclusive
+ maxZ = toZ + (oldViewDistance * up) + up; // exclusive
+ minZ = fromZ + (oldViewDistance * up) + up; // inclusive
+
+ for (int currX = minX; currX != maxX; currX += right) {
+ for (int currZ = minZ; currZ != maxZ; currZ += up) {
+ this.addObjectTo(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+
+ if (dx != 0) {
+ // handle left removal
+
+ maxX = toX - (oldViewDistance * right); // exclusive
+ minX = fromX - (oldViewDistance * right); // inclusive
+ maxZ = fromZ + (oldViewDistance * up) + up; // exclusive
+ minZ = toZ - (oldViewDistance * up); // inclusive
+
+ for (int currX = minX; currX != maxX; currX += right) {
+ for (int currZ = minZ; currZ != maxZ; currZ += up) {
+ this.removeObjectFrom(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+
+ if (dz != 0) {
+ // handle down removal
+
+ maxX = fromX + (oldViewDistance * right) + right; // exclusive
+ minX = fromX - (oldViewDistance * right); // inclusive
+ maxZ = toZ - (oldViewDistance * up); // exclusive
+ minZ = fromZ - (oldViewDistance * up); // inclusive
+
+ for (int currX = minX; currX != maxX; currX += right) {
+ for (int currZ = minZ; currZ != maxZ; currZ += up) {
+ this.removeObjectFrom(object, currX, currZ, toX, toZ, fromX, fromZ);
+ }
+ }
+ }
+ }
+
+ @FunctionalInterface
+ public static interface ChangeCallback<E> {
+
+ // if there is no previous position, then prevPos = Integer.MIN_VALUE
+ void accept(final E object, final int rangeX, final int rangeZ, final int currPosX, final int currPosZ, final int prevPosX, final int prevPosZ,
+ final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> newState);
+
+ }
+
+ @FunctionalInterface
+ public static interface ChangeSourceCallback<E> {
+ void accept(final E object, final long prevPos, final long newPos);
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/misc/DistanceTrackingAreaMap.java b/src/main/java/com/destroystokyo/paper/util/misc/DistanceTrackingAreaMap.java
new file mode 100644
index 0000000000000000000000000000000000000000..896c3ff7ddb07f1f6f05f90e1e3fe7fb615071d4
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/misc/DistanceTrackingAreaMap.java
@@ -0,0 +1,175 @@
+package com.destroystokyo.paper.util.misc;
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+import io.papermc.paper.util.IntegerUtil;
2021-06-11 14:02:28 +02:00
+import it.unimi.dsi.fastutil.longs.Long2IntOpenHashMap;
+import io.papermc.paper.util.MCUtil;
2021-06-11 14:02:28 +02:00
+import net.minecraft.world.level.ChunkPos;
+
+/** @author Spottedleaf */
+public abstract class DistanceTrackingAreaMap<E> extends AreaMap<E> {
+
+ // use this map only if you need distance tracking, the tracking here is obviously going to hit harder.
+
+ protected final Long2IntOpenHashMap chunkToNearestDistance = new Long2IntOpenHashMap(1024, 0.7f);
+ {
+ this.chunkToNearestDistance.defaultReturnValue(-1);
+ }
+
+ protected final DistanceChangeCallback<E> distanceChangeCallback;
+
+ public DistanceTrackingAreaMap() {
+ this(new PooledLinkedHashSets<>());
+ }
+
+ // let users define a "global" or "shared" pooled sets if they wish
+ public DistanceTrackingAreaMap(final PooledLinkedHashSets<E> pooledHashSets) {
+ this(pooledHashSets, null, null, null);
+ }
+
+ public DistanceTrackingAreaMap(final PooledLinkedHashSets<E> pooledHashSets, final ChangeCallback<E> addCallback, final ChangeCallback<E> removeCallback,
+ final DistanceChangeCallback<E> distanceChangeCallback) {
+ super(pooledHashSets, addCallback, removeCallback);
+ this.distanceChangeCallback = distanceChangeCallback;
+ }
+
+ // ret -1 if there is nothing mapped
+ public final int getNearestObjectDistance(final long key) {
+ return this.chunkToNearestDistance.get(key);
+ }
+
+ // ret -1 if there is nothing mapped
+ public final int getNearestObjectDistance(final ChunkPos chunkPos) {
+ return this.chunkToNearestDistance.get(MCUtil.getCoordinateKey(chunkPos));
+ }
+
+ // ret -1 if there is nothing mapped
+ public final int getNearestObjectDistance(final int chunkX, final int chunkZ) {
+ return this.chunkToNearestDistance.get(MCUtil.getCoordinateKey(chunkX, chunkZ));
+ }
+
+ protected final void recalculateDistance(final int chunkX, final int chunkZ) {
+ final long key = MCUtil.getCoordinateKey(chunkX, chunkZ);
+ final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> state = this.areaMap.get(key);
+ if (state == null) {
+ final int oldDistance = this.chunkToNearestDistance.remove(key);
+ // nothing here.
+ if (oldDistance == -1) {
+ // nothing was here previously
+ return;
+ }
+ if (this.distanceChangeCallback != null) {
+ this.distanceChangeCallback.accept(chunkX, chunkZ, oldDistance, -1, null);
+ }
+ return;
+ }
+
+ int newDistance = Integer.MAX_VALUE;
+
+ final Object[] rawData = state.getBackingSet();
+ for (int i = 0, len = rawData.length; i < len; ++i) {
+ final Object raw = rawData[i];
+
+ if (raw == null) {
+ continue;
+ }
+
+ final E object = (E)raw;
+ final long location = this.objectToLastCoordinate.getLong(object);
+
+ final int distance = Math.max(IntegerUtil.branchlessAbs(chunkX - MCUtil.getCoordinateX(location)), IntegerUtil.branchlessAbs(chunkZ - MCUtil.getCoordinateZ(location)));
+
+ if (distance < newDistance) {
+ newDistance = distance;
+ }
+ }
+
+ final int oldDistance = this.chunkToNearestDistance.put(key, newDistance);
+
+ if (oldDistance != newDistance) {
+ if (this.distanceChangeCallback != null) {
+ this.distanceChangeCallback.accept(chunkX, chunkZ, oldDistance, newDistance, state);
+ }
+ }
+ }
+
+ @Override
+ protected void addObjectCallback(final E object, final int chunkX, final int chunkZ, final int viewDistance) {
+ final int maxX = chunkX + viewDistance;
+ final int maxZ = chunkZ + viewDistance;
+ final int minX = chunkX - viewDistance;
+ final int minZ = chunkZ - viewDistance;
+ for (int x = minX; x <= maxX; ++x) {
+ for (int z = minZ; z <= maxZ; ++z) {
+ this.recalculateDistance(x, z);
+ }
+ }
+ }
+
+ @Override
+ protected void removeObjectCallback(final E object, final int chunkX, final int chunkZ, final int viewDistance) {
+ final int maxX = chunkX + viewDistance;
+ final int maxZ = chunkZ + viewDistance;
+ final int minX = chunkX - viewDistance;
+ final int minZ = chunkZ - viewDistance;
+ for (int x = minX; x <= maxX; ++x) {
+ for (int z = minZ; z <= maxZ; ++z) {
+ this.recalculateDistance(x, z);
+ }
+ }
+ }
+
+ @Override
+ protected void updateObjectCallback(final E object, final long oldPosition, final long newPosition, final int oldViewDistance, final int newViewDistance) {
+ if (oldPosition == newPosition && newViewDistance == oldViewDistance) {
+ return;
+ }
+
+ final int toX = MCUtil.getCoordinateX(newPosition);
+ final int toZ = MCUtil.getCoordinateZ(newPosition);
+ final int fromX = MCUtil.getCoordinateX(oldPosition);
+ final int fromZ = MCUtil.getCoordinateZ(oldPosition);
+
+ final int totalX = IntegerUtil.branchlessAbs(fromX - toX);
+ final int totalZ = IntegerUtil.branchlessAbs(fromZ - toZ);
+
+ if (Math.max(totalX, totalZ) > (2 * Math.max(newViewDistance, oldViewDistance))) {
+ // teleported?
+ this.removeObjectCallback(object, fromX, fromZ, oldViewDistance);
+ this.addObjectCallback(object, toX, toZ, newViewDistance);
+ return;
+ }
+
+ final int minX = Math.min(fromX - oldViewDistance, toX - newViewDistance);
+ final int maxX = Math.max(fromX + oldViewDistance, toX + newViewDistance);
+ final int minZ = Math.min(fromZ - oldViewDistance, toZ - newViewDistance);
+ final int maxZ = Math.max(fromZ + oldViewDistance, toZ + newViewDistance);
+
+ for (int x = minX; x <= maxX; ++x) {
+ for (int z = minZ; z <= maxZ; ++z) {
+ final int distXOld = IntegerUtil.branchlessAbs(x - fromX);
+ final int distZOld = IntegerUtil.branchlessAbs(z - fromZ);
+
+ if (Math.max(distXOld, distZOld) <= oldViewDistance) {
+ this.recalculateDistance(x, z);
+ continue;
+ }
+
+ final int distXNew = IntegerUtil.branchlessAbs(x - toX);
+ final int distZNew = IntegerUtil.branchlessAbs(z - toZ);
+
+ if (Math.max(distXNew, distZNew) <= newViewDistance) {
+ this.recalculateDistance(x, z);
+ continue;
+ }
+ }
+ }
+ }
+
+ @FunctionalInterface
+ public static interface DistanceChangeCallback<E> {
+
+ void accept(final int posX, final int posZ, final int oldNearestDistance, final int newNearestDistance,
+ final PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<E> state);
+
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/misc/PlayerAreaMap.java b/src/main/java/com/destroystokyo/paper/util/misc/PlayerAreaMap.java
new file mode 100644
index 0000000000000000000000000000000000000000..46954db7ecd35ac4018fdf476df7c8020d7ce6c8
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/misc/PlayerAreaMap.java
@@ -0,0 +1,32 @@
+package com.destroystokyo.paper.util.misc;
+
+import net.minecraft.server.level.ServerPlayer;
+
+/**
+ * @author Spottedleaf
+ */
+public final class PlayerAreaMap extends AreaMap<ServerPlayer> {
+
+ public PlayerAreaMap() {
+ super();
+ }
+
+ public PlayerAreaMap(final PooledLinkedHashSets<ServerPlayer> pooledHashSets) {
+ super(pooledHashSets);
+ }
+
+ public PlayerAreaMap(final PooledLinkedHashSets<ServerPlayer> pooledHashSets, final ChangeCallback<ServerPlayer> addCallback,
+ final ChangeCallback<ServerPlayer> removeCallback) {
+ this(pooledHashSets, addCallback, removeCallback, null);
+ }
+
+ public PlayerAreaMap(final PooledLinkedHashSets<ServerPlayer> pooledHashSets, final ChangeCallback<ServerPlayer> addCallback,
+ final ChangeCallback<ServerPlayer> removeCallback, final ChangeSourceCallback<ServerPlayer> changeSourceCallback) {
+ super(pooledHashSets, addCallback, removeCallback, changeSourceCallback);
+ }
+
+ @Override
+ protected PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<ServerPlayer> getEmptySetFor(final ServerPlayer player) {
+ return player.cachedSingleHashSet;
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/misc/PlayerDistanceTrackingAreaMap.java b/src/main/java/com/destroystokyo/paper/util/misc/PlayerDistanceTrackingAreaMap.java
new file mode 100644
index 0000000000000000000000000000000000000000..d05dcea15f7047b58736c7c0e07920a04d6c5abe
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/misc/PlayerDistanceTrackingAreaMap.java
@@ -0,0 +1,24 @@
+package com.destroystokyo.paper.util.misc;
+
+import net.minecraft.server.level.ServerPlayer;
+
+public class PlayerDistanceTrackingAreaMap extends DistanceTrackingAreaMap<ServerPlayer> {
+
+ public PlayerDistanceTrackingAreaMap() {
+ super();
+ }
+
+ public PlayerDistanceTrackingAreaMap(final PooledLinkedHashSets<ServerPlayer> pooledHashSets) {
+ super(pooledHashSets);
+ }
+
+ public PlayerDistanceTrackingAreaMap(final PooledLinkedHashSets<ServerPlayer> pooledHashSets, final ChangeCallback<ServerPlayer> addCallback,
+ final ChangeCallback<ServerPlayer> removeCallback, final DistanceChangeCallback<ServerPlayer> distanceChangeCallback) {
+ super(pooledHashSets, addCallback, removeCallback, distanceChangeCallback);
+ }
+
+ @Override
+ protected PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<ServerPlayer> getEmptySetFor(final ServerPlayer player) {
+ return player.cachedSingleHashSet;
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/misc/PooledLinkedHashSets.java b/src/main/java/com/destroystokyo/paper/util/misc/PooledLinkedHashSets.java
new file mode 100644
index 0000000000000000000000000000000000000000..e51104e65a07b6ea7bbbcbb6afb066ef6401cc5b
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/misc/PooledLinkedHashSets.java
@@ -0,0 +1,287 @@
+package com.destroystokyo.paper.util.misc;
+
+import it.unimi.dsi.fastutil.objects.Object2ObjectOpenHashMap;
+import it.unimi.dsi.fastutil.objects.ObjectOpenHashSet;
+import java.lang.ref.WeakReference;
+
+/** @author Spottedleaf */
+public class PooledLinkedHashSets<E> {
+
+ /* Tested via https://gist.github.com/Spottedleaf/a93bb7a8993d6ce142d3efc5932bf573 */
+
+ // we really want to avoid that equals() check as much as possible...
+ protected final Object2ObjectOpenHashMap<PooledObjectLinkedOpenHashSet<E>, PooledObjectLinkedOpenHashSet<E>> mapPool = new Object2ObjectOpenHashMap<>(128, 0.25f);
+
+ protected void decrementReferenceCount(final PooledObjectLinkedOpenHashSet<E> current) {
+ if (current.referenceCount == 0) {
+ throw new IllegalStateException("Cannot decrement reference count for " + current);
+ }
+ if (current.referenceCount == -1 || --current.referenceCount > 0) {
+ return;
+ }
+
+ this.mapPool.remove(current);
+ return;
+ }
+
+ public PooledObjectLinkedOpenHashSet<E> findMapWith(final PooledObjectLinkedOpenHashSet<E> current, final E object) {
+ final PooledObjectLinkedOpenHashSet<E> cached = current.getAddCache(object);
+
+ if (cached != null) {
+ decrementReferenceCount(current);
+
+ if (cached.referenceCount == 0) {
+ // bring the map back from the dead
+ PooledObjectLinkedOpenHashSet<E> contending = this.mapPool.putIfAbsent(cached, cached);
+ if (contending != null) {
+ // a map already exists with the elements we want
+ if (contending.referenceCount != -1) {
+ ++contending.referenceCount;
+ }
+ current.updateAddCache(object, contending);
+ return contending;
+ }
+
+ cached.referenceCount = 1;
+ } else if (cached.referenceCount != -1) {
+ ++cached.referenceCount;
+ }
+
+ return cached;
+ }
+
+ if (!current.add(object)) {
+ return current;
+ }
+
+ // we use get/put since we use a different key on put
+ PooledObjectLinkedOpenHashSet<E> ret = this.mapPool.get(current);
+
+ if (ret == null) {
+ ret = new PooledObjectLinkedOpenHashSet<>(current);
+ current.remove(object);
+ this.mapPool.put(ret, ret);
+ ret.referenceCount = 1;
+ } else {
+ if (ret.referenceCount != -1) {
+ ++ret.referenceCount;
+ }
+ current.remove(object);
+ }
+
+ current.updateAddCache(object, ret);
+
+ decrementReferenceCount(current);
+ return ret;
+ }
+
+ // rets null if current.size() == 1
+ public PooledObjectLinkedOpenHashSet<E> findMapWithout(final PooledObjectLinkedOpenHashSet<E> current, final E object) {
+ if (current.set.size() == 1) {
+ decrementReferenceCount(current);
+ return null;
+ }
+
+ final PooledObjectLinkedOpenHashSet<E> cached = current.getRemoveCache(object);
+
+ if (cached != null) {
+ decrementReferenceCount(current);
+
+ if (cached.referenceCount == 0) {
+ // bring the map back from the dead
+ PooledObjectLinkedOpenHashSet<E> contending = this.mapPool.putIfAbsent(cached, cached);
+ if (contending != null) {
+ // a map already exists with the elements we want
+ if (contending.referenceCount != -1) {
+ ++contending.referenceCount;
+ }
+ current.updateRemoveCache(object, contending);
+ return contending;
+ }
+
+ cached.referenceCount = 1;
+ } else if (cached.referenceCount != -1) {
+ ++cached.referenceCount;
+ }
+
+ return cached;
+ }
+
+ if (!current.remove(object)) {
+ return current;
+ }
+
+ // we use get/put since we use a different key on put
+ PooledObjectLinkedOpenHashSet<E> ret = this.mapPool.get(current);
+
+ if (ret == null) {
+ ret = new PooledObjectLinkedOpenHashSet<>(current);
+ current.add(object);
+ this.mapPool.put(ret, ret);
+ ret.referenceCount = 1;
+ } else {
+ if (ret.referenceCount != -1) {
+ ++ret.referenceCount;
+ }
+ current.add(object);
+ }
+
+ current.updateRemoveCache(object, ret);
+
+ decrementReferenceCount(current);
+ return ret;
+ }
+
+ static final class RawSetObjectLinkedOpenHashSet<E> extends ObjectOpenHashSet<E> {
+
+ public RawSetObjectLinkedOpenHashSet() {
+ super();
+ }
+
+ public RawSetObjectLinkedOpenHashSet(final int capacity) {
+ super(capacity);
+ }
+
+ public RawSetObjectLinkedOpenHashSet(final int capacity, final float loadFactor) {
+ super(capacity, loadFactor);
+ }
+
+ @Override
+ public RawSetObjectLinkedOpenHashSet<E> clone() {
+ return (RawSetObjectLinkedOpenHashSet<E>)super.clone();
+ }
+
+ public E[] getRawSet() {
+ return this.key;
+ }
+ }
+
+ public static final class PooledObjectLinkedOpenHashSet<E> {
+
+ private static final WeakReference NULL_REFERENCE = new WeakReference<>(null);
+
+ final RawSetObjectLinkedOpenHashSet<E> set;
+ int referenceCount; // -1 if special
+ int hash; // optimize hashcode
+
+ // add cache
+ WeakReference<E> lastAddObject = NULL_REFERENCE;
+ WeakReference<PooledObjectLinkedOpenHashSet<E>> lastAddMap = NULL_REFERENCE;
+
+ // remove cache
+ WeakReference<E> lastRemoveObject = NULL_REFERENCE;
+ WeakReference<PooledObjectLinkedOpenHashSet<E>> lastRemoveMap = NULL_REFERENCE;
+
+ public PooledObjectLinkedOpenHashSet(final PooledLinkedHashSets<E> pooledSets) {
+ this.set = new RawSetObjectLinkedOpenHashSet<>(2, 0.8f);
+ }
+
+ public PooledObjectLinkedOpenHashSet(final E single) {
+ this((PooledLinkedHashSets<E>)null);
+ this.referenceCount = -1;
+ this.add(single);
+ }
+
+ public PooledObjectLinkedOpenHashSet(final PooledObjectLinkedOpenHashSet<E> other) {
+ this.set = other.set.clone();
+ this.hash = other.hash;
+ }
+
+ // from https://github.com/Spottedleaf/ConcurrentUtil/blob/master/src/main/java/ca/spottedleaf/concurrentutil/util/IntegerUtil.java
+ // generated by https://github.com/skeeto/hash-prospector
+ private static int hash0(int x) {
+ x *= 0x36935555;
+ x ^= x >>> 16;
+ return x;
+ }
+
+ PooledObjectLinkedOpenHashSet<E> getAddCache(final E element) {
+ final E currentAdd = this.lastAddObject.get();
+
+ if (currentAdd == null || !(currentAdd == element || currentAdd.equals(element))) {
+ return null;
+ }
+
+ return this.lastAddMap.get();
+ }
+
+ PooledObjectLinkedOpenHashSet<E> getRemoveCache(final E element) {
+ final E currentRemove = this.lastRemoveObject.get();
+
+ if (currentRemove == null || !(currentRemove == element || currentRemove.equals(element))) {
+ return null;
+ }
+
+ return this.lastRemoveMap.get();
+ }
+
+ void updateAddCache(final E element, final PooledObjectLinkedOpenHashSet<E> map) {
+ this.lastAddObject = new WeakReference<>(element);
+ this.lastAddMap = new WeakReference<>(map);
+ }
+
+ void updateRemoveCache(final E element, final PooledObjectLinkedOpenHashSet<E> map) {
+ this.lastRemoveObject = new WeakReference<>(element);
+ this.lastRemoveMap = new WeakReference<>(map);
+ }
+
+ boolean add(final E element) {
+ boolean added = this.set.add(element);
+
+ if (added) {
+ this.hash += hash0(element.hashCode());
+ }
+
+ return added;
+ }
+
+ boolean remove(Object element) {
+ boolean removed = this.set.remove(element);
+
+ if (removed) {
+ this.hash -= hash0(element.hashCode());
+ }
+
+ return removed;
+ }
+
+ public boolean contains(final Object element) {
+ return this.set.contains(element);
+ }
+
+ public E[] getBackingSet() {
+ return this.set.getRawSet();
+ }
+
+ public int size() {
+ return this.set.size();
+ }
+
+ @Override
+ public int hashCode() {
+ return this.hash;
+ }
+
+ @Override
+ public boolean equals(final Object other) {
+ if (!(other instanceof PooledObjectLinkedOpenHashSet)) {
+ return false;
+ }
+ if (this.referenceCount == 0) {
+ return other == this;
+ } else {
+ if (other == this) {
+ // Unfortunately we are never equal to our own instance while in use!
+ return false;
+ }
+ return this.hash == ((PooledObjectLinkedOpenHashSet)other).hash && this.set.equals(((PooledObjectLinkedOpenHashSet)other).set);
+ }
+ }
+
+ @Override
+ public String toString() {
+ return "PooledHashSet: size: " + this.set.size() + ", reference count: " + this.referenceCount + ", hash: " +
+ this.hashCode() + ", identity: " + System.identityHashCode(this) + " map: " + this.set.toString();
+ }
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/pooled/PooledObjects.java b/src/main/java/com/destroystokyo/paper/util/pooled/PooledObjects.java
new file mode 100644
index 0000000000000000000000000000000000000000..a743703502cea333bd4231b6557de50e8eaf81eb
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/pooled/PooledObjects.java
@@ -0,0 +1,85 @@
+package com.destroystokyo.paper.util.pooled;
+
+import io.papermc.paper.util.MCUtil;
2021-06-11 14:02:28 +02:00
+import org.apache.commons.lang3.mutable.MutableInt;
+
+import java.util.ArrayDeque;
+import java.util.function.Consumer;
+import java.util.function.Supplier;
+
+public final class PooledObjects<E> {
+
+ /**
+ * Wrapper for an object that will be have a cleaner registered for it, and may be automatically returned to pool.
+ */
+ public class AutoReleased {
+ private final E object;
+ private final Runnable cleaner;
+
+ public AutoReleased(E object, Runnable cleaner) {
+ this.object = object;
+ this.cleaner = cleaner;
+ }
+
+ public final E getObject() {
+ return object;
+ }
+
+ public final Runnable getCleaner() {
+ return cleaner;
+ }
+ }
+
+ public static final PooledObjects<MutableInt> POOLED_MUTABLE_INTEGERS = new PooledObjects<>(MutableInt::new, 1024);
+
+ private final Supplier<E> creator;
+ private final Consumer<E> releaser;
+ private final int maxPoolSize;
+ private final ArrayDeque<E> queue;
+
+ public PooledObjects(final Supplier<E> creator, int maxPoolSize) {
+ this(creator, maxPoolSize, null);
+ }
+ public PooledObjects(final Supplier<E> creator, int maxPoolSize, Consumer<E> releaser) {
+ if (creator == null) {
+ throw new NullPointerException("Creator must not be null");
+ }
+ if (maxPoolSize <= 0) {
+ throw new IllegalArgumentException("Max pool size must be greater-than 0");
+ }
+
+ this.queue = new ArrayDeque<>(maxPoolSize);
+ this.maxPoolSize = maxPoolSize;
+ this.creator = creator;
+ this.releaser = releaser;
+ }
+
+ public AutoReleased acquireCleaner(Object holder) {
+ return acquireCleaner(holder, this::release);
+ }
+
+ public AutoReleased acquireCleaner(Object holder, Consumer<E> releaser) {
+ E resource = acquire();
+ Runnable cleaner = MCUtil.registerCleaner(holder, resource, releaser);
+ return new AutoReleased(resource, cleaner);
+ }
+
+ public final E acquire() {
+ E value;
+ synchronized (queue) {
+ value = this.queue.pollLast();
+ }
+ return value != null ? value : this.creator.get();
+ }
+
+ public final void release(final E value) {
+ if (this.releaser != null) {
+ this.releaser.accept(value);
+ }
+ synchronized (this.queue) {
+ if (queue.size() < this.maxPoolSize) {
+ this.queue.addLast(value);
+ }
+ }
+ }
+}
diff --git a/src/main/java/com/destroystokyo/paper/util/set/OptimizedSmallEnumSet.java b/src/main/java/com/destroystokyo/paper/util/set/OptimizedSmallEnumSet.java
new file mode 100644
2021-11-30 09:36:30 +01:00
index 0000000000000000000000000000000000000000..b3329c6fcd6758a781a51f5ba8f5052ac1c77b49
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/com/destroystokyo/paper/util/set/OptimizedSmallEnumSet.java
2021-11-30 09:36:30 +01:00
@@ -0,0 +1,71 @@
2021-06-11 14:02:28 +02:00
+package com.destroystokyo.paper.util.set;
+
+import java.util.Collection;
+
+/**
+ * @author Spottedleaf <Spottedleaf@users.noreply.github.com>
+ */
+public final class OptimizedSmallEnumSet<E extends Enum<E>> {
+
+ private final Class<E> enumClass;
+ private long backingSet;
+
+ public OptimizedSmallEnumSet(final Class<E> clazz) {
+ if (clazz == null) {
+ throw new IllegalArgumentException("Null class");
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ if (!clazz.isEnum()) {
+ throw new IllegalArgumentException("Class must be enum, not " + clazz.getCanonicalName());
+ }
+ this.enumClass = clazz;
+ }
+
+ public boolean addUnchecked(final E element) {
+ final int ordinal = element.ordinal();
+ final long key = 1L << ordinal;
+
+ final long prev = this.backingSet;
+ this.backingSet = prev | key;
+
+ return (prev & key) == 0;
+ }
+
+ public boolean removeUnchecked(final E element) {
+ final int ordinal = element.ordinal();
+ final long key = 1L << ordinal;
+
+ final long prev = this.backingSet;
+ this.backingSet = prev & ~key;
+
+ return (prev & key) != 0;
+ }
+
+ public void clear() {
+ this.backingSet = 0L;
+ }
+
+ public int size() {
+ return Long.bitCount(this.backingSet);
+ }
+
+ public void addAllUnchecked(final Collection<E> enums) {
+ for (final E element : enums) {
+ if (element == null) {
+ throw new NullPointerException("Null element");
+ }
+ this.backingSet |= (1L << element.ordinal());
+ }
+ }
+
+ public long getBackingSet() {
+ return this.backingSet;
+ }
+
+ public boolean hasCommonElements(final OptimizedSmallEnumSet<E> other) {
+ return (other.backingSet & this.backingSet) != 0;
+ }
2021-11-24 12:06:34 +01:00
+
2021-11-30 09:36:30 +01:00
+ public boolean hasElement(final E element) {
+ return (this.backingSet & (1L << element.ordinal())) != 0;
2021-11-24 12:06:34 +01:00
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+}
diff --git a/src/main/java/com/mojang/logging/LogUtils.java b/src/main/java/com/mojang/logging/LogUtils.java
index 46cab7a8c7b87ab01b26074b04f5a02b3907cfc4..49019b4a9bc4e634d54a9b0acaf9229a5c896f85 100644
--- a/src/main/java/com/mojang/logging/LogUtils.java
+++ b/src/main/java/com/mojang/logging/LogUtils.java
@@ -61,4 +61,9 @@ public class LogUtils {
public static Logger getLogger() {
return LoggerFactory.getLogger(STACK_WALKER.getCallerClass());
}
+ // Paper start
+ public static Logger getClassLogger() {
+ return LoggerFactory.getLogger(STACK_WALKER.getCallerClass().getSimpleName());
+ }
+ // Paper end
}
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
diff --git a/src/main/java/io/papermc/paper/chunk/SingleThreadChunkRegionManager.java b/src/main/java/io/papermc/paper/chunk/SingleThreadChunkRegionManager.java
new file mode 100644
index 0000000000000000000000000000000000000000..a5f706d6f716b2a463ae58adcde69d9e665c7733
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- /dev/null
+++ b/src/main/java/io/papermc/paper/chunk/SingleThreadChunkRegionManager.java
@@ -0,0 +1,477 @@
+package io.papermc.paper.chunk;
+
+import io.papermc.paper.util.maplist.IteratorSafeOrderedReferenceSet;
+import it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap;
+import it.unimi.dsi.fastutil.objects.ReferenceLinkedOpenHashSet;
+import it.unimi.dsi.fastutil.objects.ReferenceOpenHashSet;
+import io.papermc.paper.util.MCUtil;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+import net.minecraft.server.level.ServerLevel;
+import net.minecraft.world.level.ChunkPos;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Iterator;
+import java.util.List;
+import java.util.function.Supplier;
+
+public final class SingleThreadChunkRegionManager {
+
+ protected final int regionSectionMergeRadius;
+ protected final int regionSectionChunkSize;
+ public final int regionChunkShift; // log2(REGION_CHUNK_SIZE)
+
+ public final ServerLevel world;
+ public final String name;
+
+ protected final Long2ObjectOpenHashMap<RegionSection> regionsBySection = new Long2ObjectOpenHashMap<>();
+ protected final ReferenceLinkedOpenHashSet<Region> needsRecalculation = new ReferenceLinkedOpenHashSet<>();
+ protected final int minSectionRecalcCount;
+ protected final double maxDeadRegionPercent;
+ protected final Supplier<RegionData> regionDataSupplier;
+ protected final Supplier<RegionSectionData> regionSectionDataSupplier;
+
+ public SingleThreadChunkRegionManager(final ServerLevel world, final int minSectionRecalcCount,
+ final double maxDeadRegionPercent, final int sectionMergeRadius,
+ final int regionSectionChunkShift,
+ final String name, final Supplier<RegionData> regionDataSupplier,
+ final Supplier<RegionSectionData> regionSectionDataSupplier) {
+ this.regionSectionMergeRadius = sectionMergeRadius;
+ this.regionSectionChunkSize = 1 << regionSectionChunkShift;
+ this.regionChunkShift = regionSectionChunkShift;
+ this.world = world;
+ this.name = name;
+ this.minSectionRecalcCount = Math.max(2, minSectionRecalcCount);
+ this.maxDeadRegionPercent = maxDeadRegionPercent;
+ this.regionDataSupplier = regionDataSupplier;
+ this.regionSectionDataSupplier = regionSectionDataSupplier;
+ }
+
+ // tested via https://gist.github.com/Spottedleaf/aa7ade3451c37b4cac061fc77074db2f
+
+ /*
+ protected void check() {
+ ReferenceOpenHashSet<Region<T>> checked = new ReferenceOpenHashSet<>();
+
+ for (RegionSection<T> section : this.regionsBySection.values()) {
+ if (!checked.add(section.region)) {
+ section.region.check();
+ }
+ }
+ for (Region<T> region : this.needsRecalculation) {
+ region.check();
+ }
+ }
+ */
+
+ protected void addToRecalcQueue(final Region region) {
+ this.needsRecalculation.add(region);
+ }
+
+ protected void removeFromRecalcQueue(final Region region) {
+ this.needsRecalculation.remove(region);
+ }
+
+ public RegionSection getRegionSection(final int chunkX, final int chunkZ) {
+ return this.regionsBySection.get(MCUtil.getCoordinateKey(chunkX >> this.regionChunkShift, chunkZ >> this.regionChunkShift));
+ }
+
+ public Region getRegion(final int chunkX, final int chunkZ) {
+ final RegionSection section = this.regionsBySection.get(MCUtil.getCoordinateKey(chunkX >> regionChunkShift, chunkZ >> regionChunkShift));
+ return section != null ? section.region : null;
+ }
+
+ private final List<Region> toMerge = new ArrayList<>();
+
+ protected RegionSection getOrCreateAndMergeSection(final int sectionX, final int sectionZ, final RegionSection force) {
+ final long sectionKey = MCUtil.getCoordinateKey(sectionX, sectionZ);
+
+ if (force == null) {
+ RegionSection region = this.regionsBySection.get(sectionKey);
+ if (region != null) {
+ return region;
+ }
+ }
+
+ int mergeCandidateSectionSize = -1;
+ Region mergeIntoCandidate = null;
+
+ // find optimal candidate to merge into
+
+ final int minX = sectionX - this.regionSectionMergeRadius;
+ final int maxX = sectionX + this.regionSectionMergeRadius;
+ final int minZ = sectionZ - this.regionSectionMergeRadius;
+ final int maxZ = sectionZ + this.regionSectionMergeRadius;
+ for (int currX = minX; currX <= maxX; ++currX) {
+ for (int currZ = minZ; currZ <= maxZ; ++currZ) {
+ final RegionSection section = this.regionsBySection.get(MCUtil.getCoordinateKey(currX, currZ));
+ if (section == null) {
+ continue;
+ }
+ final Region region = section.region;
+ if (region.dead) {
+ throw new IllegalStateException("Dead region should not be in live region manager state: " + region);
+ }
+ final int sections = region.sections.size();
+
+ if (sections > mergeCandidateSectionSize) {
+ mergeCandidateSectionSize = sections;
+ mergeIntoCandidate = region;
+ }
+ this.toMerge.add(region);
+ }
+ }
+
+ // merge
+ if (mergeIntoCandidate != null) {
+ for (int i = 0; i < this.toMerge.size(); ++i) {
+ final Region region = this.toMerge.get(i);
+ if (region.dead || mergeIntoCandidate == region) {
+ continue;
+ }
+ region.mergeInto(mergeIntoCandidate);
+ }
+ this.toMerge.clear();
+ } else {
+ mergeIntoCandidate = new Region(this);
+ }
+
+ final RegionSection section;
+ if (force == null) {
+ this.regionsBySection.put(sectionKey, section = new RegionSection(sectionKey, this));
+ } else {
+ final RegionSection existing = this.regionsBySection.putIfAbsent(sectionKey, force);
+ if (existing != null) {
+ throw new IllegalStateException("Attempting to override section '" + existing.toStringWithRegion() +
+ ", with " + force.toStringWithRegion());
+ }
+
+ section = force;
+ }
+
+ mergeIntoCandidate.addRegionSection(section);
+ //mergeIntoCandidate.check();
+ //this.check();
+
+ return section;
+ }
+
+ public void addChunk(final int chunkX, final int chunkZ) {
+ this.getOrCreateAndMergeSection(chunkX >> this.regionChunkShift, chunkZ >> this.regionChunkShift, null).addChunk(chunkX, chunkZ);
+ }
+
+ public void removeChunk(final int chunkX, final int chunkZ) {
+ final RegionSection section = this.regionsBySection.get(
+ MCUtil.getCoordinateKey(chunkX >> this.regionChunkShift, chunkZ >> this.regionChunkShift)
+ );
+ if (section != null) {
+ section.removeChunk(chunkX, chunkZ);
+ } else {
+ throw new IllegalStateException("Cannot remove chunk at (" + chunkX + "," + chunkZ + ") from region state, section does not exist");
+ }
+ }
+
+ public void recalculateRegions() {
+ for (int i = 0, len = this.needsRecalculation.size(); i < len; ++i) {
+ final Region region = this.needsRecalculation.removeFirst();
+
+ this.recalculateRegion(region);
+ //this.check();
+ }
+ }
+
+ protected void recalculateRegion(final Region region) {
+ region.markedForRecalc = false;
+ //region.check();
+ // clear unused regions
+ for (final Iterator<RegionSection> iterator = region.deadSections.iterator(); iterator.hasNext();) {
+ final RegionSection deadSection = iterator.next();
+
+ if (deadSection.hasChunks()) {
+ throw new IllegalStateException("Dead section '" + deadSection.toStringWithRegion() + "' is marked dead but has chunks!");
+ }
+ if (!region.removeRegionSection(deadSection)) {
+ throw new IllegalStateException("Region " + region + " has inconsistent state, it should contain section " + deadSection);
+ }
+ if (!this.regionsBySection.remove(deadSection.regionCoordinate, deadSection)) {
+ throw new IllegalStateException("Cannot remove dead section '" +
+ deadSection.toStringWithRegion() + "' from section state! State at section coordinate: " +
+ this.regionsBySection.get(deadSection.regionCoordinate));
+ }
+ }
+ region.deadSections.clear();
+
+ // implicitly cover cases where size == 0
+ if (region.sections.size() < this.minSectionRecalcCount) {
+ //region.check();
+ return;
+ }
+
+ // run a test to see if we actually need to recalculate
+ // TODO
+
+ // destroy and rebuild the region
+ region.dead = true;
+
+ // destroy region state
+ for (final Iterator<RegionSection> iterator = region.sections.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final RegionSection aliveSection = iterator.next();
+ if (!aliveSection.hasChunks()) {
+ throw new IllegalStateException("Alive section '" + aliveSection.toStringWithRegion() + "' has no chunks!");
+ }
+ if (!this.regionsBySection.remove(aliveSection.regionCoordinate, aliveSection)) {
+ throw new IllegalStateException("Cannot remove alive section '" +
+ aliveSection.toStringWithRegion() + "' from section state! State at section coordinate: " +
+ this.regionsBySection.get(aliveSection.regionCoordinate));
+ }
+ }
+
+ // rebuild regions
+ for (final Iterator<RegionSection> iterator = region.sections.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final RegionSection aliveSection = iterator.next();
+ this.getOrCreateAndMergeSection(aliveSection.getSectionX(), aliveSection.getSectionZ(), aliveSection);
+ }
+ }
+
+ public static final class Region {
+ protected final IteratorSafeOrderedReferenceSet<RegionSection> sections = new IteratorSafeOrderedReferenceSet<>();
+ protected final ReferenceOpenHashSet<RegionSection> deadSections = new ReferenceOpenHashSet<>(16, 0.7f);
+ protected boolean dead;
+ protected boolean markedForRecalc;
+
+ public final SingleThreadChunkRegionManager regionManager;
+ public final RegionData regionData;
+
+ protected Region(final SingleThreadChunkRegionManager regionManager) {
+ this.regionManager = regionManager;
+ this.regionData = regionManager.regionDataSupplier.get();
+ }
+
+ public IteratorSafeOrderedReferenceSet.Iterator<RegionSection> getSections() {
+ return this.sections.iterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS);
+ }
+
+ protected final double getDeadSectionPercent() {
+ return (double)this.deadSections.size() / (double)this.sections.size();
+ }
+
+ /*
+ protected void check() {
+ if (this.dead) {
+ throw new IllegalStateException("Dead region!");
+ }
+ for (final Iterator<RegionSection<T>> iterator = this.sections.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final RegionSection<T> section = iterator.next();
+ if (section.region != this) {
+ throw new IllegalStateException("Region section must point to us!");
+ }
+ if (this.regionManager.regionsBySection.get(section.regionCoordinate) != section) {
+ throw new IllegalStateException("Region section must match the regionmanager state!");
+ }
+ }
+ }
+ */
+
+ // note: it is not true that the region at this point is not in any region. use the region field on the section
+ // to see if it is currently in another region.
+ protected final boolean addRegionSection(final RegionSection section) {
+ if (!this.sections.add(section)) {
+ return false;
+ }
+
+ section.sectionData.addToRegion(section, section.region, this);
+
+ section.region = this;
+ return true;
+ }
+
+ protected final boolean removeRegionSection(final RegionSection section) {
+ if (!this.sections.remove(section)) {
+ return false;
+ }
+
+ section.sectionData.removeFromRegion(section, this);
+
+ return true;
+ }
+
+ protected void mergeInto(final Region mergeTarget) {
+ if (this == mergeTarget) {
+ throw new IllegalStateException("Cannot merge a region onto itself");
+ }
+ if (this.dead) {
+ throw new IllegalStateException("Source region is dead! Source " + this + ", target " + mergeTarget);
+ } else if (mergeTarget.dead) {
+ throw new IllegalStateException("Target region is dead! Source " + this + ", target " + mergeTarget);
+ }
+ this.dead = true;
+ if (this.markedForRecalc) {
+ this.regionManager.removeFromRecalcQueue(this);
+ }
+
+ for (final Iterator<RegionSection> iterator = this.sections.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final RegionSection section = iterator.next();
+
+ if (!mergeTarget.addRegionSection(section)) {
+ throw new IllegalStateException("Target cannot contain source's sections! Source " + this + ", target " + mergeTarget);
+ }
+ }
+
+ for (final RegionSection deadSection : this.deadSections) {
+ if (!this.sections.contains(deadSection)) {
+ throw new IllegalStateException("Source region does not even contain its own dead sections! Missing " + deadSection + " from region " + this);
+ }
+ mergeTarget.deadSections.add(deadSection);
+ }
+ //mergeTarget.check();
+ }
+
+ protected void markSectionAlive(final RegionSection section) {
+ this.deadSections.remove(section);
+ if (this.markedForRecalc && (this.sections.size() < this.regionManager.minSectionRecalcCount || this.getDeadSectionPercent() < this.regionManager.maxDeadRegionPercent)) {
+ this.regionManager.removeFromRecalcQueue(this);
+ this.markedForRecalc = false;
+ }
+ }
+
+ protected void markSectionDead(final RegionSection section) {
+ this.deadSections.add(section);
+ if (!this.markedForRecalc && (this.sections.size() >= this.regionManager.minSectionRecalcCount || this.sections.size() == this.deadSections.size()) && this.getDeadSectionPercent() >= this.regionManager.maxDeadRegionPercent) {
+ this.regionManager.addToRecalcQueue(this);
+ this.markedForRecalc = true;
+ }
+ }
+
+ @Override
+ public String toString() {
+ final StringBuilder ret = new StringBuilder(128);
+
+ ret.append("Region{");
+ ret.append("dead=").append(this.dead).append(',');
+ ret.append("markedForRecalc=").append(this.markedForRecalc).append(',');
+
+ ret.append("sectionCount=").append(this.sections.size()).append(',');
+ ret.append("sections=[");
+ for (final Iterator<RegionSection> iterator = this.sections.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final RegionSection section = iterator.next();
+ ret.append(section);
+ if (iterator.hasNext()) {
+ ret.append(',');
+ }
+ }
+ ret.append(']');
+
+ ret.append('}');
+ return ret.toString();
+ }
+ }
+
+ public static final class RegionSection {
+ protected final long regionCoordinate;
+ protected final long[] chunksBitset;
+ protected int chunkCount;
+ protected Region region;
+
+ public final SingleThreadChunkRegionManager regionManager;
+ public final RegionSectionData sectionData;
+
+ protected RegionSection(final long regionCoordinate, final SingleThreadChunkRegionManager regionManager) {
+ this.regionCoordinate = regionCoordinate;
+ this.regionManager = regionManager;
+ this.chunksBitset = new long[Math.max(1, regionManager.regionSectionChunkSize * regionManager.regionSectionChunkSize / Long.SIZE)];
+ this.sectionData = regionManager.regionSectionDataSupplier.get();
+ }
+
+ public int getSectionX() {
+ return MCUtil.getCoordinateX(this.regionCoordinate);
+ }
+
+ public int getSectionZ() {
+ return MCUtil.getCoordinateZ(this.regionCoordinate);
+ }
+
+ public Region getRegion() {
+ return this.region;
+ }
+
+ private int getChunkIndex(final int chunkX, final int chunkZ) {
+ return (chunkX & (this.regionManager.regionSectionChunkSize - 1)) | ((chunkZ & (this.regionManager.regionSectionChunkSize - 1)) << this.regionManager.regionChunkShift);
+ }
+
+ protected boolean hasChunks() {
+ return this.chunkCount != 0;
+ }
+
+ protected void addChunk(final int chunkX, final int chunkZ) {
+ final int index = this.getChunkIndex(chunkX, chunkZ);
+ final long bitset = this.chunksBitset[index >>> 6]; // index / Long.SIZE
+ final long after = this.chunksBitset[index >>> 6] = bitset | (1L << (index & (Long.SIZE - 1)));
+ if (after == bitset) {
+ throw new IllegalStateException("Cannot add a chunk to a section which already has the chunk! RegionSection: " + this + ", global chunk: " + new ChunkPos(chunkX, chunkZ).toString());
+ }
+ if (++this.chunkCount != 1) {
+ return;
+ }
+ this.region.markSectionAlive(this);
+ }
+
+ protected void removeChunk(final int chunkX, final int chunkZ) {
+ final int index = this.getChunkIndex(chunkX, chunkZ);
+ final long before = this.chunksBitset[index >>> 6]; // index / Long.SIZE
+ final long bitset = this.chunksBitset[index >>> 6] = before & ~(1L << (index & (Long.SIZE - 1)));
+ if (before == bitset) {
+ throw new IllegalStateException("Cannot remove a chunk from a section which does not have that chunk! RegionSection: " + this + ", global chunk: " + new ChunkPos(chunkX, chunkZ).toString());
+ }
+ if (--this.chunkCount != 0) {
+ return;
+ }
+ this.region.markSectionDead(this);
+ }
+
+ @Override
+ public String toString() {
+ return "RegionSection{" +
+ "regionCoordinate=" + new ChunkPos(this.regionCoordinate).toString() + "," +
+ "chunkCount=" + this.chunkCount + "," +
+ "chunksBitset=" + toString(this.chunksBitset) + "," +
+ "hash=" + this.hashCode() +
+ "}";
+ }
+
+ public String toStringWithRegion() {
+ return "RegionSection{" +
+ "regionCoordinate=" + new ChunkPos(this.regionCoordinate).toString() + "," +
+ "chunkCount=" + this.chunkCount + "," +
+ "chunksBitset=" + toString(this.chunksBitset) + "," +
+ "hash=" + this.hashCode() + "," +
+ "region=" + this.region +
+ "}";
+ }
+
+ private static String toString(final long[] array) {
+ final StringBuilder ret = new StringBuilder();
+ for (final long value : array) {
+ // zero pad the hex string
+ final char[] zeros = new char[Long.SIZE / 4];
+ Arrays.fill(zeros, '0');
+ final String string = Long.toHexString(value);
+ System.arraycopy(string.toCharArray(), 0, zeros, zeros.length - string.length(), string.length());
+
+ ret.append(zeros);
+ }
+
+ return ret.toString();
+ }
+ }
+
+ public static interface RegionData {
+
+ }
+
+ public static interface RegionSectionData {
+
+ public void removeFromRegion(final RegionSection section, final Region from);
+
+ // removal from the old region is handled via removeFromRegion
+ public void addToRegion(final RegionSection section, final Region oldRegion, final Region newRegion);
+
+ }
+}
diff --git a/src/main/java/io/papermc/paper/chunk/system/ChunkSystem.java b/src/main/java/io/papermc/paper/chunk/system/ChunkSystem.java
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
new file mode 100644
index 0000000000000000000000000000000000000000..8a5e93961dac4d87c81c0e70b6f4124a1f1d2556
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- /dev/null
+++ b/src/main/java/io/papermc/paper/chunk/system/ChunkSystem.java
@@ -0,0 +1,294 @@
+package io.papermc.paper.chunk.system;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor;
+import com.destroystokyo.paper.util.SneakyThrow;
+import com.mojang.datafixers.util.Either;
+import com.mojang.logging.LogUtils;
+import io.papermc.paper.util.CoordinateUtils;
+import net.minecraft.server.level.ChunkHolder;
+import net.minecraft.server.level.ChunkMap;
+import net.minecraft.server.level.ServerLevel;
+import net.minecraft.server.level.ServerPlayer;
+import net.minecraft.server.level.TicketType;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+import net.minecraft.world.entity.Entity;
+import net.minecraft.world.level.ChunkPos;
+import net.minecraft.world.level.chunk.ChunkAccess;
+import net.minecraft.world.level.chunk.ChunkStatus;
+import net.minecraft.world.level.chunk.LevelChunk;
+import org.bukkit.Bukkit;
+import org.slf4j.Logger;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.CompletableFuture;
+import java.util.function.Consumer;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class ChunkSystem {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private static final Logger LOGGER = LogUtils.getLogger();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static void scheduleChunkTask(final ServerLevel level, final int chunkX, final int chunkZ, final Runnable run) {
+ scheduleChunkTask(level, chunkX, chunkZ, run, PrioritisedExecutor.Priority.NORMAL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void scheduleChunkTask(final ServerLevel level, final int chunkX, final int chunkZ, final Runnable run, final PrioritisedExecutor.Priority priority) {
+ level.chunkSource.mainThreadProcessor.execute(run);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void scheduleChunkLoad(final ServerLevel level, final int chunkX, final int chunkZ, final boolean gen,
+ final ChunkStatus toStatus, final boolean addTicket, final PrioritisedExecutor.Priority priority,
+ final Consumer<ChunkAccess> onComplete) {
+ if (gen) {
+ scheduleChunkLoad(level, chunkX, chunkZ, toStatus, addTicket, priority, onComplete);
+ return;
+ }
+ scheduleChunkLoad(level, chunkX, chunkZ, ChunkStatus.EMPTY, addTicket, priority, (final ChunkAccess chunk) -> {
+ if (chunk == null) {
+ onComplete.accept(null);
+ } else {
+ if (chunk.getStatus().isOrAfter(toStatus)) {
+ scheduleChunkLoad(level, chunkX, chunkZ, toStatus, addTicket, priority, onComplete);
+ } else {
+ onComplete.accept(null);
+ }
+ }
+ });
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ static final TicketType<Long> CHUNK_LOAD = TicketType.create("chunk_load", Long::compareTo);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private static long chunkLoadCounter = 0L;
+ public static void scheduleChunkLoad(final ServerLevel level, final int chunkX, final int chunkZ, final ChunkStatus toStatus,
+ final boolean addTicket, final PrioritisedExecutor.Priority priority, final Consumer<ChunkAccess> onComplete) {
+ if (!Bukkit.isPrimaryThread()) {
+ scheduleChunkTask(level, chunkX, chunkZ, () -> {
+ scheduleChunkLoad(level, chunkX, chunkZ, toStatus, addTicket, priority, onComplete);
+ }, priority);
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int minLevel = 33 + ChunkStatus.getDistance(toStatus);
+ final Long chunkReference = addTicket ? Long.valueOf(++chunkLoadCounter) : null;
+ final ChunkPos chunkPos = new ChunkPos(chunkX, chunkZ);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (addTicket) {
+ level.chunkSource.addTicketAtLevel(CHUNK_LOAD, chunkPos, minLevel, chunkReference);
+ }
+ level.chunkSource.runDistanceManagerUpdates();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final Consumer<ChunkAccess> loadCallback = (final ChunkAccess chunk) -> {
+ try {
+ if (onComplete != null) {
+ onComplete.accept(chunk);
+ }
+ } catch (final ThreadDeath death) {
+ throw death;
+ } catch (final Throwable thr) {
+ LOGGER.error("Exception handling chunk load callback", thr);
+ SneakyThrow.sneaky(thr);
+ } finally {
+ if (addTicket) {
+ level.chunkSource.addTicketAtLevel(TicketType.UNKNOWN, chunkPos, minLevel, chunkPos);
+ level.chunkSource.removeTicketAtLevel(CHUNK_LOAD, chunkPos, minLevel, chunkReference);
+ }
+ }
+ };
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final ChunkHolder holder = level.chunkSource.chunkMap.getUpdatingChunkIfPresent(CoordinateUtils.getChunkKey(chunkX, chunkZ));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (holder == null || holder.getTicketLevel() > minLevel) {
+ loadCallback.accept(null);
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final CompletableFuture<Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure>> loadFuture = holder.getOrScheduleFuture(toStatus, level.chunkSource.chunkMap);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (loadFuture.isDone()) {
+ loadCallback.accept(loadFuture.join().left().orElse(null));
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ loadFuture.whenCompleteAsync((final Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure> either, final Throwable thr) -> {
+ if (thr != null) {
+ loadCallback.accept(null);
+ return;
+ }
+ loadCallback.accept(either.left().orElse(null));
+ }, (final Runnable r) -> {
+ scheduleChunkTask(level, chunkX, chunkZ, r, PrioritisedExecutor.Priority.HIGHEST);
+ });
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void scheduleTickingState(final ServerLevel level, final int chunkX, final int chunkZ,
+ final ChunkHolder.FullChunkStatus toStatus, final boolean addTicket,
+ final PrioritisedExecutor.Priority priority, final Consumer<LevelChunk> onComplete) {
+ if (toStatus == ChunkHolder.FullChunkStatus.INACCESSIBLE) {
+ throw new IllegalArgumentException("Cannot wait for INACCESSIBLE status");
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (!Bukkit.isPrimaryThread()) {
+ scheduleChunkTask(level, chunkX, chunkZ, () -> {
+ scheduleTickingState(level, chunkX, chunkZ, toStatus, addTicket, priority, onComplete);
+ }, priority);
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int minLevel = 33 - (toStatus.ordinal() - 1);
+ final int radius = toStatus.ordinal() - 1;
+ final Long chunkReference = addTicket ? Long.valueOf(++chunkLoadCounter) : null;
+ final ChunkPos chunkPos = new ChunkPos(chunkX, chunkZ);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (addTicket) {
+ level.chunkSource.addTicketAtLevel(CHUNK_LOAD, chunkPos, minLevel, chunkReference);
+ }
+ level.chunkSource.runDistanceManagerUpdates();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final Consumer<LevelChunk> loadCallback = (final LevelChunk chunk) -> {
+ try {
+ if (onComplete != null) {
+ onComplete.accept(chunk);
+ }
+ } catch (final ThreadDeath death) {
+ throw death;
+ } catch (final Throwable thr) {
+ LOGGER.error("Exception handling chunk load callback", thr);
+ SneakyThrow.sneaky(thr);
+ } finally {
+ if (addTicket) {
+ level.chunkSource.addTicketAtLevel(TicketType.UNKNOWN, chunkPos, minLevel, chunkPos);
+ level.chunkSource.removeTicketAtLevel(CHUNK_LOAD, chunkPos, minLevel, chunkReference);
+ }
+ }
+ };
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final ChunkHolder holder = level.chunkSource.chunkMap.getUpdatingChunkIfPresent(CoordinateUtils.getChunkKey(chunkX, chunkZ));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (holder == null || holder.getTicketLevel() > minLevel) {
+ loadCallback.accept(null);
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> tickingState;
+ switch (toStatus) {
+ case BORDER: {
+ tickingState = holder.getFullChunkFuture();
+ break;
+ }
+ case TICKING: {
+ tickingState = holder.getTickingChunkFuture();
+ break;
+ }
+ case ENTITY_TICKING: {
+ tickingState = holder.getEntityTickingChunkFuture();
+ break;
+ }
+ default: {
+ throw new IllegalStateException("Cannot reach here");
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (tickingState.isDone()) {
+ loadCallback.accept(tickingState.join().left().orElse(null));
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ tickingState.whenCompleteAsync((final Either<LevelChunk, ChunkHolder.ChunkLoadingFailure> either, final Throwable thr) -> {
+ if (thr != null) {
+ loadCallback.accept(null);
+ return;
+ }
+ loadCallback.accept(either.left().orElse(null));
+ }, (final Runnable r) -> {
+ scheduleChunkTask(level, chunkX, chunkZ, r, PrioritisedExecutor.Priority.HIGHEST);
+ });
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static List<ChunkHolder> getVisibleChunkHolders(final ServerLevel level) {
+ return new ArrayList<>(level.chunkSource.chunkMap.visibleChunkMap.values());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static List<ChunkHolder> getUpdatingChunkHolders(final ServerLevel level) {
+ return new ArrayList<>(level.chunkSource.chunkMap.updatingChunkMap.values());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getVisibleChunkHolderCount(final ServerLevel level) {
+ return level.chunkSource.chunkMap.visibleChunkMap.size();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getUpdatingChunkHolderCount(final ServerLevel level) {
+ return level.chunkSource.chunkMap.updatingChunkMap.size();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static boolean hasAnyChunkHolders(final ServerLevel level) {
+ return getUpdatingChunkHolderCount(level) != 0;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onEntityPreAdd(final ServerLevel level, final Entity entity) {
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkHolderCreate(final ServerLevel level, final ChunkHolder holder) {
+ final ChunkMap chunkMap = level.chunkSource.chunkMap;
+ for (int index = 0, len = chunkMap.regionManagers.size(); index < len; ++index) {
+ chunkMap.regionManagers.get(index).addChunk(holder.pos.x, holder.pos.z);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkHolderDelete(final ServerLevel level, final ChunkHolder holder) {
+ final ChunkMap chunkMap = level.chunkSource.chunkMap;
+ for (int index = 0, len = chunkMap.regionManagers.size(); index < len; ++index) {
+ chunkMap.regionManagers.get(index).removeChunk(holder.pos.x, holder.pos.z);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkBorder(final LevelChunk chunk, final ChunkHolder holder) {
+ chunk.playerChunk = holder;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkNotBorder(final LevelChunk chunk, final ChunkHolder holder) {
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkTicking(final LevelChunk chunk, final ChunkHolder holder) {
+ chunk.level.getChunkSource().tickingChunks.add(chunk);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkNotTicking(final LevelChunk chunk, final ChunkHolder holder) {
+ chunk.level.getChunkSource().tickingChunks.remove(chunk);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkEntityTicking(final LevelChunk chunk, final ChunkHolder holder) {
+ chunk.level.getChunkSource().entityTickingChunks.add(chunk);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void onChunkNotEntityTicking(final LevelChunk chunk, final ChunkHolder holder) {
+ chunk.level.getChunkSource().entityTickingChunks.remove(chunk);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static ChunkHolder getUnloadingChunkHolder(final ServerLevel level, final int chunkX, final int chunkZ) {
+ return level.chunkSource.chunkMap.getUnloadingChunkHolder(chunkX, chunkZ);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getSendViewDistance(final ServerPlayer player) {
+ return getLoadViewDistance(player);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getLoadViewDistance(final ServerPlayer player) {
+ final ServerLevel level = player.getLevel();
+ if (level == null) {
+ return Bukkit.getViewDistance() + 1;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ return level.chunkSource.chunkMap.getEffectiveViewDistance() + 1;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getTickViewDistance(final ServerPlayer player) {
+ final ServerLevel level = player.getLevel();
+ if (level == null) {
+ return Bukkit.getSimulationDistance();
+ }
+ return level.chunkSource.chunkMap.distanceManager.getSimulationDistance();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private ChunkSystem() {
+ throw new RuntimeException();
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/CachedLists.java b/src/main/java/io/papermc/paper/util/CachedLists.java
new file mode 100644
index 0000000000000000000000000000000000000000..be668387f65a633c6ac497fca632a4767a1bf3a2
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/CachedLists.java
@@ -0,0 +1,8 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class CachedLists {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static void reset() {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/CoordinateUtils.java b/src/main/java/io/papermc/paper/util/CoordinateUtils.java
new file mode 100644
index 0000000000000000000000000000000000000000..413e4b6da027876dbbe8eb78f2568a440f431547
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/CoordinateUtils.java
@@ -0,0 +1,128 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import net.minecraft.core.BlockPos;
+import net.minecraft.core.SectionPos;
+import net.minecraft.util.Mth;
+import net.minecraft.world.entity.Entity;
+import net.minecraft.world.level.ChunkPos;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class CoordinateUtils {
+
+ // dx, dz are relative to the target chunk
+ // dx, dz in [-radius, radius]
+ public static int getNeighbourMappedIndex(final int dx, final int dz, final int radius) {
+ return (dx + radius) + (2 * radius + 1)*(dz + radius);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // the chunk keys are compatible with vanilla
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static long getChunkKey(final BlockPos pos) {
+ return ((long)(pos.getZ() >> 4) << 32) | ((pos.getX() >> 4) & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkKey(final Entity entity) {
+ return ((Mth.lfloor(entity.getZ()) >> 4) << 32) | ((Mth.lfloor(entity.getX()) >> 4) & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkKey(final ChunkPos pos) {
+ return ((long)pos.z << 32) | (pos.x & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkKey(final SectionPos pos) {
+ return ((long)pos.getZ() << 32) | (pos.getX() & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkKey(final int x, final int z) {
+ return ((long)z << 32) | (x & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkX(final long chunkKey) {
+ return (int)chunkKey;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkZ(final long chunkKey) {
+ return (int)(chunkKey >>> 32);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkCoordinate(final double blockCoordinate) {
+ return Mth.floor(blockCoordinate) >> 4;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // the section keys are compatible with vanilla's
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ static final int SECTION_X_BITS = 22;
+ static final long SECTION_X_MASK = (1L << SECTION_X_BITS) - 1;
+ static final int SECTION_Y_BITS = 20;
+ static final long SECTION_Y_MASK = (1L << SECTION_Y_BITS) - 1;
+ static final int SECTION_Z_BITS = 22;
+ static final long SECTION_Z_MASK = (1L << SECTION_Z_BITS) - 1;
+ // format is y,z,x (in order of LSB to MSB)
+ static final int SECTION_Y_SHIFT = 0;
+ static final int SECTION_Z_SHIFT = SECTION_Y_SHIFT + SECTION_Y_BITS;
+ static final int SECTION_X_SHIFT = SECTION_Z_SHIFT + SECTION_X_BITS;
+ static final int SECTION_TO_BLOCK_SHIFT = 4;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static long getChunkSectionKey(final int x, final int y, final int z) {
+ return ((x & SECTION_X_MASK) << SECTION_X_SHIFT)
+ | ((y & SECTION_Y_MASK) << SECTION_Y_SHIFT)
+ | ((z & SECTION_Z_MASK) << SECTION_Z_SHIFT);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkSectionKey(final SectionPos pos) {
+ return ((pos.getX() & SECTION_X_MASK) << SECTION_X_SHIFT)
+ | ((pos.getY() & SECTION_Y_MASK) << SECTION_Y_SHIFT)
+ | ((pos.getZ() & SECTION_Z_MASK) << SECTION_Z_SHIFT);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkSectionKey(final ChunkPos pos, final int y) {
+ return ((pos.x & SECTION_X_MASK) << SECTION_X_SHIFT)
+ | ((y & SECTION_Y_MASK) << SECTION_Y_SHIFT)
+ | ((pos.z & SECTION_Z_MASK) << SECTION_Z_SHIFT);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getChunkSectionKey(final BlockPos pos) {
+ return (((long)pos.getX() << (SECTION_X_SHIFT - SECTION_TO_BLOCK_SHIFT)) & (SECTION_X_MASK << SECTION_X_SHIFT)) |
+ ((pos.getY() >> SECTION_TO_BLOCK_SHIFT) & (SECTION_Y_MASK << SECTION_Y_SHIFT)) |
+ (((long)pos.getZ() << (SECTION_Z_SHIFT - SECTION_TO_BLOCK_SHIFT)) & (SECTION_Z_MASK << SECTION_Z_SHIFT));
+ }
+
+ public static long getChunkSectionKey(final Entity entity) {
+ return ((Mth.lfloor(entity.getX()) << (SECTION_X_SHIFT - SECTION_TO_BLOCK_SHIFT)) & (SECTION_X_MASK << SECTION_X_SHIFT)) |
+ ((Mth.lfloor(entity.getY()) >> SECTION_TO_BLOCK_SHIFT) & (SECTION_Y_MASK << SECTION_Y_SHIFT)) |
+ ((Mth.lfloor(entity.getZ()) << (SECTION_Z_SHIFT - SECTION_TO_BLOCK_SHIFT)) & (SECTION_Z_MASK << SECTION_Z_SHIFT));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkSectionX(final long key) {
+ return (int)(key << (Long.SIZE - (SECTION_X_SHIFT + SECTION_X_BITS)) >> (Long.SIZE - SECTION_X_BITS));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkSectionY(final long key) {
+ return (int)(key << (Long.SIZE - (SECTION_Y_SHIFT + SECTION_Y_BITS)) >> (Long.SIZE - SECTION_Y_BITS));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkSectionZ(final long key) {
+ return (int)(key << (Long.SIZE - (SECTION_Z_SHIFT + SECTION_Z_BITS)) >> (Long.SIZE - SECTION_Z_BITS));
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // the block coordinates are not necessarily compatible with vanilla's
+
+ public static int getBlockCoordinate(final double blockCoordinate) {
+ return Mth.floor(blockCoordinate);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getBlockKey(final int x, final int y, final int z) {
+ return ((long)x & 0x7FFFFFF) | (((long)z & 0x7FFFFFF) << 27) | ((long)y << 54);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getBlockKey(final BlockPos pos) {
+ return ((long)pos.getX() & 0x7FFFFFF) | (((long)pos.getZ() & 0x7FFFFFF) << 27) | ((long)pos.getY() << 54);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getBlockKey(final Entity entity) {
+ return ((long)entity.getX() & 0x7FFFFFF) | (((long)entity.getZ() & 0x7FFFFFF) << 27) | ((long)entity.getY() << 54);
+ }
+
+ private CoordinateUtils() {
+ throw new RuntimeException();
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/IntegerUtil.java b/src/main/java/io/papermc/paper/util/IntegerUtil.java
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
new file mode 100644
index 0000000000000000000000000000000000000000..a1bc1d1d0c86217ef18883d281195bc6b27e52ac
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/IntegerUtil.java
@@ -0,0 +1,226 @@
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+package io.papermc.paper.util;
+
+public final class IntegerUtil {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static final int HIGH_BIT_U32 = Integer.MIN_VALUE;
+ public static final long HIGH_BIT_U64 = Long.MIN_VALUE;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int ceilLog2(final int value) {
+ return Integer.SIZE - Integer.numberOfLeadingZeros(value - 1); // see doc of numberOfLeadingZeros
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long ceilLog2(final long value) {
+ return Long.SIZE - Long.numberOfLeadingZeros(value - 1); // see doc of numberOfLeadingZeros
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int floorLog2(final int value) {
+ // xor is optimized subtract for 2^n -1
+ // note that (2^n -1) - k = (2^n -1) ^ k for k <= (2^n - 1)
+ return (Integer.SIZE - 1) ^ Integer.numberOfLeadingZeros(value); // see doc of numberOfLeadingZeros
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int floorLog2(final long value) {
+ // xor is optimized subtract for 2^n -1
+ // note that (2^n -1) - k = (2^n -1) ^ k for k <= (2^n - 1)
+ return (Long.SIZE - 1) ^ Long.numberOfLeadingZeros(value); // see doc of numberOfLeadingZeros
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int roundCeilLog2(final int value) {
+ // optimized variant of 1 << (32 - leading(val - 1))
+ // given
+ // 1 << n = HIGH_BIT_32 >>> (31 - n) for n [0, 32)
+ // 1 << (32 - leading(val - 1)) = HIGH_BIT_32 >>> (31 - (32 - leading(val - 1)))
+ // HIGH_BIT_32 >>> (31 - (32 - leading(val - 1)))
+ // HIGH_BIT_32 >>> (31 - 32 + leading(val - 1))
+ // HIGH_BIT_32 >>> (-1 + leading(val - 1))
+ return HIGH_BIT_U32 >>> (Integer.numberOfLeadingZeros(value - 1) - 1);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long roundCeilLog2(final long value) {
+ // see logic documented above
+ return HIGH_BIT_U64 >>> (Long.numberOfLeadingZeros(value - 1) - 1);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int roundFloorLog2(final int value) {
+ // optimized variant of 1 << (31 - leading(val))
+ // given
+ // 1 << n = HIGH_BIT_32 >>> (31 - n) for n [0, 32)
+ // 1 << (31 - leading(val)) = HIGH_BIT_32 >> (31 - (31 - leading(val)))
+ // HIGH_BIT_32 >> (31 - (31 - leading(val)))
+ // HIGH_BIT_32 >> (31 - 31 + leading(val))
+ return HIGH_BIT_U32 >>> Integer.numberOfLeadingZeros(value);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long roundFloorLog2(final long value) {
+ // see logic documented above
+ return HIGH_BIT_U64 >>> Long.numberOfLeadingZeros(value);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static boolean isPowerOfTwo(final int n) {
+ // 2^n has one bit
+ // note: this rets true for 0 still
+ return IntegerUtil.getTrailingBit(n) == n;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static boolean isPowerOfTwo(final long n) {
+ // 2^n has one bit
+ // note: this rets true for 0 still
+ return IntegerUtil.getTrailingBit(n) == n;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getTrailingBit(final int n) {
+ return -n & n;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static long getTrailingBit(final long n) {
+ return -n & n;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int trailingZeros(final int n) {
+ return Integer.numberOfTrailingZeros(n);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int trailingZeros(final long n) {
+ return Long.numberOfTrailingZeros(n);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // from hacker's delight (signed division magic value)
+ public static int getDivisorMultiple(final long numbers) {
+ return (int)(numbers >>> 32);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // from hacker's delight (signed division magic value)
+ public static int getDivisorShift(final long numbers) {
+ return (int)numbers;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // copied from hacker's delight (signed division magic value)
+ // http://www.hackersdelight.org/hdcodetxt/magic.c.txt
+ public static long getDivisorNumbers(final int d) {
+ final int ad = IntegerUtil.branchlessAbs(d);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (ad < 2) {
+ throw new IllegalArgumentException("|number| must be in [2, 2^31 -1], not: " + d);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ final int two31 = 0x80000000;
+ final long mask = 0xFFFFFFFFL; // mask for enforcing unsigned behaviour
+
+ int p = 31;
+
+ // all these variables are UNSIGNED!
+ int t = two31 + (d >>> 31);
+ int anc = t - 1 - t%ad;
+ int q1 = (int)((two31 & mask)/(anc & mask));
+ int r1 = two31 - q1*anc;
+ int q2 = (int)((two31 & mask)/(ad & mask));
+ int r2 = two31 - q2*ad;
+ int delta;
+
+ do {
+ p = p + 1;
+ q1 = 2*q1; // Update q1 = 2**p/|nc|.
+ r1 = 2*r1; // Update r1 = rem(2**p, |nc|).
+ if ((r1 & mask) >= (anc & mask)) {// (Must be an unsigned comparison here)
+ q1 = q1 + 1;
+ r1 = r1 - anc;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ q2 = 2*q2; // Update q2 = 2**p/|d|.
+ r2 = 2*r2; // Update r2 = rem(2**p, |d|).
+ if ((r2 & mask) >= (ad & mask)) {// (Must be an unsigned comparison here)
+ q2 = q2 + 1;
+ r2 = r2 - ad;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ delta = ad - r2;
+ } while ((q1 & mask) < (delta & mask) || (q1 == delta && r1 == 0));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ int magicNum = q2 + 1;
+ if (d < 0) {
+ magicNum = -magicNum;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ int shift = p - 32;
+ return ((long)magicNum << 32) | shift;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int branchlessAbs(final int val) {
+ // -n = -1 ^ n + 1
+ final int mask = val >> (Integer.SIZE - 1); // -1 if < 0, 0 if >= 0
+ return (mask ^ val) - mask; // if val < 0, then (0 ^ val) - 0 else (-1 ^ val) + 1
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long branchlessAbs(final long val) {
+ // -n = -1 ^ n + 1
+ final long mask = val >> (Long.SIZE - 1); // -1 if < 0, 0 if >= 0
+ return (mask ^ val) - mask; // if val < 0, then (0 ^ val) - 0 else (-1 ^ val) + 1
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ //https://github.com/skeeto/hash-prospector for hash functions
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ //score = ~590.47984224483832
+ public static int hash0(int x) {
+ x *= 0x36935555;
+ x ^= x >>> 16;
+ return x;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ //score = ~310.01596637036749
+ public static int hash1(int x) {
+ x ^= x >>> 15;
+ x *= 0x356aaaad;
+ x ^= x >>> 17;
+ return x;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int hash2(int x) {
+ x ^= x >>> 16;
+ x *= 0x7feb352d;
+ x ^= x >>> 15;
+ x *= 0x846ca68b;
+ x ^= x >>> 16;
+ return x;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int hash3(int x) {
+ x ^= x >>> 17;
+ x *= 0xed5ad4bb;
+ x ^= x >>> 11;
+ x *= 0xac4c1b51;
+ x ^= x >>> 15;
+ x *= 0x31848bab;
+ x ^= x >>> 14;
+ return x;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ //score = ~365.79959673201887
+ public static long hash1(long x) {
+ x ^= x >>> 27;
+ x *= 0xb24924b71d2d354bL;
+ x ^= x >>> 28;
+ return x;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ //h2 hash
+ public static long hash2(long x) {
+ x ^= x >>> 32;
+ x *= 0xd6e8feb86659fd93L;
+ x ^= x >>> 32;
+ x *= 0xd6e8feb86659fd93L;
+ x ^= x >>> 32;
+ return x;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static long hash3(long x) {
+ x ^= x >>> 45;
+ x *= 0xc161abe5704b6c79L;
+ x ^= x >>> 41;
+ x *= 0xe3e5389aedbc90f7L;
+ x ^= x >>> 56;
+ x *= 0x1f9aba75a52db073L;
+ x ^= x >>> 53;
+ return x;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private IntegerUtil() {
+ throw new RuntimeException();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/IntervalledCounter.java b/src/main/java/io/papermc/paper/util/IntervalledCounter.java
new file mode 100644
index 0000000000000000000000000000000000000000..cea9c098ade00ee87b8efc8164ab72f5279758f0
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/IntervalledCounter.java
@@ -0,0 +1,115 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class IntervalledCounter {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected long[] times;
+ protected long[] counts;
+ protected final long interval;
+ protected long minTime;
+ protected long sum;
+ protected int head; // inclusive
+ protected int tail; // exclusive
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public IntervalledCounter(final long interval) {
+ this.times = new long[8];
+ this.counts = new long[8];
+ this.interval = interval;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public void updateCurrentTime() {
+ this.updateCurrentTime(System.nanoTime());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public void updateCurrentTime(final long currentTime) {
+ long sum = this.sum;
+ int head = this.head;
+ final int tail = this.tail;
+ final long minTime = currentTime - this.interval;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int arrayLen = this.times.length;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // guard against overflow by using subtraction
+ while (head != tail && this.times[head] - minTime < 0) {
+ sum -= this.counts[head];
+ // there are two ways we can do this:
+ // 1. free the count when adding
+ // 2. free it now
+ // option #2
+ this.counts[head] = 0;
+ if (++head >= arrayLen) {
+ head = 0;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ this.sum = sum;
+ this.head = head;
+ this.minTime = minTime;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public void addTime(final long currTime) {
+ this.addTime(currTime, 1L);
+ }
+
+ public void addTime(final long currTime, final long count) {
+ // guard against overflow by using subtraction
+ if (currTime - this.minTime < 0) {
+ return;
+ }
+ int nextTail = (this.tail + 1) % this.times.length;
+ if (nextTail == this.head) {
+ this.resize();
+ nextTail = (this.tail + 1) % this.times.length;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ this.times[this.tail] = currTime;
+ this.counts[this.tail] += count;
+ this.sum += count;
+ this.tail = nextTail;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public void updateAndAdd(final int count) {
+ final long currTime = System.nanoTime();
+ this.updateCurrentTime(currTime);
+ this.addTime(currTime, count);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public void updateAndAdd(final int count, final long currTime) {
+ this.updateCurrentTime(currTime);
+ this.addTime(currTime, count);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ private void resize() {
+ final long[] oldElements = this.times;
+ final long[] oldCounts = this.counts;
+ final long[] newElements = new long[this.times.length * 2];
+ final long[] newCounts = new long[this.times.length * 2];
+ this.times = newElements;
+ this.counts = newCounts;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int head = this.head;
+ final int tail = this.tail;
+ final int size = tail >= head ? (tail - head) : (tail + (oldElements.length - head));
+ this.head = 0;
+ this.tail = size;
+
+ if (tail >= head) {
+ System.arraycopy(oldElements, head, newElements, 0, size);
+ System.arraycopy(oldCounts, head, newCounts, 0, size);
+ } else {
+ System.arraycopy(oldElements, head, newElements, 0, oldElements.length - head);
+ System.arraycopy(oldElements, 0, newElements, oldElements.length - head, tail);
+
+ System.arraycopy(oldCounts, head, newCounts, 0, oldCounts.length - head);
+ System.arraycopy(oldCounts, 0, newCounts, oldCounts.length - head, tail);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+
+ // returns in units per second
+ public double getRate() {
+ return this.size() / (this.interval * 1.0e-9);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public long size() {
+ return this.sum;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/MCUtil.java b/src/main/java/io/papermc/paper/util/MCUtil.java
new file mode 100644
index 0000000000000000000000000000000000000000..9798a1010120125039cbf226a0e3679cc92f92c7
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/MCUtil.java
@@ -0,0 +1,512 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import com.google.common.util.concurrent.ThreadFactoryBuilder;
+import it.unimi.dsi.fastutil.objects.ObjectRBTreeSet;
+import java.lang.ref.Cleaner;
+import net.minecraft.core.BlockPos;
+import net.minecraft.core.Direction;
+import net.minecraft.server.MinecraftServer;
+import net.minecraft.server.level.ServerLevel;
+import net.minecraft.world.entity.Entity;
+import net.minecraft.world.level.ChunkPos;
+import net.minecraft.world.level.ClipContext;
+import net.minecraft.world.level.Level;
+import org.apache.commons.lang.exception.ExceptionUtils;
+import org.bukkit.Location;
+import org.bukkit.block.BlockFace;
+import org.bukkit.craftbukkit.CraftWorld;
+import org.bukkit.craftbukkit.util.Waitable;
+import org.spigotmc.AsyncCatcher;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import javax.annotation.Nonnull;
+import javax.annotation.Nullable;
+import java.util.List;
+import java.util.Queue;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.LinkedBlockingQueue;
+import java.util.concurrent.ThreadPoolExecutor;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.function.BiConsumer;
+import java.util.function.Consumer;
+import java.util.function.Supplier;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class MCUtil {
+ public static final ThreadPoolExecutor asyncExecutor = new ThreadPoolExecutor(
+ 0, 2, 60L, TimeUnit.SECONDS,
+ new LinkedBlockingQueue<>(),
+ new ThreadFactoryBuilder()
+ .setNameFormat("Paper Async Task Handler Thread - %1$d")
+ .setUncaughtExceptionHandler(new net.minecraft.DefaultUncaughtExceptionHandlerWithName(MinecraftServer.LOGGER))
+ .build()
+ );
+ public static final ThreadPoolExecutor cleanerExecutor = new ThreadPoolExecutor(
+ 1, 1, 0L, TimeUnit.SECONDS,
+ new LinkedBlockingQueue<>(),
+ new ThreadFactoryBuilder()
+ .setNameFormat("Paper Object Cleaner")
+ .setUncaughtExceptionHandler(new net.minecraft.DefaultUncaughtExceptionHandlerWithName(MinecraftServer.LOGGER))
+ .build()
+ );
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static final long INVALID_CHUNK_KEY = getCoordinateKey(Integer.MAX_VALUE, Integer.MAX_VALUE);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+
+ public static Runnable once(Runnable run) {
+ AtomicBoolean ran = new AtomicBoolean(false);
+ return () -> {
+ if (ran.compareAndSet(false, true)) {
+ run.run();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ };
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static <T> Runnable once(List<T> list, Consumer<T> cb) {
+ return once(() -> {
+ list.forEach(cb);
+ });
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private static Runnable makeCleanerCallback(Runnable run) {
+ return once(() -> cleanerExecutor.execute(run));
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * DANGER WILL ROBINSON: Be sure you do not use a lambda that lives in the object being monitored, or leaky leaky!
+ * @param obj
+ * @param run
+ * @return
+ */
+ public static Runnable registerCleaner(Object obj, Runnable run) {
+ // Wrap callback in its own method above or the lambda will leak object
+ Runnable cleaner = makeCleanerCallback(run);
+ CleanerHolder.CLEANER.register(obj, cleaner);
+ return cleaner;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private static final class CleanerHolder {
+ private static final Cleaner CLEANER = Cleaner.create();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ /**
+ * DANGER WILL ROBINSON: Be sure you do not use a lambda that lives in the object being monitored, or leaky leaky!
+ * @param obj
+ * @param list
+ * @param cleaner
+ * @param <T>
+ * @return
+ */
+ public static <T> Runnable registerListCleaner(Object obj, List<T> list, Consumer<T> cleaner) {
+ return registerCleaner(obj, () -> {
+ list.forEach(cleaner);
+ list.clear();
+ });
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * DANGER WILL ROBINSON: Be sure you do not use a lambda that lives in the object being monitored, or leaky leaky!
+ * @param obj
+ * @param resource
+ * @param cleaner
+ * @param <T>
+ * @return
+ */
+ public static <T> Runnable registerCleaner(Object obj, T resource, java.util.function.Consumer<T> cleaner) {
+ return registerCleaner(obj, () -> cleaner.accept(resource));
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static List<ChunkPos> getSpiralOutChunks(BlockPos blockposition, int radius) {
+ List<ChunkPos> list = com.google.common.collect.Lists.newArrayList();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ list.add(new ChunkPos(blockposition.getX() >> 4, blockposition.getZ() >> 4));
+ for (int r = 1; r <= radius; r++) {
+ int x = -r;
+ int z = r;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // Iterates the edge of half of the box; then negates for other half.
+ while (x <= r && z > -r) {
+ list.add(new ChunkPos((blockposition.getX() + (x << 4)) >> 4, (blockposition.getZ() + (z << 4)) >> 4));
+ list.add(new ChunkPos((blockposition.getX() - (x << 4)) >> 4, (blockposition.getZ() - (z << 4)) >> 4));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (x < r) {
+ x++;
+ } else {
+ z--;
+ }
+ }
+ }
+ return list;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int fastFloor(double x) {
+ int truncated = (int)x;
+ return x < (double)truncated ? truncated - 1 : truncated;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int fastFloor(float x) {
+ int truncated = (int)x;
+ return x < (double)truncated ? truncated - 1 : truncated;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static float normalizeYaw(float f) {
+ float f1 = f % 360.0F;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (f1 >= 180.0F) {
+ f1 -= 360.0F;
+ }
+
+ if (f1 < -180.0F) {
+ f1 += 360.0F;
+ }
+
+ return f1;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ /**
+ * Quickly generate a stack trace for current location
+ *
+ * @return Stacktrace
+ */
+ public static String stack() {
+ return ExceptionUtils.getFullStackTrace(new Throwable());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ /**
+ * Quickly generate a stack trace for current location with message
+ *
+ * @param str
+ * @return Stacktrace
+ */
+ public static String stack(String str) {
+ return ExceptionUtils.getFullStackTrace(new Throwable(str));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getCoordinateKey(final BlockPos blockPos) {
+ return ((long)(blockPos.getZ() >> 4) << 32) | ((blockPos.getX() >> 4) & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getCoordinateKey(final Entity entity) {
+ return ((long)(MCUtil.fastFloor(entity.getZ()) >> 4) << 32) | ((MCUtil.fastFloor(entity.getX()) >> 4) & 0xFFFFFFFFL);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getCoordinateKey(final ChunkPos pair) {
+ return ((long)pair.z << 32) | (pair.x & 0xFFFFFFFFL);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static long getCoordinateKey(final int x, final int z) {
+ return ((long)z << 32) | (x & 0xFFFFFFFFL);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getCoordinateX(final long key) {
+ return (int)key;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getCoordinateZ(final long key) {
+ return (int)(key >>> 32);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getChunkCoordinate(final double coordinate) {
+ return MCUtil.fastFloor(coordinate) >> 4;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getBlockCoordinate(final double coordinate) {
+ return MCUtil.fastFloor(coordinate);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getBlockKey(final int x, final int y, final int z) {
+ return ((long)x & 0x7FFFFFF) | (((long)z & 0x7FFFFFF) << 27) | ((long)y << 54);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static long getBlockKey(final BlockPos pos) {
+ return ((long)pos.getX() & 0x7FFFFFF) | (((long)pos.getZ() & 0x7FFFFFF) << 27) | ((long)pos.getY() << 54);
+ }
+
+ public static long getBlockKey(final Entity entity) {
+ return getBlockKey(getBlockCoordinate(entity.getX()), getBlockCoordinate(entity.getY()), getBlockCoordinate(entity.getZ()));
+ }
+
+ // assumes the sets have the same comparator, and if this comparator is null then assume T is Comparable
+ public static <T> void mergeSortedSets(final java.util.function.Consumer<T> consumer, final java.util.Comparator<? super T> comparator, final java.util.SortedSet<T>...sets) {
+ final ObjectRBTreeSet<T> all = new ObjectRBTreeSet<>(comparator);
+ // note: this is done in log(n!) ~ nlogn time. It could be improved if it were to mimic what mergesort does.
+ for (java.util.SortedSet<T> set : sets) {
+ if (set != null) {
+ all.addAll(set);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ all.forEach(consumer);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ private MCUtil() {}
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static final java.util.concurrent.Executor MAIN_EXECUTOR = (run) -> {
+ if (!isMainThread()) {
+ MinecraftServer.getServer().execute(run);
+ } else {
+ run.run();
+ }
+ };
+
+ public static <T> CompletableFuture<T> ensureMain(CompletableFuture<T> future) {
+ return future.thenApplyAsync(r -> r, MAIN_EXECUTOR);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static <T> void thenOnMain(CompletableFuture<T> future, Consumer<T> consumer) {
+ future.thenAcceptAsync(consumer, MAIN_EXECUTOR);
+ }
+ public static <T> void thenOnMain(CompletableFuture<T> future, BiConsumer<T, Throwable> consumer) {
+ future.whenCompleteAsync(consumer, MAIN_EXECUTOR);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static boolean isMainThread() {
+ return MinecraftServer.getServer().isSameThread();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static org.bukkit.scheduler.BukkitTask scheduleTask(int ticks, Runnable runnable) {
+ return scheduleTask(ticks, runnable, null);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static org.bukkit.scheduler.BukkitTask scheduleTask(int ticks, Runnable runnable, String taskName) {
+ return MinecraftServer.getServer().server.getScheduler().scheduleInternalTask(runnable, ticks, taskName);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static void processQueue() {
+ Runnable runnable;
+ Queue<Runnable> processQueue = getProcessQueue();
+ while ((runnable = processQueue.poll()) != null) {
+ try {
+ runnable.run();
+ } catch (Exception e) {
+ MinecraftServer.LOGGER.error("Error executing task", e);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ public static <T> T processQueueWhileWaiting(CompletableFuture <T> future) {
+ try {
+ if (isMainThread()) {
+ while (!future.isDone()) {
+ try {
+ return future.get(1, TimeUnit.MILLISECONDS);
+ } catch (TimeoutException ignored) {
+ processQueue();
+ }
+ }
+ }
+ return future.get();
+ } catch (Exception e) {
+ throw new RuntimeException(e);
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static void ensureMain(Runnable run) {
+ ensureMain(null, run);
+ }
+ /**
+ * Ensures the target code is running on the main thread
+ * @param reason
+ * @param run
+ * @return
+ */
+ public static void ensureMain(String reason, Runnable run) {
+ if (!isMainThread()) {
+ if (reason != null) {
+ MinecraftServer.LOGGER.warn("Asynchronous " + reason + "!", new IllegalStateException());
+ }
+ getProcessQueue().add(run);
+ return;
+ }
+ run.run();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private static Queue<Runnable> getProcessQueue() {
+ return MinecraftServer.getServer().processQueue;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static <T> T ensureMain(Supplier<T> run) {
+ return ensureMain(null, run);
+ }
+ /**
+ * Ensures the target code is running on the main thread
+ * @param reason
+ * @param run
+ * @param <T>
+ * @return
+ */
+ public static <T> T ensureMain(String reason, Supplier<T> run) {
+ if (!isMainThread()) {
+ if (reason != null) {
+ MinecraftServer.LOGGER.warn("Asynchronous " + reason + "! Blocking thread until it returns ", new IllegalStateException());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ Waitable<T> wait = new Waitable<T>() {
+ @Override
+ protected T evaluate() {
+ return run.get();
+ }
+ };
+ getProcessQueue().add(wait);
+ try {
+ return wait.get();
+ } catch (InterruptedException | ExecutionException e) {
+ MinecraftServer.LOGGER.warn("Encountered exception", e);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ return null;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ return run.get();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Calculates distance between 2 entities
+ * @param e1
+ * @param e2
+ * @return
+ */
+ public static double distance(Entity e1, Entity e2) {
+ return Math.sqrt(distanceSq(e1, e2));
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+
+ /**
+ * Calculates distance between 2 block positions
+ * @param e1
+ * @param e2
+ * @return
+ */
+ public static double distance(BlockPos e1, BlockPos e2) {
+ return Math.sqrt(distanceSq(e1, e2));
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ /**
+ * Gets the distance between 2 positions
+ * @param x1
+ * @param y1
+ * @param z1
+ * @param x2
+ * @param y2
+ * @param z2
+ * @return
+ */
+ public static double distance(double x1, double y1, double z1, double x2, double y2, double z2) {
+ return Math.sqrt(distanceSq(x1, y1, z1, x2, y2, z2));
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Get's the distance squared between 2 entities
+ * @param e1
+ * @param e2
+ * @return
+ */
+ public static double distanceSq(Entity e1, Entity e2) {
+ return distanceSq(e1.getX(),e1.getY(),e1.getZ(), e2.getX(),e2.getY(),e2.getZ());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Gets the distance sqaured between 2 block positions
+ * @param pos1
+ * @param pos2
+ * @return
+ */
+ public static double distanceSq(BlockPos pos1, BlockPos pos2) {
+ return distanceSq(pos1.getX(), pos1.getY(), pos1.getZ(), pos2.getX(), pos2.getY(), pos2.getZ());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Gets the distance squared between 2 positions
+ * @param x1
+ * @param y1
+ * @param z1
+ * @param x2
+ * @param y2
+ * @param z2
+ * @return
+ */
+ public static double distanceSq(double x1, double y1, double z1, double x2, double y2, double z2) {
+ return (x1 - x2) * (x1 - x2) + (y1 - y2) * (y1 - y2) + (z1 - z2) * (z1 - z2);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Converts a NMS World/BlockPosition to Bukkit Location
+ * @param world
+ * @param x
+ * @param y
+ * @param z
+ * @return
+ */
+ public static Location toLocation(Level world, double x, double y, double z) {
+ return new Location(world.getWorld(), x, y, z);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Converts a NMS World/BlockPosition to Bukkit Location
+ * @param world
+ * @param pos
+ * @return
+ */
+ public static Location toLocation(Level world, BlockPos pos) {
+ return new Location(world.getWorld(), pos.getX(), pos.getY(), pos.getZ());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Converts an NMS entity's current location to a Bukkit Location
+ * @param entity
+ * @return
+ */
+ public static Location toLocation(Entity entity) {
+ return new Location(entity.getCommandSenderWorld().getWorld(), entity.getX(), entity.getY(), entity.getZ());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static org.bukkit.block.Block toBukkitBlock(Level world, BlockPos pos) {
+ return world.getWorld().getBlockAt(pos.getX(), pos.getY(), pos.getZ());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static BlockPos toBlockPosition(Location loc) {
+ return new BlockPos(loc.getBlockX(), loc.getBlockY(), loc.getBlockZ());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static boolean isEdgeOfChunk(BlockPos pos) {
+ final int modX = pos.getX() & 15;
+ final int modZ = pos.getZ() & 15;
+ return (modX == 0 || modX == 15 || modZ == 0 || modZ == 15);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /**
+ * Posts a task to be executed asynchronously
+ * @param run
+ */
+ public static void scheduleAsyncTask(Runnable run) {
+ asyncExecutor.execute(run);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ @Nonnull
+ public static ServerLevel getNMSWorld(@Nonnull org.bukkit.World world) {
+ return ((CraftWorld) world).getHandle();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static ServerLevel getNMSWorld(@Nonnull org.bukkit.entity.Entity entity) {
+ return getNMSWorld(entity.getWorld());
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static BlockFace toBukkitBlockFace(Direction enumDirection) {
+ switch (enumDirection) {
+ case DOWN:
+ return BlockFace.DOWN;
+ case UP:
+ return BlockFace.UP;
+ case NORTH:
+ return BlockFace.NORTH;
+ case SOUTH:
+ return BlockFace.SOUTH;
+ case WEST:
+ return BlockFace.WEST;
+ case EAST:
+ return BlockFace.EAST;
+ default:
+ return null;
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getTicketLevelFor(net.minecraft.world.level.chunk.ChunkStatus status) {
+ return net.minecraft.server.level.ChunkMap.MAX_VIEW_DISTANCE + net.minecraft.world.level.chunk.ChunkStatus.getDistance(status);
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/StackWalkerUtil.java b/src/main/java/io/papermc/paper/util/StackWalkerUtil.java
new file mode 100644
index 0000000000000000000000000000000000000000..f7114d5b8f2f93f62883e24da29afaf9f74ee1a6
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/StackWalkerUtil.java
@@ -0,0 +1,24 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import org.bukkit.plugin.java.JavaPlugin;
+import org.bukkit.plugin.java.PluginClassLoader;
+import org.jetbrains.annotations.Nullable;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import java.util.Optional;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public class StackWalkerUtil {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ @Nullable
+ public static JavaPlugin getFirstPluginCaller() {
+ Optional<JavaPlugin> foundFrame = StackWalker.getInstance(StackWalker.Option.RETAIN_CLASS_REFERENCE)
+ .walk(stream -> stream
+ .filter(frame -> frame.getDeclaringClass().getClassLoader() instanceof PluginClassLoader)
+ .map((frame) -> {
+ PluginClassLoader classLoader = (PluginClassLoader) frame.getDeclaringClass().getClassLoader();
+ return classLoader.getPlugin();
+ })
+ .findFirst());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ return foundFrame.orElse(null);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/WorldUtil.java b/src/main/java/io/papermc/paper/util/WorldUtil.java
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
new file mode 100644
index 0000000000000000000000000000000000000000..67bb91fcfb532a919954cd9d7733d09a6c3fec35
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/WorldUtil.java
@@ -0,0 +1,46 @@
+package io.papermc.paper.util;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import net.minecraft.world.level.LevelHeightAccessor;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class WorldUtil {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // min, max are inclusive
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getMaxSection(final LevelHeightAccessor world) {
+ return world.getMaxSection() - 1; // getMaxSection() is exclusive
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getMinSection(final LevelHeightAccessor world) {
+ return world.getMinSection();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getMaxLightSection(final LevelHeightAccessor world) {
+ return getMaxSection(world) + 1;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getMinLightSection(final LevelHeightAccessor world) {
+ return getMinSection(world) - 1;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+
+
+ public static int getTotalSections(final LevelHeightAccessor world) {
+ return getMaxSection(world) - getMinSection(world) + 1;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static int getTotalLightSections(final LevelHeightAccessor world) {
+ return getMaxLightSection(world) - getMinLightSection(world) + 1;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getMinBlockY(final LevelHeightAccessor world) {
+ return getMinSection(world) << 4;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static int getMaxBlockY(final LevelHeightAccessor world) {
+ return (getMaxSection(world) << 4) | 15;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private WorldUtil() {
+ throw new RuntimeException();
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/maplist/IteratorSafeOrderedReferenceSet.java b/src/main/java/io/papermc/paper/util/maplist/IteratorSafeOrderedReferenceSet.java
new file mode 100644
index 0000000000000000000000000000000000000000..0fd814f1d65c111266a2b20f86561839a4cef755
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/maplist/IteratorSafeOrderedReferenceSet.java
@@ -0,0 +1,334 @@
+package io.papermc.paper.util.maplist;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import it.unimi.dsi.fastutil.objects.Reference2IntLinkedOpenHashMap;
+import it.unimi.dsi.fastutil.objects.Reference2IntMap;
+import org.bukkit.Bukkit;
+import java.util.Arrays;
+import java.util.NoSuchElementException;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class IteratorSafeOrderedReferenceSet<E> {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public static final int ITERATOR_FLAG_SEE_ADDITIONS = 1 << 0;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected final Reference2IntLinkedOpenHashMap<E> indexMap;
+ protected int firstInvalidIndex = -1;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ /* list impl */
+ protected E[] listElements;
+ protected int listSize;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected final double maxFragFactor;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected int iteratorCount;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ private final boolean threadRestricted;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public IteratorSafeOrderedReferenceSet() {
+ this(16, 0.75f, 16, 0.2);
+ }
+
+ public IteratorSafeOrderedReferenceSet(final boolean threadRestricted) {
+ this(16, 0.75f, 16, 0.2, threadRestricted);
+ }
+
+ public IteratorSafeOrderedReferenceSet(final int setCapacity, final float setLoadFactor, final int arrayCapacity,
+ final double maxFragFactor) {
+ this(setCapacity, setLoadFactor, arrayCapacity, maxFragFactor, false);
+ }
+ public IteratorSafeOrderedReferenceSet(final int setCapacity, final float setLoadFactor, final int arrayCapacity,
+ final double maxFragFactor, final boolean threadRestricted) {
+ this.indexMap = new Reference2IntLinkedOpenHashMap<>(setCapacity, setLoadFactor);
+ this.indexMap.defaultReturnValue(-1);
+ this.maxFragFactor = maxFragFactor;
+ this.listElements = (E[])new Object[arrayCapacity];
+ this.threadRestricted = threadRestricted;
+ }
+
+ /*
+ public void check() {
+ int iterated = 0;
+ ReferenceOpenHashSet<E> check = new ReferenceOpenHashSet<>();
+ if (this.listElements != null) {
+ for (int i = 0; i < this.listSize; ++i) {
+ Object obj = this.listElements[i];
+ if (obj != null) {
+ iterated++;
+ if (!check.add((E)obj)) {
+ throw new IllegalStateException("contains duplicate");
+ }
+ if (!this.contains((E)obj)) {
+ throw new IllegalStateException("desync");
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
+
+ if (iterated != this.size()) {
+ throw new IllegalStateException("Size is mismatched! Got " + iterated + ", expected " + this.size());
+ }
+
+ check.clear();
+ iterated = 0;
+ for (final java.util.Iterator<E> iterator = this.unsafeIterator(IteratorSafeOrderedReferenceSet.ITERATOR_FLAG_SEE_ADDITIONS); iterator.hasNext();) {
+ final E element = iterator.next();
+ iterated++;
+ if (!check.add(element)) {
+ throw new IllegalStateException("contains duplicate (iterator is wrong)");
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ if (!this.contains(element)) {
+ throw new IllegalStateException("desync (iterator is wrong)");
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (iterated != this.size()) {
+ throw new IllegalStateException("Size is mismatched! (iterator is wrong) Got " + iterated + ", expected " + this.size());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ */
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected final boolean allowSafeIteration() {
+ return !this.threadRestricted || Bukkit.isPrimaryThread();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected final double getFragFactor() {
+ return 1.0 - ((double)this.indexMap.size() / (double)this.listSize);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public int createRawIterator() {
+ if (this.allowSafeIteration()) {
+ ++this.iteratorCount;
+ }
+ if (this.indexMap.isEmpty()) {
+ return -1;
+ } else {
+ return this.firstInvalidIndex == 0 ? this.indexMap.getInt(this.indexMap.firstKey()) : 0;
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public int advanceRawIterator(final int index) {
+ final E[] elements = this.listElements;
+ int ret = index + 1;
+ for (int len = this.listSize; ret < len; ++ret) {
+ if (elements[ret] != null) {
+ return ret;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ return -1;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public void finishRawIterator() {
+ if (this.allowSafeIteration() && --this.iteratorCount == 0) {
+ if (this.getFragFactor() >= this.maxFragFactor) {
+ this.defrag();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public boolean remove(final E element) {
+ final int index = this.indexMap.removeInt(element);
+ if (index >= 0) {
+ if (this.firstInvalidIndex < 0 || index < this.firstInvalidIndex) {
+ this.firstInvalidIndex = index;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ if (this.listElements[index] != element) {
+ throw new IllegalStateException();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ this.listElements[index] = null;
+ if (this.allowSafeIteration() && this.iteratorCount == 0 && this.getFragFactor() >= this.maxFragFactor) {
+ this.defrag();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ //this.check();
+ return true;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ return false;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public boolean contains(final E element) {
+ return this.indexMap.containsKey(element);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ public boolean add(final E element) {
+ final int listSize = this.listSize;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int previous = this.indexMap.putIfAbsent(element, listSize);
+ if (previous != -1) {
+ return false;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (listSize >= this.listElements.length) {
+ this.listElements = Arrays.copyOf(this.listElements, listSize * 2);
+ }
+ this.listElements[listSize] = element;
+ this.listSize = listSize + 1;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ //this.check();
+ return true;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ protected void defrag() {
+ if (this.firstInvalidIndex < 0) {
+ return; // nothing to do
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (this.indexMap.isEmpty()) {
+ Arrays.fill(this.listElements, 0, this.listSize, null);
+ this.listSize = 0;
+ this.firstInvalidIndex = -1;
+ //this.check();
+ return;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final E[] backingArray = this.listElements;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ int lastValidIndex;
+ java.util.Iterator<Reference2IntMap.Entry<E>> iterator;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (this.firstInvalidIndex == 0) {
+ iterator = this.indexMap.reference2IntEntrySet().fastIterator();
+ lastValidIndex = 0;
+ } else {
+ lastValidIndex = this.firstInvalidIndex;
+ final E key = backingArray[lastValidIndex - 1];
+ iterator = this.indexMap.reference2IntEntrySet().fastIterator(new Reference2IntMap.Entry<E>() {
+ @Override
+ public int getIntValue() {
+ throw new UnsupportedOperationException();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ @Override
+ public int setValue(int i) {
+ throw new UnsupportedOperationException();
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ @Override
+ public E getKey() {
+ return key;
+ }
+ });
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ while (iterator.hasNext()) {
+ final Reference2IntMap.Entry<E> entry = iterator.next();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int newIndex = lastValidIndex++;
+ backingArray[newIndex] = entry.getKey();
+ entry.setValue(newIndex);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ // cleanup end
+ Arrays.fill(backingArray, lastValidIndex, this.listSize, null);
+ this.listSize = lastValidIndex;
+ this.firstInvalidIndex = -1;
+ //this.check();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public E rawGet(final int index) {
+ return this.listElements[index];
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public int size() {
+ // always returns the correct amount - listSize can be different
+ return this.indexMap.size();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public IteratorSafeOrderedReferenceSet.Iterator<E> iterator() {
+ return this.iterator(0);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public IteratorSafeOrderedReferenceSet.Iterator<E> iterator(final int flags) {
+ if (this.allowSafeIteration()) {
+ ++this.iteratorCount;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ return new BaseIterator<>(this, true, (flags & ITERATOR_FLAG_SEE_ADDITIONS) != 0 ? Integer.MAX_VALUE : this.listSize);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public java.util.Iterator<E> unsafeIterator() {
+ return this.unsafeIterator(0);
+ }
+ public java.util.Iterator<E> unsafeIterator(final int flags) {
+ return new BaseIterator<>(this, false, (flags & ITERATOR_FLAG_SEE_ADDITIONS) != 0 ? Integer.MAX_VALUE : this.listSize);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ public static interface Iterator<E> extends java.util.Iterator<E> {
+
+ public void finishedIterating();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ }
+
+ protected static final class BaseIterator<E> implements IteratorSafeOrderedReferenceSet.Iterator<E> {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected final IteratorSafeOrderedReferenceSet<E> set;
+ protected final boolean canFinish;
+ protected final int maxIndex;
+ protected int nextIndex;
+ protected E pendingValue;
+ protected boolean finished;
+ protected E lastReturned;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ protected BaseIterator(final IteratorSafeOrderedReferenceSet<E> set, final boolean canFinish, final int maxIndex) {
+ this.set = set;
+ this.canFinish = canFinish;
+ this.maxIndex = maxIndex;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+
+ @Override
+ public boolean hasNext() {
+ if (this.finished) {
+ return false;
+ }
+ if (this.pendingValue != null) {
+ return true;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final E[] elements = this.set.listElements;
+ int index, len;
+ for (index = this.nextIndex, len = Math.min(this.maxIndex, this.set.listSize); index < len; ++index) {
+ final E element = elements[index];
+ if (element != null) {
+ this.pendingValue = element;
+ this.nextIndex = index + 1;
+ return true;
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ this.nextIndex = index;
+ return false;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ @Override
+ public E next() {
+ if (!this.hasNext()) {
+ throw new NoSuchElementException();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ final E ret = this.pendingValue;
+
+ this.pendingValue = null;
+ this.lastReturned = ret;
+
+ return ret;
+ }
+
+ @Override
+ public void remove() {
+ final E lastReturned = this.lastReturned;
+ if (lastReturned == null) {
+ throw new IllegalStateException();
+ }
+ this.lastReturned = null;
+ this.set.remove(lastReturned);
+ }
+
+ @Override
+ public void finishedIterating() {
+ if (this.finished || !this.canFinish) {
+ throw new IllegalStateException();
+ }
+ this.lastReturned = null;
+ this.finished = true;
+ if (this.set.allowSafeIteration()) {
+ this.set.finishRawIterator();
+ }
+ }
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/misc/Delayed26WayDistancePropagator3D.java b/src/main/java/io/papermc/paper/util/misc/Delayed26WayDistancePropagator3D.java
new file mode 100644
index 0000000000000000000000000000000000000000..470402573bc31106d5a63e415b958fb7f9c36aa9
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/misc/Delayed26WayDistancePropagator3D.java
@@ -0,0 +1,297 @@
+package io.papermc.paper.util.misc;
+
+import io.papermc.paper.util.CoordinateUtils;
+import it.unimi.dsi.fastutil.longs.Long2ByteOpenHashMap;
+import it.unimi.dsi.fastutil.longs.LongIterator;
+import it.unimi.dsi.fastutil.longs.LongLinkedOpenHashSet;
+
+public final class Delayed26WayDistancePropagator3D {
+
+ // this map is considered "stale" unless updates are propagated.
+ protected final Delayed8WayDistancePropagator2D.LevelMap levels = new Delayed8WayDistancePropagator2D.LevelMap(8192*2, 0.6f);
+
+ // this map is never stale
+ protected final Long2ByteOpenHashMap sources = new Long2ByteOpenHashMap(4096, 0.6f);
+
+ // Generally updates to positions are made close to other updates, so we link to decrease cache misses when
+ // propagating updates
+ protected final LongLinkedOpenHashSet updatedSources = new LongLinkedOpenHashSet();
+
+ @FunctionalInterface
+ public static interface LevelChangeCallback {
+
+ /**
+ * This can be called for intermediate updates. So do not rely on newLevel being close to or
+ * the exact level that is expected after a full propagation has occured.
+ */
+ public void onLevelUpdate(final long coordinate, final byte oldLevel, final byte newLevel);
+
+ }
+
+ protected final LevelChangeCallback changeCallback;
+
+ public Delayed26WayDistancePropagator3D() {
+ this(null);
+ }
+
+ public Delayed26WayDistancePropagator3D(final LevelChangeCallback changeCallback) {
+ this.changeCallback = changeCallback;
+ }
+
+ public int getLevel(final long pos) {
+ return this.levels.get(pos);
+ }
+
+ public int getLevel(final int x, final int y, final int z) {
+ return this.levels.get(CoordinateUtils.getChunkSectionKey(x, y, z));
+ }
+
+ public void setSource(final int x, final int y, final int z, final int level) {
+ this.setSource(CoordinateUtils.getChunkSectionKey(x, y, z), level);
+ }
+
+ public void setSource(final long coordinate, final int level) {
+ if ((level & 63) != level || level == 0) {
+ throw new IllegalArgumentException("Level must be in (0, 63], not " + level);
+ }
+
+ final byte byteLevel = (byte)level;
+ final byte oldLevel = this.sources.put(coordinate, byteLevel);
+
+ if (oldLevel == byteLevel) {
+ return; // nothing to do
+ }
+
+ // queue to update later
+ this.updatedSources.add(coordinate);
+ }
+
+ public void removeSource(final int x, final int y, final int z) {
+ this.removeSource(CoordinateUtils.getChunkSectionKey(x, y, z));
+ }
+
+ public void removeSource(final long coordinate) {
+ if (this.sources.remove(coordinate) != 0) {
+ this.updatedSources.add(coordinate);
+ }
+ }
+
+ // queues used for BFS propagating levels
+ protected final Delayed8WayDistancePropagator2D.WorkQueue[] levelIncreaseWorkQueues = new Delayed8WayDistancePropagator2D.WorkQueue[64];
+ {
+ for (int i = 0; i < this.levelIncreaseWorkQueues.length; ++i) {
+ this.levelIncreaseWorkQueues[i] = new Delayed8WayDistancePropagator2D.WorkQueue();
+ }
+ }
+ protected final Delayed8WayDistancePropagator2D.WorkQueue[] levelRemoveWorkQueues = new Delayed8WayDistancePropagator2D.WorkQueue[64];
+ {
+ for (int i = 0; i < this.levelRemoveWorkQueues.length; ++i) {
+ this.levelRemoveWorkQueues[i] = new Delayed8WayDistancePropagator2D.WorkQueue();
+ }
+ }
+ protected long levelIncreaseWorkQueueBitset;
+ protected long levelRemoveWorkQueueBitset;
+
+ protected final void addToIncreaseWorkQueue(final long coordinate, final byte level) {
+ final Delayed8WayDistancePropagator2D.WorkQueue queue = this.levelIncreaseWorkQueues[level];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
+
+ this.levelIncreaseWorkQueueBitset |= (1L << level);
+ }
+
+ protected final void addToIncreaseWorkQueue(final long coordinate, final byte index, final byte level) {
+ final Delayed8WayDistancePropagator2D.WorkQueue queue = this.levelIncreaseWorkQueues[index];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
+
+ this.levelIncreaseWorkQueueBitset |= (1L << index);
+ }
+
+ protected final void addToRemoveWorkQueue(final long coordinate, final byte level) {
+ final Delayed8WayDistancePropagator2D.WorkQueue queue = this.levelRemoveWorkQueues[level];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
+
+ this.levelRemoveWorkQueueBitset |= (1L << level);
+ }
+
+ public boolean propagateUpdates() {
+ if (this.updatedSources.isEmpty()) {
+ return false;
+ }
+
+ boolean ret = false;
+
+ for (final LongIterator iterator = this.updatedSources.iterator(); iterator.hasNext();) {
+ final long coordinate = iterator.nextLong();
+
+ final byte currentLevel = this.levels.get(coordinate);
+ final byte updatedSource = this.sources.get(coordinate);
+
+ if (currentLevel == updatedSource) {
+ continue;
+ }
+ ret = true;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ if (updatedSource > currentLevel) {
+ // level increase
+ this.addToIncreaseWorkQueue(coordinate, updatedSource);
+ } else {
+ // level decrease
+ this.addToRemoveWorkQueue(coordinate, currentLevel);
+ // if the current coordinate is a source, then the decrease propagation will detect that and queue
+ // the source propagation
+ }
+ }
+
+ this.updatedSources.clear();
+
+ // propagate source level increases first for performance reasons (in crowded areas hopefully the additions
+ // make the removes remove less)
+ this.propagateIncreases();
+
+ // now we propagate the decreases (which will then re-propagate clobbered sources)
+ this.propagateDecreases();
+
+ return ret;
+ }
+
+ protected void propagateIncreases() {
+ for (int queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelIncreaseWorkQueueBitset);
+ this.levelIncreaseWorkQueueBitset != 0L;
+ this.levelIncreaseWorkQueueBitset ^= (1L << queueIndex), queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelIncreaseWorkQueueBitset)) {
+
+ final Delayed8WayDistancePropagator2D.WorkQueue queue = this.levelIncreaseWorkQueues[queueIndex];
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ while (!queue.queuedLevels.isEmpty()) {
+ final long coordinate = queue.queuedCoordinates.removeFirstLong();
+ byte level = queue.queuedLevels.removeFirstByte();
+
+ final boolean neighbourCheck = level < 0;
+
+ final byte currentLevel;
+ if (neighbourCheck) {
+ level = (byte)-level;
+ currentLevel = this.levels.get(coordinate);
+ } else {
+ currentLevel = this.levels.putIfGreater(coordinate, level);
+ }
+
+ if (neighbourCheck) {
+ // used when propagating from decrease to indicate that this level needs to check its neighbours
+ // this means the level at coordinate could be equal, but would still need neighbours checked
+
+ if (currentLevel != level) {
+ // something caused the level to change, which means something propagated to it (which means
+ // us propagating here is redundant), or something removed the level (which means we
+ // cannot propagate further)
+ continue;
+ }
+ } else if (currentLevel >= level) {
+ // something higher/equal propagated
+ continue;
+ }
+ if (this.changeCallback != null) {
+ this.changeCallback.onLevelUpdate(coordinate, currentLevel, level);
+ }
+
+ if (level == 1) {
+ // can't propagate 0 to neighbours
+ continue;
+ }
+
+ // propagate to neighbours
+ final byte neighbourLevel = (byte)(level - 1);
+ final int x = CoordinateUtils.getChunkSectionX(coordinate);
+ final int y = CoordinateUtils.getChunkSectionY(coordinate);
+ final int z = CoordinateUtils.getChunkSectionZ(coordinate);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ for (int dy = -1; dy <= 1; ++dy) {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ for (int dz = -1; dz <= 1; ++dz) {
+ for (int dx = -1; dx <= 1; ++dx) {
+ if ((dy | dz | dx) == 0) {
+ // already propagated to coordinate
+ continue;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // sure we can check the neighbour level in the map right now and avoid a propagation,
+ // but then we would still have to recheck it when popping the value off of the queue!
+ // so just avoid the double lookup
+ final long neighbourCoordinate = CoordinateUtils.getChunkSectionKey(dx + x, dy + y, dz + z);
+ this.addToIncreaseWorkQueue(neighbourCoordinate, neighbourLevel);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
+ }
+ }
+
+ protected void propagateDecreases() {
+ for (int queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelRemoveWorkQueueBitset);
+ this.levelRemoveWorkQueueBitset != 0L;
+ this.levelRemoveWorkQueueBitset ^= (1L << queueIndex), queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelRemoveWorkQueueBitset)) {
+
+ final Delayed8WayDistancePropagator2D.WorkQueue queue = this.levelRemoveWorkQueues[queueIndex];
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ while (!queue.queuedLevels.isEmpty()) {
+ final long coordinate = queue.queuedCoordinates.removeFirstLong();
+ final byte level = queue.queuedLevels.removeFirstByte();
+
+ final byte currentLevel = this.levels.removeIfGreaterOrEqual(coordinate, level);
+ if (currentLevel == 0) {
+ // something else removed
+ continue;
+ }
+
+ if (currentLevel > level) {
+ // something higher propagated here or we hit the propagation of another source
+ // in the second case we need to re-propagate because we could have just clobbered another source's
+ // propagation
+ this.addToIncreaseWorkQueue(coordinate, currentLevel, (byte)-currentLevel); // indicate to the increase code that the level's neighbours need checking
+ continue;
+ }
+
+ if (this.changeCallback != null) {
+ this.changeCallback.onLevelUpdate(coordinate, currentLevel, (byte)0);
+ }
+
+ final byte source = this.sources.get(coordinate);
+ if (source != 0) {
+ // must re-propagate source later
+ this.addToIncreaseWorkQueue(coordinate, source);
+ }
+
+ if (level == 0) {
+ // can't propagate -1 to neighbours
+ // we have to check neighbours for removing 1 just in case the neighbour is 2
+ continue;
+ }
+
+ // propagate to neighbours
+ final byte neighbourLevel = (byte)(level - 1);
+ final int x = CoordinateUtils.getChunkSectionX(coordinate);
+ final int y = CoordinateUtils.getChunkSectionY(coordinate);
+ final int z = CoordinateUtils.getChunkSectionZ(coordinate);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ for (int dy = -1; dy <= 1; ++dy) {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ for (int dz = -1; dz <= 1; ++dz) {
+ for (int dx = -1; dx <= 1; ++dx) {
+ if ((dy | dz | dx) == 0) {
+ // already propagated to coordinate
+ continue;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // sure we can check the neighbour level in the map right now and avoid a propagation,
+ // but then we would still have to recheck it when popping the value off of the queue!
+ // so just avoid the double lookup
+ final long neighbourCoordinate = CoordinateUtils.getChunkSectionKey(dx + x, dy + y, dz + z);
+ this.addToRemoveWorkQueue(neighbourCoordinate, neighbourLevel);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
+ }
+
+ // propagate sources we clobbered in the process
+ this.propagateIncreases();
+ }
+}
diff --git a/src/main/java/io/papermc/paper/util/misc/Delayed8WayDistancePropagator2D.java b/src/main/java/io/papermc/paper/util/misc/Delayed8WayDistancePropagator2D.java
new file mode 100644
index 0000000000000000000000000000000000000000..808d1449ac44ae86a650932365081fbaf178d141
--- /dev/null
+++ b/src/main/java/io/papermc/paper/util/misc/Delayed8WayDistancePropagator2D.java
@@ -0,0 +1,718 @@
+package io.papermc.paper.util.misc;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+import it.unimi.dsi.fastutil.HashCommon;
+import it.unimi.dsi.fastutil.bytes.ByteArrayFIFOQueue;
+import it.unimi.dsi.fastutil.longs.Long2ByteOpenHashMap;
+import it.unimi.dsi.fastutil.longs.LongArrayFIFOQueue;
+import it.unimi.dsi.fastutil.longs.LongIterator;
+import it.unimi.dsi.fastutil.longs.LongLinkedOpenHashSet;
+import io.papermc.paper.util.MCUtil;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+public final class Delayed8WayDistancePropagator2D {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ // Test
+ /*
+ protected static void test(int x, int z, com.destroystokyo.paper.util.misc.DistanceTrackingAreaMap<Ticket> reference, Delayed8WayDistancePropagator2D test) {
+ int got = test.getLevel(x, z);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ int expect = 0;
+ Object[] nearest = reference.getObjectsInRange(x, z) == null ? null : reference.getObjectsInRange(x, z).getBackingSet();
+ if (nearest != null) {
+ for (Object _obj : nearest) {
+ if (_obj instanceof Ticket) {
+ Ticket ticket = (Ticket)_obj;
+ long ticketCoord = reference.getLastCoordinate(ticket);
+ int viewDistance = reference.getLastViewDistance(ticket);
+ int distance = Math.max(com.destroystokyo.paper.util.math.IntegerUtil.branchlessAbs(MCUtil.getCoordinateX(ticketCoord) - x),
+ com.destroystokyo.paper.util.math.IntegerUtil.branchlessAbs(MCUtil.getCoordinateZ(ticketCoord) - z));
+ int level = viewDistance - distance;
+ if (level > expect) {
+ expect = level;
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
+
+ if (expect != got) {
+ throw new IllegalStateException("Expected " + expect + " at pos (" + x + "," + z + ") but got " + got);
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ static class Ticket {
+
+ int x;
+ int z;
+
+ final com.destroystokyo.paper.util.misc.PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<Ticket> empty
+ = new com.destroystokyo.paper.util.misc.PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<>(this);
+
+ }
+
+ public static void main(final String[] args) {
+ com.destroystokyo.paper.util.misc.DistanceTrackingAreaMap<Ticket> reference = new com.destroystokyo.paper.util.misc.DistanceTrackingAreaMap<Ticket>() {
+ @Override
+ protected com.destroystokyo.paper.util.misc.PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<Ticket> getEmptySetFor(Ticket object) {
+ return object.empty;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ };
+ Delayed8WayDistancePropagator2D test = new Delayed8WayDistancePropagator2D();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ final int maxDistance = 64;
+ // test origin
+ {
+ Ticket originTicket = new Ticket();
+ int originDistance = 31;
+ // test single source
+ reference.add(originTicket, 0, 0, originDistance);
+ test.setSource(0, 0, originDistance); test.propagateUpdates(); // set and propagate
+ for (int dx = -originDistance; dx <= originDistance; ++dx) {
+ for (int dz = -originDistance; dz <= originDistance; ++dz) {
+ test(dx, dz, reference, test);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ // test single source decrease
+ reference.update(originTicket, 0, 0, originDistance/2);
+ test.setSource(0, 0, originDistance/2); test.propagateUpdates(); // set and propagate
+ for (int dx = -originDistance; dx <= originDistance; ++dx) {
+ for (int dz = -originDistance; dz <= originDistance; ++dz) {
+ test(dx, dz, reference, test);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ // test source increase
+ originDistance = 2*originDistance;
+ reference.update(originTicket, 0, 0, originDistance);
+ test.setSource(0, 0, originDistance); test.propagateUpdates(); // set and propagate
+ for (int dx = -4*originDistance; dx <= 4*originDistance; ++dx) {
+ for (int dz = -4*originDistance; dz <= 4*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
2021-06-11 14:02:28 +02:00
+
+ reference.remove(originTicket);
+ test.removeSource(0, 0); test.propagateUpdates();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
2021-06-11 14:02:28 +02:00
+
+ // test multiple sources at origin
+ {
+ int originDistance = 31;
+ java.util.List<Ticket> list = new java.util.ArrayList<>();
+ for (int i = 0; i < 10; ++i) {
+ Ticket a = new Ticket();
+ list.add(a);
+ a.x = (i & 1) == 1 ? -i : i;
+ a.z = (i & 1) == 1 ? -i : i;
+ }
+ for (Ticket ticket : list) {
+ reference.add(ticket, ticket.x, ticket.z, originDistance);
+ test.setSource(ticket.x, ticket.z, originDistance);
+ }
+ test.propagateUpdates();
2021-06-11 14:02:28 +02:00
+
+ for (int dx = -8*originDistance; dx <= 8*originDistance; ++dx) {
+ for (int dz = -8*originDistance; dz <= 8*originDistance; ++dz) {
+ test(dx, dz, reference, test);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
2021-06-11 14:02:28 +02:00
+
+ // test ticket level decrease
2021-06-11 14:02:28 +02:00
+
+ for (Ticket ticket : list) {
+ reference.update(ticket, ticket.x, ticket.z, originDistance/2);
+ test.setSource(ticket.x, ticket.z, originDistance/2);
+ }
+ test.propagateUpdates();
2021-06-11 14:02:28 +02:00
+
+ for (int dx = -8*originDistance; dx <= 8*originDistance; ++dx) {
+ for (int dz = -8*originDistance; dz <= 8*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
2021-06-11 14:02:28 +02:00
+
+ // test ticket level increase
2021-06-11 14:02:28 +02:00
+
+ for (Ticket ticket : list) {
+ reference.update(ticket, ticket.x, ticket.z, originDistance*2);
+ test.setSource(ticket.x, ticket.z, originDistance*2);
2021-06-11 14:02:28 +02:00
+ }
+ test.propagateUpdates();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ for (int dx = -16*originDistance; dx <= 16*originDistance; ++dx) {
+ for (int dz = -16*originDistance; dz <= 16*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
2021-06-11 14:02:28 +02:00
+
+ // test ticket remove
+ for (int i = 0, len = list.size(); i < len; ++i) {
+ if ((i & 3) != 0) {
+ continue;
+ }
+ Ticket ticket = list.get(i);
+ reference.remove(ticket);
+ test.removeSource(ticket.x, ticket.z);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ test.propagateUpdates();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+
+ for (int dx = -16*originDistance; dx <= 16*originDistance; ++dx) {
+ for (int dz = -16*originDistance; dz <= 16*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
2021-06-11 14:02:28 +02:00
+
+ // now test at coordinate offsets
+ // test offset
+ {
+ Ticket originTicket = new Ticket();
+ int originDistance = 31;
+ int offX = 54432;
+ int offZ = -134567;
+ // test single source
+ reference.add(originTicket, offX, offZ, originDistance);
+ test.setSource(offX, offZ, originDistance); test.propagateUpdates(); // set and propagate
+ for (int dx = -originDistance; dx <= originDistance; ++dx) {
+ for (int dz = -originDistance; dz <= originDistance; ++dz) {
+ test(dx + offX, dz + offZ, reference, test);
+ }
+ }
+ // test single source decrease
+ reference.update(originTicket, offX, offZ, originDistance/2);
+ test.setSource(offX, offZ, originDistance/2); test.propagateUpdates(); // set and propagate
+ for (int dx = -originDistance; dx <= originDistance; ++dx) {
+ for (int dz = -originDistance; dz <= originDistance; ++dz) {
+ test(dx + offX, dz + offZ, reference, test);
+ }
+ }
+ // test source increase
+ originDistance = 2*originDistance;
+ reference.update(originTicket, offX, offZ, originDistance);
+ test.setSource(offX, offZ, originDistance); test.propagateUpdates(); // set and propagate
+ for (int dx = -4*originDistance; dx <= 4*originDistance; ++dx) {
+ for (int dz = -4*originDistance; dz <= 4*originDistance; ++dz) {
+ test(dx + offX, dz + offZ, reference, test);
+ }
+ }
+
+ reference.remove(originTicket);
+ test.removeSource(offX, offZ); test.propagateUpdates();
+ }
+
+ // test multiple sources at origin
+ {
+ int originDistance = 31;
+ int offX = 54432;
+ int offZ = -134567;
+ java.util.List<Ticket> list = new java.util.ArrayList<>();
+ for (int i = 0; i < 10; ++i) {
+ Ticket a = new Ticket();
+ list.add(a);
+ a.x = offX + ((i & 1) == 1 ? -i : i);
+ a.z = offZ + ((i & 1) == 1 ? -i : i);
+ }
+ for (Ticket ticket : list) {
+ reference.add(ticket, ticket.x, ticket.z, originDistance);
+ test.setSource(ticket.x, ticket.z, originDistance);
+ }
+ test.propagateUpdates();
+
+ for (int dx = -8*originDistance; dx <= 8*originDistance; ++dx) {
+ for (int dz = -8*originDistance; dz <= 8*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
+
+ // test ticket level decrease
+
+ for (Ticket ticket : list) {
+ reference.update(ticket, ticket.x, ticket.z, originDistance/2);
+ test.setSource(ticket.x, ticket.z, originDistance/2);
+ }
+ test.propagateUpdates();
+
+ for (int dx = -8*originDistance; dx <= 8*originDistance; ++dx) {
+ for (int dz = -8*originDistance; dz <= 8*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
+
+ // test ticket level increase
+
+ for (Ticket ticket : list) {
+ reference.update(ticket, ticket.x, ticket.z, originDistance*2);
+ test.setSource(ticket.x, ticket.z, originDistance*2);
+ }
+ test.propagateUpdates();
+
+ for (int dx = -16*originDistance; dx <= 16*originDistance; ++dx) {
+ for (int dz = -16*originDistance; dz <= 16*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
+
+ // test ticket remove
+ for (int i = 0, len = list.size(); i < len; ++i) {
+ if ((i & 3) != 0) {
+ continue;
+ }
+ Ticket ticket = list.get(i);
+ reference.remove(ticket);
+ test.removeSource(ticket.x, ticket.z);
+ }
+ test.propagateUpdates();
+
+ for (int dx = -16*originDistance; dx <= 16*originDistance; ++dx) {
+ for (int dz = -16*originDistance; dz <= 16*originDistance; ++dz) {
+ test(dx, dz, reference, test);
+ }
+ }
+ }
+ }
+ */
+
+ // this map is considered "stale" unless updates are propagated.
+ protected final LevelMap levels = new LevelMap(8192*2, 0.6f);
+
+ // this map is never stale
+ protected final Long2ByteOpenHashMap sources = new Long2ByteOpenHashMap(4096, 0.6f);
+
+ // Generally updates to positions are made close to other updates, so we link to decrease cache misses when
+ // propagating updates
+ protected final LongLinkedOpenHashSet updatedSources = new LongLinkedOpenHashSet();
+
+ @FunctionalInterface
+ public static interface LevelChangeCallback {
+
+ /**
+ * This can be called for intermediate updates. So do not rely on newLevel being close to or
+ * the exact level that is expected after a full propagation has occured.
+ */
+ public void onLevelUpdate(final long coordinate, final byte oldLevel, final byte newLevel);
+
+ }
+
+ protected final LevelChangeCallback changeCallback;
+
+ public Delayed8WayDistancePropagator2D() {
+ this(null);
+ }
+
+ public Delayed8WayDistancePropagator2D(final LevelChangeCallback changeCallback) {
+ this.changeCallback = changeCallback;
+ }
+
+ public int getLevel(final long pos) {
+ return this.levels.get(pos);
+ }
+
+ public int getLevel(final int x, final int z) {
+ return this.levels.get(MCUtil.getCoordinateKey(x, z));
+ }
+
+ public void setSource(final int x, final int z, final int level) {
+ this.setSource(MCUtil.getCoordinateKey(x, z), level);
+ }
+
+ public void setSource(final long coordinate, final int level) {
+ if ((level & 63) != level || level == 0) {
+ throw new IllegalArgumentException("Level must be in (0, 63], not " + level);
+ }
+
+ final byte byteLevel = (byte)level;
+ final byte oldLevel = this.sources.put(coordinate, byteLevel);
+
+ if (oldLevel == byteLevel) {
+ return; // nothing to do
+ }
+
+ // queue to update later
+ this.updatedSources.add(coordinate);
+ }
+
+ public void removeSource(final int x, final int z) {
+ this.removeSource(MCUtil.getCoordinateKey(x, z));
+ }
+
+ public void removeSource(final long coordinate) {
+ if (this.sources.remove(coordinate) != 0) {
+ this.updatedSources.add(coordinate);
+ }
+ }
+
+ // queues used for BFS propagating levels
+ protected final WorkQueue[] levelIncreaseWorkQueues = new WorkQueue[64];
+ {
+ for (int i = 0; i < this.levelIncreaseWorkQueues.length; ++i) {
+ this.levelIncreaseWorkQueues[i] = new WorkQueue();
+ }
+ }
+ protected final WorkQueue[] levelRemoveWorkQueues = new WorkQueue[64];
+ {
+ for (int i = 0; i < this.levelRemoveWorkQueues.length; ++i) {
+ this.levelRemoveWorkQueues[i] = new WorkQueue();
+ }
+ }
+ protected long levelIncreaseWorkQueueBitset;
+ protected long levelRemoveWorkQueueBitset;
+
+ protected final void addToIncreaseWorkQueue(final long coordinate, final byte level) {
+ final WorkQueue queue = this.levelIncreaseWorkQueues[level];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
+
+ this.levelIncreaseWorkQueueBitset |= (1L << level);
+ }
+
+ protected final void addToIncreaseWorkQueue(final long coordinate, final byte index, final byte level) {
+ final WorkQueue queue = this.levelIncreaseWorkQueues[index];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
+
+ this.levelIncreaseWorkQueueBitset |= (1L << index);
+ }
+
+ protected final void addToRemoveWorkQueue(final long coordinate, final byte level) {
+ final WorkQueue queue = this.levelRemoveWorkQueues[level];
+ queue.queuedCoordinates.enqueue(coordinate);
+ queue.queuedLevels.enqueue(level);
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
+
+ this.levelRemoveWorkQueueBitset |= (1L << level);
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
+ }
+
+ public boolean propagateUpdates() {
+ if (this.updatedSources.isEmpty()) {
+ return false;
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
+ }
+
+ boolean ret = false;
+
+ for (final LongIterator iterator = this.updatedSources.iterator(); iterator.hasNext();) {
+ final long coordinate = iterator.nextLong();
+
+ final byte currentLevel = this.levels.get(coordinate);
+ final byte updatedSource = this.sources.get(coordinate);
+
+ if (currentLevel == updatedSource) {
+ continue;
+ }
+ ret = true;
+
+ if (updatedSource > currentLevel) {
+ // level increase
+ this.addToIncreaseWorkQueue(coordinate, updatedSource);
+ } else {
+ // level decrease
+ this.addToRemoveWorkQueue(coordinate, currentLevel);
+ // if the current coordinate is a source, then the decrease propagation will detect that and queue
+ // the source propagation
+ }
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
+ }
+
+ this.updatedSources.clear();
+
+ // propagate source level increases first for performance reasons (in crowded areas hopefully the additions
+ // make the removes remove less)
+ this.propagateIncreases();
+
+ // now we propagate the decreases (which will then re-propagate clobbered sources)
+ this.propagateDecreases();
+
+ return ret;
+ }
2021-06-11 14:02:28 +02:00
+
+ protected void propagateIncreases() {
+ for (int queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelIncreaseWorkQueueBitset);
+ this.levelIncreaseWorkQueueBitset != 0L;
+ this.levelIncreaseWorkQueueBitset ^= (1L << queueIndex), queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelIncreaseWorkQueueBitset)) {
2021-06-11 14:02:28 +02:00
+
+ final WorkQueue queue = this.levelIncreaseWorkQueues[queueIndex];
+ while (!queue.queuedLevels.isEmpty()) {
+ final long coordinate = queue.queuedCoordinates.removeFirstLong();
+ byte level = queue.queuedLevels.removeFirstByte();
2021-06-11 14:02:28 +02:00
+
+ final boolean neighbourCheck = level < 0;
2021-06-11 14:02:28 +02:00
+
+ final byte currentLevel;
+ if (neighbourCheck) {
+ level = (byte)-level;
+ currentLevel = this.levels.get(coordinate);
+ } else {
+ currentLevel = this.levels.putIfGreater(coordinate, level);
+ }
2021-06-11 14:02:28 +02:00
+
+ if (neighbourCheck) {
+ // used when propagating from decrease to indicate that this level needs to check its neighbours
+ // this means the level at coordinate could be equal, but would still need neighbours checked
2021-06-11 14:02:28 +02:00
+
+ if (currentLevel != level) {
+ // something caused the level to change, which means something propagated to it (which means
+ // us propagating here is redundant), or something removed the level (which means we
+ // cannot propagate further)
+ continue;
+ }
+ } else if (currentLevel >= level) {
+ // something higher/equal propagated
+ continue;
+ }
+ if (this.changeCallback != null) {
+ this.changeCallback.onLevelUpdate(coordinate, currentLevel, level);
+ }
+
+ if (level == 1) {
+ // can't propagate 0 to neighbours
+ continue;
+ }
+
+ // propagate to neighbours
+ final byte neighbourLevel = (byte)(level - 1);
+ final int x = (int)coordinate;
+ final int z = (int)(coordinate >>> 32);
+
+ for (int dx = -1; dx <= 1; ++dx) {
+ for (int dz = -1; dz <= 1; ++dz) {
+ if ((dx | dz) == 0) {
+ // already propagated to coordinate
+ continue;
+ }
+
+ // sure we can check the neighbour level in the map right now and avoid a propagation,
+ // but then we would still have to recheck it when popping the value off of the queue!
+ // so just avoid the double lookup
+ final long neighbourCoordinate = MCUtil.getCoordinateKey(x + dx, z + dz);
+ this.addToIncreaseWorkQueue(neighbourCoordinate, neighbourLevel);
+ }
+ }
2021-06-11 14:02:28 +02:00
+ }
+ }
2021-06-11 14:02:28 +02:00
+ }
+
+ protected void propagateDecreases() {
+ for (int queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelRemoveWorkQueueBitset);
+ this.levelRemoveWorkQueueBitset != 0L;
+ this.levelRemoveWorkQueueBitset ^= (1L << queueIndex), queueIndex = 63 ^ Long.numberOfLeadingZeros(this.levelRemoveWorkQueueBitset)) {
2021-06-11 14:02:28 +02:00
+
+ final WorkQueue queue = this.levelRemoveWorkQueues[queueIndex];
+ while (!queue.queuedLevels.isEmpty()) {
+ final long coordinate = queue.queuedCoordinates.removeFirstLong();
+ final byte level = queue.queuedLevels.removeFirstByte();
2021-06-11 14:02:28 +02:00
+
+ final byte currentLevel = this.levels.removeIfGreaterOrEqual(coordinate, level);
+ if (currentLevel == 0) {
+ // something else removed
+ continue;
+ }
2021-06-11 14:02:28 +02:00
+
+ if (currentLevel > level) {
+ // something higher propagated here or we hit the propagation of another source
+ // in the second case we need to re-propagate because we could have just clobbered another source's
+ // propagation
+ this.addToIncreaseWorkQueue(coordinate, currentLevel, (byte)-currentLevel); // indicate to the increase code that the level's neighbours need checking
+ continue;
+ }
+
+ if (this.changeCallback != null) {
+ this.changeCallback.onLevelUpdate(coordinate, currentLevel, (byte)0);
+ }
2021-06-11 14:02:28 +02:00
+
+ final byte source = this.sources.get(coordinate);
+ if (source != 0) {
+ // must re-propagate source later
+ this.addToIncreaseWorkQueue(coordinate, source);
+ }
2021-06-11 14:02:28 +02:00
+
+ if (level == 0) {
+ // can't propagate -1 to neighbours
+ // we have to check neighbours for removing 1 just in case the neighbour is 2
+ continue;
+ }
2021-06-11 14:02:28 +02:00
+
+ // propagate to neighbours
+ final byte neighbourLevel = (byte)(level - 1);
+ final int x = (int)coordinate;
+ final int z = (int)(coordinate >>> 32);
2021-06-11 14:02:28 +02:00
+
+ for (int dx = -1; dx <= 1; ++dx) {
+ for (int dz = -1; dz <= 1; ++dz) {
+ if ((dx | dz) == 0) {
+ // already propagated to coordinate
+ continue;
+ }
2021-06-11 14:02:28 +02:00
+
+ // sure we can check the neighbour level in the map right now and avoid a propagation,
+ // but then we would still have to recheck it when popping the value off of the queue!
+ // so just avoid the double lookup
+ final long neighbourCoordinate = MCUtil.getCoordinateKey(x + dx, z + dz);
+ this.addToRemoveWorkQueue(neighbourCoordinate, neighbourLevel);
+ }
2021-06-11 14:02:28 +02:00
+ }
+ }
+ }
+
+ // propagate sources we clobbered in the process
+ this.propagateIncreases();
2021-06-11 14:02:28 +02:00
+ }
+
+ protected static final class LevelMap extends Long2ByteOpenHashMap {
+ public LevelMap() {
+ super();
+ }
2021-06-11 14:02:28 +02:00
+
+ public LevelMap(final int expected, final float loadFactor) {
+ super(expected, loadFactor);
2021-06-11 14:02:28 +02:00
+ }
+
+ // copied from superclass
+ private int find(final long k) {
+ if (k == 0L) {
+ return this.containsNullKey ? this.n : -(this.n + 1);
+ } else {
+ final long[] key = this.key;
+ long curr;
+ int pos;
+ if ((curr = key[pos = (int)HashCommon.mix(k) & this.mask]) == 0L) {
+ return -(pos + 1);
+ } else if (k == curr) {
+ return pos;
+ } else {
+ while((curr = key[pos = pos + 1 & this.mask]) != 0L) {
+ if (k == curr) {
+ return pos;
+ }
+ }
+
+ return -(pos + 1);
+ }
+ }
2021-06-11 14:02:28 +02:00
+ }
+
+ // copied from superclass
+ private void insert(final int pos, final long k, final byte v) {
+ if (pos == this.n) {
+ this.containsNullKey = true;
+ }
2021-06-11 14:02:28 +02:00
+
+ this.key[pos] = k;
+ this.value[pos] = v;
+ if (this.size++ >= this.maxFill) {
+ this.rehash(HashCommon.arraySize(this.size + 1, this.f));
2021-06-11 14:02:28 +02:00
+ }
+ }
+
+ // copied from superclass
+ public byte putIfGreater(final long key, final byte value) {
+ final int pos = this.find(key);
+ if (pos < 0) {
+ if (this.defRetValue < value) {
+ this.insert(-pos - 1, key, value);
+ }
+ return this.defRetValue;
+ } else {
+ final byte curr = this.value[pos];
+ if (value > curr) {
+ this.value[pos] = value;
+ return curr;
+ }
+ return curr;
2021-06-11 14:02:28 +02:00
+ }
+ }
+
+ // copied from superclass
+ private void removeEntry(final int pos) {
+ --this.size;
+ this.shiftKeys(pos);
+ if (this.n > this.minN && this.size < this.maxFill / 4 && this.n > 16) {
+ this.rehash(this.n / 2);
2021-06-11 14:02:28 +02:00
+ }
+ }
+
+ // copied from superclass
+ private void removeNullEntry() {
+ this.containsNullKey = false;
+ --this.size;
+ if (this.n > this.minN && this.size < this.maxFill / 4 && this.n > 16) {
+ this.rehash(this.n / 2);
2021-06-11 14:02:28 +02:00
+ }
+ }
+
+ // copied from superclass
+ public byte removeIfGreaterOrEqual(final long key, final byte value) {
+ if (key == 0L) {
+ if (!this.containsNullKey) {
+ return this.defRetValue;
+ }
+ final byte current = this.value[this.n];
+ if (value >= current) {
+ this.removeNullEntry();
+ return current;
+ }
+ return current;
+ } else {
+ long[] keys = this.key;
+ byte[] values = this.value;
+ long curr;
+ int pos;
+ if ((curr = keys[pos = (int)HashCommon.mix(key) & this.mask]) == 0L) {
+ return this.defRetValue;
+ } else if (key == curr) {
+ final byte current = values[pos];
+ if (value >= current) {
+ this.removeEntry(pos);
+ return current;
+ }
+ return current;
+ } else {
+ while((curr = keys[pos = pos + 1 & this.mask]) != 0L) {
+ if (key == curr) {
+ final byte current = values[pos];
+ if (value >= current) {
+ this.removeEntry(pos);
+ return current;
+ }
+ return current;
+ }
+ }
2021-06-11 14:02:28 +02:00
+
+ return this.defRetValue;
2021-06-11 14:02:28 +02:00
+ }
+ }
+ }
+ }
+
+ protected static final class WorkQueue {
2021-06-11 14:02:28 +02:00
+
+ public final NoResizeLongArrayFIFODeque queuedCoordinates = new NoResizeLongArrayFIFODeque();
+ public final NoResizeByteArrayFIFODeque queuedLevels = new NoResizeByteArrayFIFODeque();
2021-06-11 14:02:28 +02:00
+
+ }
+
+ protected static final class NoResizeLongArrayFIFODeque extends LongArrayFIFOQueue {
2021-06-11 14:02:28 +02:00
+
+ /**
+ * Assumes non-empty. If empty, undefined behaviour.
+ */
+ public long removeFirstLong() {
+ // copied from superclass
+ long t = this.array[this.start];
+ if (++this.start == this.length) {
+ this.start = 0;
+ }
2021-06-11 14:02:28 +02:00
+
+ return t;
+ }
2021-06-11 14:02:28 +02:00
+ }
+
+ protected static final class NoResizeByteArrayFIFODeque extends ByteArrayFIFOQueue {
2021-06-11 14:02:28 +02:00
+
+ /**
+ * Assumes non-empty. If empty, undefined behaviour.
+ */
+ public byte removeFirstByte() {
+ // copied from superclass
+ byte t = this.array[this.start];
+ if (++this.start == this.length) {
+ this.start = 0;
+ }
2021-06-11 14:02:28 +02:00
+
+ return t;
+ }
2021-06-11 14:02:28 +02:00
+ }
+}
diff --git a/src/main/java/net/minecraft/Util.java b/src/main/java/net/minecraft/Util.java
2022-12-07 18:08:55 +01:00
index dc7a805d3058244ba13c2268666f274e7ec069dc..4423140df56e64527db470ada17a117656b93114 100644
--- a/src/main/java/net/minecraft/Util.java
+++ b/src/main/java/net/minecraft/Util.java
@@ -116,7 +116,7 @@ public class Util {
}
public static long getNanos() {
- return timeSource.getAsLong();
+ return System.nanoTime(); // Paper
}
public static long getEpochMillis() {
diff --git a/src/main/java/net/minecraft/core/BlockPos.java b/src/main/java/net/minecraft/core/BlockPos.java
2022-12-07 18:08:55 +01:00
index 6801a3aa5f2f1e9a7a72ff81c09b4944ecb8349d..a72dfbfa2aa636ac8bac07a62777126cf2696241 100644
--- a/src/main/java/net/minecraft/core/BlockPos.java
+++ b/src/main/java/net/minecraft/core/BlockPos.java
2022-12-07 18:08:55 +01:00
@@ -535,6 +535,7 @@ public class BlockPos extends Vec3i {
}
}
+ // Paper start - comment out useless overrides @Override - TODO figure out why this is suddenly important to keep
@Override
public BlockPos.MutableBlockPos setX(int i) {
super.setX(i);
2022-12-07 18:08:55 +01:00
@@ -552,6 +553,7 @@ public class BlockPos extends Vec3i {
super.setZ(i);
return this;
}
+ // Paper end
@Override
public BlockPos immutable() {
diff --git a/src/main/java/net/minecraft/nbt/CompoundTag.java b/src/main/java/net/minecraft/nbt/CompoundTag.java
2022-12-07 18:08:55 +01:00
index 5b111f4543c327169d1c327c873d728009fe3a11..ca7d1d33cfd448673365b615b8609083dcb11da4 100644
--- a/src/main/java/net/minecraft/nbt/CompoundTag.java
+++ b/src/main/java/net/minecraft/nbt/CompoundTag.java
@@ -121,7 +121,7 @@ public class CompoundTag implements Tag {
return "TAG_Compound";
}
};
- private final Map<String, Tag> tags;
+ public final Map<String, Tag> tags; // Paper
protected CompoundTag(Map<String, Tag> entries) {
this.tags = entries;
2022-12-07 18:08:55 +01:00
@@ -197,6 +197,10 @@ public class CompoundTag implements Tag {
this.tags.put(key, NbtUtils.createUUID(value));
}
2021-06-11 14:02:28 +02:00
+
+ /**
+ * You must use {@link #hasUUID(String)} before or else it <b>will</b> throw an NPE.
2021-06-11 14:02:28 +02:00
+ */
public UUID getUUID(String key) {
return NbtUtils.loadUUID(this.get(key));
}
diff --git a/src/main/java/net/minecraft/network/PacketEncoder.java b/src/main/java/net/minecraft/network/PacketEncoder.java
index 207f1c1fc9d4451d27047bb8362bded8cd53e32f..021a26a6b1c258deffc26c035ab52a4ea027d9a1 100644
--- a/src/main/java/net/minecraft/network/PacketEncoder.java
+++ b/src/main/java/net/minecraft/network/PacketEncoder.java
@@ -45,7 +45,7 @@ public class PacketEncoder extends MessageToByteEncoder<Packet<?>> {
JvmProfiler.INSTANCE.onPacketSent(k, integer, channelHandlerContext.channel().remoteAddress(), j);
}
} catch (Throwable var10) {
- LOGGER.error("Error receiving packet {}", integer, var10);
+ LOGGER.error("Packet encoding of packet ID {} threw (skippable? {})", integer, packet.isSkippable(), var10); // Paper - Give proper error message
if (packet.isSkippable()) {
throw new SkipPacketException(var10);
} else {
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/MinecraftServer.java b/src/main/java/net/minecraft/server/MinecraftServer.java
2022-12-07 18:08:55 +01:00
index 06b1da8b559f9d545f1593d81de7b25e0a12f176..0f7dd33d51281b383be0fb47d4e6b133f123ce1f 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/MinecraftServer.java
+++ b/src/main/java/net/minecraft/server/MinecraftServer.java
2022-12-07 18:08:55 +01:00
@@ -293,6 +293,7 @@ public abstract class MinecraftServer extends ReentrantBlockableEventLoop<TickTa
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
public final double[] recentTps = new double[ 3 ];
// Spigot end
public final io.papermc.paper.configuration.PaperConfigurations paperConfigurations;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ public static long currentTickLong = 0L; // Paper
public static <S extends MinecraftServer> S spin(Function<Thread, S> serverFactory) {
AtomicReference<S> atomicreference = new AtomicReference();
2022-12-07 18:08:55 +01:00
@@ -929,6 +930,9 @@ public abstract class MinecraftServer extends ReentrantBlockableEventLoop<TickTa
2021-06-11 14:02:28 +02:00
MinecraftServer.LOGGER.error("Failed to unlock level {}", this.storageSource.getLevelId(), ioexception1);
}
// Spigot start
+ io.papermc.paper.util.MCUtil.asyncExecutor.shutdown(); // Paper
+ try { io.papermc.paper.util.MCUtil.asyncExecutor.awaitTermination(30, java.util.concurrent.TimeUnit.SECONDS); // Paper
2021-06-11 14:02:28 +02:00
+ } catch (java.lang.InterruptedException ignored) {} // Paper
if (org.spigotmc.SpigotConfig.saveUserCacheOnStopOnly) {
MinecraftServer.LOGGER.info("Saving usercache.json");
2021-06-11 14:02:28 +02:00
this.getProfileCache().save();
2022-12-07 18:08:55 +01:00
@@ -995,6 +999,7 @@ public abstract class MinecraftServer extends ReentrantBlockableEventLoop<TickTa
2022-06-07 20:12:34 +02:00
this.lastOverloadWarning = this.nextTickTime;
}
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
2022-06-07 20:12:34 +02:00
+ ++MinecraftServer.currentTickLong; // Paper
if ( tickCount++ % MinecraftServer.SAMPLE_INTERVAL == 0 )
{
double currentTps = 1E3 / ( curTime - tickSection ) * MinecraftServer.SAMPLE_INTERVAL;
2022-12-07 18:08:55 +01:00
@@ -1230,7 +1235,7 @@ public abstract class MinecraftServer extends ReentrantBlockableEventLoop<TickTa
2021-11-23 11:51:25 +01:00
MinecraftServer.LOGGER.debug("Autosave finished");
SpigotTimings.worldSaveTimer.stopTiming(); // Spigot
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
}
2021-11-23 11:51:25 +01:00
-
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ io.papermc.paper.util.CachedLists.reset(); // Paper
this.profiler.push("tallying");
long l = this.tickTimes[this.tickCount % 100] = Util.getNanos() - i;
2021-11-23 11:51:25 +01:00
2022-12-07 18:08:55 +01:00
@@ -1292,6 +1297,11 @@ public abstract class MinecraftServer extends ReentrantBlockableEventLoop<TickTa
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
try {
worldserver.timings.doTick.startTiming(); // Spigot
worldserver.tick(shouldKeepTicking);
+ // Paper start
+ for (final io.papermc.paper.chunk.SingleThreadChunkRegionManager regionManager : worldserver.getChunkSource().chunkMap.regionManagers) {
+ regionManager.recalculateRegions();
+ }
+ // Paper end
worldserver.timings.doTick.stopTiming(); // Spigot
} catch (Throwable throwable) {
// Spigot Start
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/ChunkHolder.java b/src/main/java/net/minecraft/server/level/ChunkHolder.java
index f902b1f7062fc2c81e0ce43e8b8599192469e57c..74d1ae0104e8d0795df50f00317fd860de4f112e 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/ChunkHolder.java
+++ b/src/main/java/net/minecraft/server/level/ChunkHolder.java
2022-03-01 06:43:03 +01:00
@@ -52,9 +52,9 @@ public class ChunkHolder {
private static final int BLOCKS_BEFORE_RESEND_FUDGE = 64;
2021-06-11 14:02:28 +02:00
private final AtomicReferenceArray<CompletableFuture<Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure>>> futures;
private final LevelHeightAccessor levelHeightAccessor;
2021-06-11 14:02:28 +02:00
- private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> fullChunkFuture;
- private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> tickingChunkFuture;
- private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> entityTickingChunkFuture;
+ private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> fullChunkFuture; private int fullChunkCreateCount; private volatile boolean isFullChunkReady; // Paper - cache chunk ticking stage
+ private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> tickingChunkFuture; private volatile boolean isTickingReady; // Paper - cache chunk ticking stage
+ private volatile CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> entityTickingChunkFuture; private volatile boolean isEntityTickingReady; // Paper - cache chunk ticking stage
private CompletableFuture<ChunkAccess> chunkToSave;
@Nullable
private final DebugBuffer<ChunkHolder.ChunkSaveDebug> chunkToSaveHistory;
@@ -73,6 +73,18 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
private boolean resendLight;
private CompletableFuture<Void> pendingFullStateConfirmation;
2021-06-11 14:02:28 +02:00
+ private final ChunkMap chunkMap; // Paper
+
+ // Paper start
+ public void onChunkAdd() {
+
+ }
+
+ public void onChunkRemove() {
+
+ }
+ // Paper end
2021-06-11 14:02:28 +02:00
+
public ChunkHolder(ChunkPos pos, int level, LevelHeightAccessor world, LevelLightEngine lightingProvider, ChunkHolder.LevelChangeListener levelUpdateListener, ChunkHolder.PlayerProvider playersWatchingChunkProvider) {
2021-06-11 14:02:28 +02:00
this.futures = new AtomicReferenceArray(ChunkHolder.CHUNK_STATUSES.size());
this.fullChunkFuture = ChunkHolder.UNLOADED_LEVEL_CHUNK_FUTURE;
@@ -93,8 +105,23 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
this.queueLevel = this.oldTicketLevel;
this.setTicketLevel(level);
this.changedBlocksPerSection = new ShortSet[world.getSectionsCount()];
2021-06-11 14:02:28 +02:00
+ this.chunkMap = (ChunkMap)playersWatchingChunkProvider; // Paper
}
+ // Paper start
+ public @Nullable ChunkAccess getAvailableChunkNow() {
2021-11-23 12:47:17 +01:00
+ // TODO can we just getStatusFuture(EMPTY)?
+ for (ChunkStatus curr = ChunkStatus.FULL, next = curr.getParent(); curr != next; curr = next, next = next.getParent()) {
+ CompletableFuture<Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure>> future = this.getFutureIfPresentUnchecked(curr);
+ Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure> either = future.getNow(null);
+ if (either == null || either.left().isEmpty()) {
2021-11-23 12:47:17 +01:00
+ continue;
+ }
+ return either.left().get();
+ }
+ return null;
+ }
+ // Paper end
// CraftBukkit start
public LevelChunk getFullChunkNow() {
// Note: We use the oldTicketLevel for isLoaded checks.
@@ -119,20 +146,20 @@ public class ChunkHolder {
return ChunkHolder.getStatus(this.ticketLevel).isOrAfter(leastStatus) ? this.getFutureIfPresentUnchecked(leastStatus) : ChunkHolder.UNLOADED_CHUNK_FUTURE;
2021-06-11 14:02:28 +02:00
}
- public CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getTickingChunkFuture() {
+ public final CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getTickingChunkFuture() { // Paper - final for inline
return this.tickingChunkFuture;
}
- public CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getEntityTickingChunkFuture() {
+ public final CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getEntityTickingChunkFuture() { // Paper - final for inline
return this.entityTickingChunkFuture;
}
- public CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getFullChunkFuture() {
+ public final CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> getFullChunkFuture() { // Paper - final for inline
return this.fullChunkFuture;
}
@Nullable
- public LevelChunk getTickingChunk() {
+ public final LevelChunk getTickingChunk() { // Paper - final for inline
CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> completablefuture = this.getTickingChunkFuture();
Either<LevelChunk, ChunkHolder.ChunkLoadingFailure> either = (Either) completablefuture.getNow(null); // CraftBukkit - decompile error
@@ -140,7 +167,7 @@ public class ChunkHolder {
2022-03-01 06:43:03 +01:00
}
@Nullable
- public LevelChunk getFullChunk() {
+ public final LevelChunk getFullChunk() { // Paper - final for inline
CompletableFuture<Either<LevelChunk, ChunkHolder.ChunkLoadingFailure>> completablefuture = this.getFullChunkFuture();
Either<LevelChunk, ChunkHolder.ChunkLoadingFailure> either = (Either) completablefuture.getNow(null); // CraftBukkit - decompile error
@@ -161,6 +188,21 @@ public class ChunkHolder {
2021-11-23 12:47:17 +01:00
return null;
}
+ // Paper start
+ public ChunkStatus getChunkHolderStatus() {
+ for (ChunkStatus curr = ChunkStatus.FULL, next = curr.getParent(); curr != next; curr = next, next = next.getParent()) {
+ CompletableFuture<Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure>> future = this.getFutureIfPresentUnchecked(curr);
+ Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure> either = future.getNow(null);
+ if (either == null || !either.left().isPresent()) {
+ continue;
+ }
+ return curr;
+ }
+
+ return null;
+ }
+ // Paper end
+
@Nullable
public ChunkAccess getLastAvailable() {
for (int i = ChunkHolder.CHUNK_STATUSES.size() - 1; i >= 0; --i) {
@@ -179,7 +221,7 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
return null;
}
- public CompletableFuture<ChunkAccess> getChunkToSave() {
+ public final CompletableFuture<ChunkAccess> getChunkToSave() { // Paper - final for inline
return this.chunkToSave;
}
@@ -360,11 +402,11 @@ public class ChunkHolder {
return ChunkHolder.getFullChunkStatus(this.ticketLevel);
2021-06-11 14:02:28 +02:00
}
- public ChunkPos getPos() {
+ public final ChunkPos getPos() { // Paper - final for inline
return this.pos;
}
- public int getTicketLevel() {
+ public final int getTicketLevel() { // Paper - final for inline
return this.ticketLevel;
}
@@ -453,14 +495,31 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
this.wasAccessibleSinceLastSave |= flag3;
if (!flag2 && flag3) {
+ int expectCreateCount = ++this.fullChunkCreateCount; // Paper
this.fullChunkFuture = chunkStorage.prepareAccessibleChunk(this);
this.scheduleFullChunkPromotion(chunkStorage, this.fullChunkFuture, executor, ChunkHolder.FullChunkStatus.BORDER);
2021-06-11 14:02:28 +02:00
+ // Paper start - cache ticking ready status
+ this.fullChunkFuture.thenAccept(either -> {
+ final Optional<LevelChunk> left = either.left();
+ if (left.isPresent() && ChunkHolder.this.fullChunkCreateCount == expectCreateCount) {
2021-06-11 14:02:28 +02:00
+ LevelChunk fullChunk = either.left().get();
+ ChunkHolder.this.isFullChunkReady = true;
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkBorder(fullChunk, this);
2021-06-11 14:02:28 +02:00
+ }
+ });
this.updateChunkToSave(this.fullChunkFuture, "full");
2021-06-11 14:02:28 +02:00
}
if (flag2 && !flag3) {
+ // Paper start
+ if (this.isFullChunkReady) {
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkNotBorder(this.fullChunkFuture.join().left().get(), this); // Paper
+ }
+ // Paper end
2021-11-23 11:51:25 +01:00
this.fullChunkFuture.complete(ChunkHolder.UNLOADED_LEVEL_CHUNK);
2021-06-11 14:02:28 +02:00
this.fullChunkFuture = ChunkHolder.UNLOADED_LEVEL_CHUNK_FUTURE;
+ ++this.fullChunkCreateCount; // Paper - cache ticking ready status
+ this.isFullChunkReady = false; // Paper - cache ticking ready status
2021-11-23 11:51:25 +01:00
}
boolean flag4 = playerchunk_state.isOrAfter(ChunkHolder.FullChunkStatus.TICKING);
@@ -469,11 +528,25 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
if (!flag4 && flag5) {
this.tickingChunkFuture = chunkStorage.prepareTickingChunk(this);
this.scheduleFullChunkPromotion(chunkStorage, this.tickingChunkFuture, executor, ChunkHolder.FullChunkStatus.TICKING);
2021-06-11 14:02:28 +02:00
+ // Paper start - cache ticking ready status
+ this.tickingChunkFuture.thenAccept(either -> {
+ either.ifLeft(chunk -> {
2021-06-11 14:02:28 +02:00
+ // note: Here is a very good place to add callbacks to logic waiting on this.
+ ChunkHolder.this.isTickingReady = true;
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkTicking(chunk, this);
+ });
2021-06-11 14:02:28 +02:00
+ });
+ // Paper end
this.updateChunkToSave(this.tickingChunkFuture, "ticking");
2021-06-11 14:02:28 +02:00
}
if (flag4 && !flag5) {
- this.tickingChunkFuture.complete(ChunkHolder.UNLOADED_LEVEL_CHUNK);
+ // Paper start
+ if (this.isTickingReady) {
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkNotTicking(this.tickingChunkFuture.join().left().get(), this); // Paper
+ }
+ // Paper end
2021-06-11 14:02:28 +02:00
+ this.tickingChunkFuture.complete(ChunkHolder.UNLOADED_LEVEL_CHUNK); this.isTickingReady = false; // Paper - cache chunk ticking stage
this.tickingChunkFuture = ChunkHolder.UNLOADED_LEVEL_CHUNK_FUTURE;
}
@@ -487,11 +560,24 @@ public class ChunkHolder {
2021-06-11 14:02:28 +02:00
this.entityTickingChunkFuture = chunkStorage.prepareEntityTickingChunk(this.pos);
this.scheduleFullChunkPromotion(chunkStorage, this.entityTickingChunkFuture, executor, ChunkHolder.FullChunkStatus.ENTITY_TICKING);
2021-06-11 14:02:28 +02:00
+ // Paper start - cache ticking ready status
+ this.entityTickingChunkFuture.thenAccept(either -> {
+ either.ifLeft(chunk -> {
2021-06-11 14:02:28 +02:00
+ ChunkHolder.this.isEntityTickingReady = true;
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkEntityTicking(chunk, this);
+ });
2021-06-11 14:02:28 +02:00
+ });
+ // Paper end
this.updateChunkToSave(this.entityTickingChunkFuture, "entity ticking");
2021-06-11 14:02:28 +02:00
}
if (flag6 && !flag7) {
- this.entityTickingChunkFuture.complete(ChunkHolder.UNLOADED_LEVEL_CHUNK);
+ // Paper start
+ if (this.isEntityTickingReady) {
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkNotEntityTicking(this.entityTickingChunkFuture.join().left().get(), this);
+ }
+ // Paper end
2021-06-11 14:02:28 +02:00
+ this.entityTickingChunkFuture.complete(ChunkHolder.UNLOADED_LEVEL_CHUNK); this.isEntityTickingReady = false; // Paper - cache chunk ticking stage
this.entityTickingChunkFuture = ChunkHolder.UNLOADED_LEVEL_CHUNK_FUTURE;
}
@@ -608,4 +694,18 @@ public class ChunkHolder {
}
};
}
+
+ // Paper start
+ public final boolean isEntityTickingReady() {
+ return this.isEntityTickingReady;
+ }
+
+ public final boolean isTickingReady() {
+ return this.isTickingReady;
+ }
+
+ public final boolean isFullChunkReady() {
+ return this.isFullChunkReady;
+ }
+ // Paper end
}
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/ChunkMap.java b/src/main/java/net/minecraft/server/level/ChunkMap.java
index 01438ee9ad7a64ac4fd124d97029f87b49c25d5d..ae69c14943bd93ca948789ad4d4da899a7505a60 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/ChunkMap.java
+++ b/src/main/java/net/minecraft/server/level/ChunkMap.java
2022-12-07 18:08:55 +01:00
@@ -65,6 +65,7 @@ import net.minecraft.network.protocol.game.ClientboundSetChunkCacheCenterPacket;
2021-06-11 14:02:28 +02:00
import net.minecraft.network.protocol.game.ClientboundSetEntityLinkPacket;
import net.minecraft.network.protocol.game.ClientboundSetPassengersPacket;
import net.minecraft.network.protocol.game.DebugPackets;
+import io.papermc.paper.util.MCUtil;
2021-06-11 14:02:28 +02:00
import net.minecraft.server.level.progress.ChunkProgressListener;
import net.minecraft.server.network.ServerPlayerConnection;
2021-06-11 14:02:28 +02:00
import net.minecraft.util.CsvOutput;
2022-12-07 18:08:55 +01:00
@@ -173,6 +174,56 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
2021-06-11 14:02:28 +02:00
};
// CraftBukkit end
+ // Paper start - distance maps
+ private final com.destroystokyo.paper.util.misc.PooledLinkedHashSets<ServerPlayer> pooledLinkedPlayerHashSets = new com.destroystokyo.paper.util.misc.PooledLinkedHashSets<>();
+
+ void addPlayerToDistanceMaps(ServerPlayer player) {
+ int chunkX = MCUtil.getChunkCoordinate(player.getX());
+ int chunkZ = MCUtil.getChunkCoordinate(player.getZ());
+ // Note: players need to be explicitly added to distance maps before they can be updated
+ }
+
+ void removePlayerFromDistanceMaps(ServerPlayer player) {
+
+ }
+
+ void updateMaps(ServerPlayer player) {
+ int chunkX = MCUtil.getChunkCoordinate(player.getX());
+ int chunkZ = MCUtil.getChunkCoordinate(player.getZ());
+ // Note: players need to be explicitly added to distance maps before they can be updated
+ }
+ // Paper end
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ // Paper start
+ public final List<io.papermc.paper.chunk.SingleThreadChunkRegionManager> regionManagers = new java.util.ArrayList<>();
+ public final io.papermc.paper.chunk.SingleThreadChunkRegionManager dataRegionManager;
+
+ public static final class DataRegionData implements io.papermc.paper.chunk.SingleThreadChunkRegionManager.RegionData {
+ }
+
+ public static final class DataRegionSectionData implements io.papermc.paper.chunk.SingleThreadChunkRegionManager.RegionSectionData {
+
+ @Override
+ public void removeFromRegion(final io.papermc.paper.chunk.SingleThreadChunkRegionManager.RegionSection section,
+ final io.papermc.paper.chunk.SingleThreadChunkRegionManager.Region from) {
+ final DataRegionSectionData sectionData = (DataRegionSectionData)section.sectionData;
+ final DataRegionData fromData = (DataRegionData)from.regionData;
+ }
+
+ @Override
+ public void addToRegion(final io.papermc.paper.chunk.SingleThreadChunkRegionManager.RegionSection section,
+ final io.papermc.paper.chunk.SingleThreadChunkRegionManager.Region oldRegion,
+ final io.papermc.paper.chunk.SingleThreadChunkRegionManager.Region newRegion) {
+ final DataRegionSectionData sectionData = (DataRegionSectionData)section.sectionData;
+ final DataRegionData oldRegionData = oldRegion == null ? null : (DataRegionData)oldRegion.regionData;
+ final DataRegionData newRegionData = (DataRegionData)newRegion.regionData;
+ }
+ }
+
+ public final ChunkHolder getUnloadingChunkHolder(int chunkX, int chunkZ) {
+ return this.pendingUnloads.get(io.papermc.paper.util.CoordinateUtils.getChunkKey(chunkX, chunkZ));
+ }
+ // Paper end
2021-06-11 14:02:28 +02:00
+
2022-06-07 20:12:34 +02:00
public ChunkMap(ServerLevel world, LevelStorageSource.LevelStorageAccess session, DataFixer dataFixer, StructureTemplateManager structureTemplateManager, Executor executor, BlockableEventLoop<Runnable> mainThreadExecutor, LightChunkGetter chunkProvider, ChunkGenerator chunkGenerator, ChunkProgressListener worldGenerationProgressListener, ChunkStatusUpdateListener chunkStatusChangeListener, Supplier<DimensionDataStorage> persistentStateManagerFactory, int viewDistance, boolean dsync) {
2021-11-23 11:51:25 +01:00
super(session.getDimensionPath(world.dimension()).resolve("region"), dataFixer, dsync);
2021-06-11 14:02:28 +02:00
this.visibleChunkMap = this.updatingChunkMap.clone();
2022-12-07 18:08:55 +01:00
@@ -226,6 +277,10 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
this.overworldDataStorage = persistentStateManagerFactory;
2022-12-07 18:08:55 +01:00
this.poiManager = new PoiManager(path.resolve("poi"), dataFixer, dsync, iregistrycustom, world);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
this.setViewDistance(viewDistance);
+ // Paper start
+ this.dataRegionManager = new io.papermc.paper.chunk.SingleThreadChunkRegionManager(this.level, 2, (1.0 / 3.0), 1, 6, "Data", DataRegionData::new, DataRegionSectionData::new);
+ this.regionManagers.add(this.dataRegionManager);
+ // Paper end
}
2021-11-23 11:51:25 +01:00
protected ChunkGenerator generator() {
2022-12-07 18:08:55 +01:00
@@ -323,6 +378,14 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
}
2021-06-11 14:02:28 +02:00
}
+ // Paper start
+ public final int getEffectiveViewDistance() {
+ // TODO this needs to be checked on update
+ // Mojang currently sets it to +1 of the configured view distance. So subtract one to get the one we really want.
+ return this.viewDistance - 1;
+ }
+ // Paper end
+
private CompletableFuture<Either<List<ChunkAccess>, ChunkHolder.ChunkLoadingFailure>> getChunkRangeFuture(ChunkPos centerChunk, int margin, IntFunction<ChunkStatus> distanceToStatus) {
List<CompletableFuture<Either<ChunkAccess, ChunkHolder.ChunkLoadingFailure>>> list = new ArrayList();
List<ChunkHolder> list1 = new ArrayList();
2022-12-07 18:08:55 +01:00
@@ -410,9 +473,9 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
};
stringbuilder.append("Updating:").append(System.lineSeparator());
- this.updatingChunkMap.values().forEach(consumer);
+ io.papermc.paper.chunk.system.ChunkSystem.getUpdatingChunkHolders(this.level).forEach(consumer); // Paper
stringbuilder.append("Visible:").append(System.lineSeparator());
- this.visibleChunkMap.values().forEach(consumer);
+ io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level).forEach(consumer); // Paper
CrashReport crashreport = CrashReport.forThrowable(exception, "Chunk loading");
CrashReportCategory crashreportsystemdetails = crashreport.addCategory("Chunk loading");
2022-12-07 18:08:55 +01:00
@@ -454,8 +517,14 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
holder.setTicketLevel(level);
} else {
holder = new ChunkHolder(new ChunkPos(pos), level, this.level, this.lightEngine, this.queueSorter, this);
+ // Paper start
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkHolderCreate(this.level, holder);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ // Paper end
}
+ // Paper start
+ holder.onChunkAdd();
+ // Paper end
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
this.updatingChunkMap.put(pos, holder);
this.modified = true;
}
2022-12-07 18:08:55 +01:00
@@ -477,7 +546,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
protected void saveAllChunks(boolean flush) {
if (flush) {
- List<ChunkHolder> list = (List) this.visibleChunkMap.values().stream().filter(ChunkHolder::wasAccessibleSinceLastSave).peek(ChunkHolder::refreshAccessibility).collect(Collectors.toList());
+ List<ChunkHolder> list = (List) io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level).stream().filter(ChunkHolder::wasAccessibleSinceLastSave).peek(ChunkHolder::refreshAccessibility).collect(Collectors.toList()); // Paper
MutableBoolean mutableboolean = new MutableBoolean();
do {
2022-12-07 18:08:55 +01:00
@@ -506,7 +575,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
});
this.flushWorker();
} else {
- this.visibleChunkMap.values().forEach(this::saveChunkIfNeeded);
+ io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level).forEach(this::saveChunkIfNeeded);
}
}
2022-12-07 18:08:55 +01:00
@@ -525,7 +594,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
}
public boolean hasWork() {
- return this.lightEngine.hasLightWork() || !this.pendingUnloads.isEmpty() || !this.updatingChunkMap.isEmpty() || this.poiManager.hasWork() || !this.toDrop.isEmpty() || !this.unloadQueue.isEmpty() || this.queueSorter.hasWork() || this.distanceManager.hasTickets();
+ return this.lightEngine.hasLightWork() || !this.pendingUnloads.isEmpty() || io.papermc.paper.chunk.system.ChunkSystem.hasAnyChunkHolders(this.level) || this.poiManager.hasWork() || !this.toDrop.isEmpty() || !this.unloadQueue.isEmpty() || this.queueSorter.hasWork() || this.distanceManager.hasTickets(); // Paper
}
private void processUnloads(BooleanSupplier shouldKeepTicking) {
2022-12-07 18:08:55 +01:00
@@ -536,6 +605,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
ChunkHolder playerchunk = (ChunkHolder) this.updatingChunkMap.remove(j);
if (playerchunk != null) {
+ playerchunk.onChunkRemove(); // Paper
this.pendingUnloads.put(j, playerchunk);
this.modified = true;
++i;
2022-12-07 18:08:55 +01:00
@@ -553,7 +623,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
}
int l = 0;
- ObjectIterator objectiterator = this.visibleChunkMap.values().iterator();
+ Iterator objectiterator = io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level).iterator(); // Paper
while (l < 20 && shouldKeepTicking.getAsBoolean() && objectiterator.hasNext()) {
if (this.saveChunkIfNeeded((ChunkHolder) objectiterator.next())) {
2022-12-07 18:08:55 +01:00
@@ -571,7 +641,11 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
if (completablefuture1 != completablefuture) {
this.scheduleUnload(pos, holder);
} else {
- if (this.pendingUnloads.remove(pos, holder) && ichunkaccess != null) {
+ // Paper start
+ boolean removed;
+ if ((removed = this.pendingUnloads.remove(pos, holder)) && ichunkaccess != null) {
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkHolderDelete(this.level, holder);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ // Paper end
if (ichunkaccess instanceof LevelChunk) {
((LevelChunk) ichunkaccess).setLoaded(false);
}
2022-12-07 18:08:55 +01:00
@@ -587,7 +661,9 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
this.lightEngine.tryScheduleUpdate();
this.progressListener.onStatusChange(ichunkaccess.getPos(), (ChunkStatus) null);
2022-03-01 06:43:03 +01:00
this.chunkSaveCooldowns.remove(ichunkaccess.getPos().toLong());
- }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ } else if (removed) { // Paper start
+ io.papermc.paper.chunk.system.ChunkSystem.onChunkHolderDelete(this.level, holder);
+ } // Paper end
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
}
};
2022-12-07 18:08:55 +01:00
@@ -968,7 +1044,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
this.viewDistance = j;
this.distanceManager.updatePlayerTickets(this.viewDistance + 1);
- ObjectIterator objectiterator = this.updatingChunkMap.values().iterator();
+ Iterator objectiterator = io.papermc.paper.chunk.system.ChunkSystem.getUpdatingChunkHolders(this.level).iterator(); // Paper
while (objectiterator.hasNext()) {
ChunkHolder playerchunk = (ChunkHolder) objectiterator.next();
2022-12-07 18:08:55 +01:00
@@ -1011,7 +1087,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
}
public int size() {
- return this.visibleChunkMap.size();
+ return io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolderCount(this.level); // Paper
}
public DistanceManager getDistanceManager() {
2022-12-07 18:08:55 +01:00
@@ -1019,19 +1095,19 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
}
protected Iterable<ChunkHolder> getChunks() {
- return Iterables.unmodifiableIterable(this.visibleChunkMap.values());
+ return Iterables.unmodifiableIterable(io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level)); // Paper
}
void dumpChunks(Writer writer) throws IOException {
CsvOutput csvwriter = CsvOutput.builder().addColumn("x").addColumn("z").addColumn("level").addColumn("in_memory").addColumn("status").addColumn("full_status").addColumn("accessible_ready").addColumn("ticking_ready").addColumn("entity_ticking_ready").addColumn("ticket").addColumn("spawning").addColumn("block_entity_count").addColumn("ticking_ticket").addColumn("ticking_level").addColumn("block_ticks").addColumn("fluid_ticks").build(writer);
TickingTracker tickingtracker = this.distanceManager.tickingTracker();
- ObjectBidirectionalIterator objectbidirectionaliterator = this.visibleChunkMap.long2ObjectEntrySet().iterator();
+ Iterator<ChunkHolder> objectbidirectionaliterator = io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.level).iterator(); // Paper
while (objectbidirectionaliterator.hasNext()) {
- Entry<ChunkHolder> entry = (Entry) objectbidirectionaliterator.next();
- long i = entry.getLongKey();
+ ChunkHolder playerchunk = objectbidirectionaliterator.next(); // Paper
+ long i = playerchunk.pos.toLong(); // Paper
ChunkPos chunkcoordintpair = new ChunkPos(i);
- ChunkHolder playerchunk = (ChunkHolder) entry.getValue();
+ // Paper
Optional<ChunkAccess> optional = Optional.ofNullable(playerchunk.getLastAvailable());
Optional<LevelChunk> optional1 = optional.flatMap((ichunkaccess) -> {
return ichunkaccess instanceof LevelChunk ? Optional.of((LevelChunk) ichunkaccess) : Optional.empty();
2022-12-07 18:08:55 +01:00
@@ -1157,6 +1233,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
2021-06-11 14:02:28 +02:00
if (!flag1) {
2022-06-07 20:12:34 +02:00
this.distanceManager.addPlayer(SectionPos.of((EntityAccess) player), player);
2021-06-11 14:02:28 +02:00
}
+ this.addPlayerToDistanceMaps(player); // Paper - distance maps
} else {
SectionPos sectionposition = player.getLastSectionPos();
2022-12-07 18:08:55 +01:00
@@ -1164,6 +1241,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
2021-06-11 14:02:28 +02:00
if (!flag2) {
this.distanceManager.removePlayer(sectionposition, player);
}
+ this.removePlayerFromDistanceMaps(player); // Paper - distance maps
}
2021-12-10 15:24:07 +01:00
for (int k = i - this.viewDistance - 1; k <= i + this.viewDistance + 1; ++k) {
2022-12-07 18:08:55 +01:00
@@ -1276,6 +1354,8 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
2021-06-11 14:02:28 +02:00
}
}
+ this.updateMaps(player); // Paper - distance maps
+
}
@Override
2022-12-08 05:24:59 +01:00
@@ -1502,7 +1582,7 @@ public class ChunkMap extends ChunkStorage implements ChunkHolder.PlayerProvider
private class ChunkDistanceManager extends DistanceManager {
protected ChunkDistanceManager(Executor workerExecutor, Executor mainThreadExecutor) {
- super(workerExecutor, mainThreadExecutor);
+ super(workerExecutor, mainThreadExecutor, ChunkMap.this);
}
@Override
diff --git a/src/main/java/net/minecraft/server/level/DistanceManager.java b/src/main/java/net/minecraft/server/level/DistanceManager.java
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
index 6c98676827ceb6999f340fa2b06a0b3e1cb4cae2..fbe62a31ab199d83a1db0a4e0b1a813824e6f2c2 100644
--- a/src/main/java/net/minecraft/server/level/DistanceManager.java
+++ b/src/main/java/net/minecraft/server/level/DistanceManager.java
2022-03-01 06:43:03 +01:00
@@ -60,8 +60,9 @@ public abstract class DistanceManager {
final Executor mainThreadExecutor;
private long ticketTickCounter;
private int simulationDistance = 10;
+ private final ChunkMap chunkMap; // Paper
- protected DistanceManager(Executor workerExecutor, Executor mainThreadExecutor) {
+ protected DistanceManager(Executor workerExecutor, Executor mainThreadExecutor, ChunkMap chunkMap) {
Objects.requireNonNull(mainThreadExecutor);
ProcessorHandle<Runnable> mailbox = ProcessorHandle.of("player ticket throttler", mainThreadExecutor::execute);
ChunkTaskPriorityQueueSorter chunktaskqueuesorter = new ChunkTaskPriorityQueueSorter(ImmutableList.of(mailbox), workerExecutor, 4);
2022-03-01 06:43:03 +01:00
@@ -70,6 +71,7 @@ public abstract class DistanceManager {
this.ticketThrottlerInput = chunktaskqueuesorter.getProcessor(mailbox, true);
this.ticketThrottlerReleaser = chunktaskqueuesorter.getReleaseProcessor(mailbox);
this.mainThreadExecutor = mainThreadExecutor;
+ this.chunkMap = chunkMap; // Paper
}
protected void purgeStaleTickets() {
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -319,6 +321,12 @@ public abstract class DistanceManager {
this.playerTicketManager.updateViewDistance(viewDistance);
}
+ // Paper start
+ public int getSimulationDistance() {
+ return this.simulationDistance;
+ }
+ // Paper end
+
public void updateSimulationDistance(int simulationDistance) {
if (simulationDistance != this.simulationDistance) {
this.simulationDistance = simulationDistance;
@@ -382,7 +390,7 @@ public abstract class DistanceManager {
}
public void removeTicketsOnClosing() {
- ImmutableSet<TicketType<?>> immutableset = ImmutableSet.of(TicketType.UNKNOWN, TicketType.POST_TELEPORT, TicketType.LIGHT);
+ ImmutableSet<TicketType<?>> immutableset = ImmutableSet.of(TicketType.UNKNOWN, TicketType.POST_TELEPORT, TicketType.LIGHT, TicketType.FUTURE_AWAIT); // Paper - add additional tickets to preserve
ObjectIterator objectiterator = this.tickets.long2ObjectEntrySet().fastIterator();
while (objectiterator.hasNext()) {
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/ServerChunkCache.java b/src/main/java/net/minecraft/server/level/ServerChunkCache.java
2022-12-07 18:08:55 +01:00
index b323b8329f534b7020dd595b8b15197c29939590..794ad2dbaea2555d4557124e9d942d3e6919ea09 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/ServerChunkCache.java
+++ b/src/main/java/net/minecraft/server/level/ServerChunkCache.java
2022-12-07 18:08:55 +01:00
@@ -51,6 +51,7 @@ import net.minecraft.world.level.storage.LevelStorageSource;
2021-06-11 14:02:28 +02:00
public class ServerChunkCache extends ChunkSource {
2022-03-01 06:43:03 +01:00
+ public static final org.slf4j.Logger LOGGER = com.mojang.logging.LogUtils.getLogger(); // Paper
private static final List<ChunkStatus> CHUNK_STATUSES = ChunkStatus.getStatusList();
2022-03-01 06:43:03 +01:00
private final DistanceManager distanceManager;
final ServerLevel level;
2022-12-07 18:08:55 +01:00
@@ -69,6 +70,231 @@ public class ServerChunkCache extends ChunkSource {
2021-06-11 14:02:28 +02:00
@Nullable
@VisibleForDebug
2021-06-11 14:02:28 +02:00
private NaturalSpawner.SpawnState lastSpawnState;
+ // Paper start
+ final com.destroystokyo.paper.util.concurrent.WeakSeqLock loadedChunkMapSeqLock = new com.destroystokyo.paper.util.concurrent.WeakSeqLock();
2022-03-01 06:43:03 +01:00
+ final it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap<LevelChunk> loadedChunkMap = new it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap<>(8192, 0.5f);
2021-06-11 14:02:28 +02:00
+
+ private final LevelChunk[] lastLoadedChunks = new LevelChunk[4 * 4];
+
+ private static int getChunkCacheKey(int x, int z) {
+ return x & 3 | ((z & 3) << 2);
+ }
+
+ public void addLoadedChunk(LevelChunk chunk) {
+ this.loadedChunkMapSeqLock.acquireWrite();
+ try {
+ this.loadedChunkMap.put(chunk.coordinateKey, chunk);
+ } finally {
+ this.loadedChunkMapSeqLock.releaseWrite();
+ }
+
+ // rewrite cache if we have to
+ // we do this since we also cache null chunks
+ int cacheKey = getChunkCacheKey(chunk.locX, chunk.locZ);
+
+ this.lastLoadedChunks[cacheKey] = chunk;
+ }
+
+ public void removeLoadedChunk(LevelChunk chunk) {
+ this.loadedChunkMapSeqLock.acquireWrite();
+ try {
+ this.loadedChunkMap.remove(chunk.coordinateKey);
+ } finally {
+ this.loadedChunkMapSeqLock.releaseWrite();
+ }
+
+ // rewrite cache if we have to
+ // we do this since we also cache null chunks
+ int cacheKey = getChunkCacheKey(chunk.locX, chunk.locZ);
+
+ LevelChunk cachedChunk = this.lastLoadedChunks[cacheKey];
+ if (cachedChunk != null && cachedChunk.coordinateKey == chunk.coordinateKey) {
+ this.lastLoadedChunks[cacheKey] = null;
+ }
+ }
+
+ public final LevelChunk getChunkAtIfLoadedMainThread(int x, int z) {
+ int cacheKey = getChunkCacheKey(x, z);
+
+ LevelChunk cachedChunk = this.lastLoadedChunks[cacheKey];
+ if (cachedChunk != null && cachedChunk.locX == x & cachedChunk.locZ == z) {
+ return cachedChunk;
2021-06-11 14:02:28 +02:00
+ }
+
+ long chunkKey = ChunkPos.asLong(x, z);
+
+ cachedChunk = this.loadedChunkMap.get(chunkKey);
+ // Skipping a null check to avoid extra instructions to improve inline capability
+ this.lastLoadedChunks[cacheKey] = cachedChunk;
+ return cachedChunk;
+ }
+
+ public final LevelChunk getChunkAtIfLoadedMainThreadNoCache(int x, int z) {
+ return this.loadedChunkMap.get(ChunkPos.asLong(x, z));
+ }
+
+ public final LevelChunk getChunkAtMainThread(int x, int z) {
+ LevelChunk ret = this.getChunkAtIfLoadedMainThread(x, z);
+ if (ret != null) {
+ return ret;
+ }
+ return (LevelChunk)this.getChunk(x, z, ChunkStatus.FULL, true);
+ }
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ long chunkFutureAwaitCounter; // Paper - private -> package private
2021-06-11 14:02:28 +02:00
+
+ public void getEntityTickingChunkAsync(int x, int z, java.util.function.Consumer<LevelChunk> onLoad) {
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleTickingState(
+ this.level, x, z, ChunkHolder.FullChunkStatus.ENTITY_TICKING, true,
+ ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL, onLoad
+ );
2021-06-11 14:02:28 +02:00
+ }
+
+ public void getTickingChunkAsync(int x, int z, java.util.function.Consumer<LevelChunk> onLoad) {
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleTickingState(
+ this.level, x, z, ChunkHolder.FullChunkStatus.TICKING, true,
+ ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL, onLoad
+ );
2021-06-11 14:02:28 +02:00
+ }
+
+ public void getFullChunkAsync(int x, int z, java.util.function.Consumer<LevelChunk> onLoad) {
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleTickingState(
+ this.level, x, z, ChunkHolder.FullChunkStatus.BORDER, true,
+ ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL, onLoad
+ );
2021-06-11 14:02:28 +02:00
+ }
+
+ void chunkLoadAccept(int chunkX, int chunkZ, ChunkAccess chunk, java.util.function.Consumer<ChunkAccess> consumer) {
+ try {
+ consumer.accept(chunk);
+ } catch (Throwable throwable) {
+ if (throwable instanceof ThreadDeath) {
+ throw (ThreadDeath)throwable;
+ }
+ LOGGER.error("Load callback for chunk " + chunkX + "," + chunkZ + " in world '" + this.level.getWorld().getName() + "' threw an exception", throwable);
+ }
+ }
+
+ void getChunkAtAsynchronously(int chunkX, int chunkZ, int ticketLevel,
+ java.util.function.Consumer<ChunkAccess> consumer) {
+ if (ticketLevel <= 33) {
+ this.getFullChunkAsync(chunkX, chunkZ, (java.util.function.Consumer)consumer);
+ return;
+ }
+
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleChunkLoad(
+ this.level, chunkX, chunkZ, ChunkHolder.getStatus(ticketLevel), true,
+ ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL, consumer
+ );
+ }
+
+
+ public final void getChunkAtAsynchronously(int chunkX, int chunkZ, ChunkStatus status, boolean gen, boolean allowSubTicketLevel, java.util.function.Consumer<ChunkAccess> onLoad) {
+ // try to fire sync
+ int chunkStatusTicketLevel = 33 + ChunkStatus.getDistance(status);
+ ChunkHolder playerChunk = this.chunkMap.getUpdatingChunkIfPresent(io.papermc.paper.util.CoordinateUtils.getChunkKey(chunkX, chunkZ));
+ if (playerChunk != null) {
+ ChunkStatus holderStatus = playerChunk.getChunkHolderStatus();
+ ChunkAccess immediate = playerChunk.getAvailableChunkNow();
+ if (immediate != null) {
+ if (allowSubTicketLevel ? immediate.getStatus().isOrAfter(status) : (playerChunk.getTicketLevel() <= chunkStatusTicketLevel && holderStatus != null && holderStatus.isOrAfter(status))) {
+ this.chunkLoadAccept(chunkX, chunkZ, immediate, onLoad);
+ return;
+ } else {
+ if (gen || (!allowSubTicketLevel && immediate.getStatus().isOrAfter(status))) {
+ this.getChunkAtAsynchronously(chunkX, chunkZ, chunkStatusTicketLevel, onLoad);
+ return;
+ } else {
+ this.chunkLoadAccept(chunkX, chunkZ, null, onLoad);
+ return;
+ }
+ }
+ }
+ }
+
+ // need to fire async
+
+ if (gen && !allowSubTicketLevel) {
+ this.getChunkAtAsynchronously(chunkX, chunkZ, chunkStatusTicketLevel, onLoad);
+ return;
+ }
+
+ this.getChunkAtAsynchronously(chunkX, chunkZ, io.papermc.paper.util.MCUtil.getTicketLevelFor(ChunkStatus.EMPTY), (ChunkAccess chunk) -> {
+ if (chunk == null) {
+ throw new IllegalStateException("Chunk cannot be null");
+ }
+
+ if (!chunk.getStatus().isOrAfter(status)) {
+ if (gen) {
+ this.getChunkAtAsynchronously(chunkX, chunkZ, chunkStatusTicketLevel, onLoad);
+ return;
+ } else {
+ ServerChunkCache.this.chunkLoadAccept(chunkX, chunkZ, null, onLoad);
+ return;
+ }
+ } else {
+ if (allowSubTicketLevel) {
+ ServerChunkCache.this.chunkLoadAccept(chunkX, chunkZ, chunk, onLoad);
+ return;
+ } else {
+ this.getChunkAtAsynchronously(chunkX, chunkZ, chunkStatusTicketLevel, onLoad);
+ return;
+ }
+ }
+ });
+ }
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ // Paper end
+
+ // Paper start
2021-11-23 12:47:17 +01:00
+ @Nullable
+ public ChunkAccess getChunkAtImmediately(int x, int z) {
+ ChunkHolder holder = this.chunkMap.getVisibleChunkIfPresent(ChunkPos.asLong(x, z));
+ if (holder == null) {
+ return null;
+ }
+
+ return holder.getLastAvailable();
+ }
+
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ // this will try to avoid chunk neighbours for lighting
+ public final ChunkAccess getFullStatusChunkAt(int chunkX, int chunkZ) {
+ LevelChunk ifLoaded = this.getChunkAtIfLoadedImmediately(chunkX, chunkZ);
+ if (ifLoaded != null) {
+ return ifLoaded;
+ }
+
+ ChunkAccess empty = this.getChunk(chunkX, chunkZ, ChunkStatus.EMPTY, true);
+ if (empty != null && empty.getStatus().isOrAfter(ChunkStatus.FULL)) {
+ return empty;
+ }
+ return this.getChunk(chunkX, chunkZ, ChunkStatus.FULL, true);
+ }
+
+ public final ChunkAccess getFullStatusChunkAtIfLoaded(int chunkX, int chunkZ) {
+ LevelChunk ifLoaded = this.getChunkAtIfLoadedImmediately(chunkX, chunkZ);
+ if (ifLoaded != null) {
+ return ifLoaded;
+ }
+
+ ChunkAccess ret = this.getChunkAtImmediately(chunkX, chunkZ);
+ if (ret != null && ret.getStatus().isOrAfter(ChunkStatus.FULL)) {
+ return ret;
+ } else {
+ return null;
+ }
+ }
+
2021-11-23 12:47:17 +01:00
+ public <T> void addTicketAtLevel(TicketType<T> ticketType, ChunkPos chunkPos, int ticketLevel, T identifier) {
2022-03-01 06:43:03 +01:00
+ this.distanceManager.addTicket(ticketType, chunkPos, ticketLevel, identifier);
2021-11-23 12:47:17 +01:00
+ }
+
+ public <T> void removeTicketAtLevel(TicketType<T> ticketType, ChunkPos chunkPos, int ticketLevel, T identifier) {
2022-03-01 06:43:03 +01:00
+ this.distanceManager.removeTicket(ticketType, chunkPos, ticketLevel, identifier);
2021-11-23 12:47:17 +01:00
+ }
+
+ public final io.papermc.paper.util.maplist.IteratorSafeOrderedReferenceSet<LevelChunk> tickingChunks = new io.papermc.paper.util.maplist.IteratorSafeOrderedReferenceSet<>(4096, 0.75f, 4096, 0.15, true);
+ public final io.papermc.paper.util.maplist.IteratorSafeOrderedReferenceSet<LevelChunk> entityTickingChunks = new io.papermc.paper.util.maplist.IteratorSafeOrderedReferenceSet<>(4096, 0.75f, 4096, 0.15, true);
2021-06-11 14:02:28 +02:00
+ // Paper end
2022-06-07 20:12:34 +02:00
public ServerChunkCache(ServerLevel world, LevelStorageSource.LevelStorageAccess session, DataFixer dataFixer, StructureTemplateManager structureTemplateManager, Executor workerExecutor, ChunkGenerator chunkGenerator, int viewDistance, int simulationDistance, boolean dsync, ChunkProgressListener worldGenerationProgressListener, ChunkStatusUpdateListener chunkStatusChangeListener, Supplier<DimensionDataStorage> persistentStateManagerFactory) {
this.level = world;
2022-12-07 18:08:55 +01:00
@@ -121,6 +347,49 @@ public class ServerChunkCache extends ChunkSource {
2021-06-11 14:02:28 +02:00
this.lastChunk[0] = chunk;
}
+ // Paper start - "real" get chunk if loaded
+ // Note: Partially copied from the getChunkAt method below
+ @Nullable
+ public LevelChunk getChunkAtIfCachedImmediately(int x, int z) {
+ long k = ChunkPos.asLong(x, z);
+
+ // Note: Bypass cache since we need to check ticket level, and to make this MT-Safe
+
+ ChunkHolder playerChunk = this.getVisibleChunkIfPresent(k);
+ if (playerChunk == null) {
+ return null;
+ }
+
+ return playerChunk.getFullChunkNowUnchecked();
2021-06-11 14:02:28 +02:00
+ }
+
+ @Nullable
+ public LevelChunk getChunkAtIfLoadedImmediately(int x, int z) {
+ long k = ChunkPos.asLong(x, z);
+
+ if (Thread.currentThread() == this.mainThread) {
+ return this.getChunkAtIfLoadedMainThread(x, z);
+ }
+
+ LevelChunk ret = null;
+ long readlock;
+ do {
+ readlock = this.loadedChunkMapSeqLock.acquireRead();
+ try {
+ ret = this.loadedChunkMap.get(k);
+ } catch (Throwable thr) {
+ if (thr instanceof ThreadDeath) {
+ throw (ThreadDeath)thr;
+ }
+ // re-try, this means a CME occurred...
+ continue;
+ }
+ } while (!this.loadedChunkMapSeqLock.tryReleaseRead(readlock));
+
+ return ret;
+ }
+ // Paper end
+
@Nullable
@Override
public ChunkAccess getChunk(int x, int z, ChunkStatus leastStatus, boolean create) {
2022-12-07 18:08:55 +01:00
@@ -328,6 +597,12 @@ public class ServerChunkCache extends ChunkSource {
}
}
+ // Paper start
+ public boolean isPositionTicking(Entity entity) {
+ return this.isPositionTicking(ChunkPos.asLong(net.minecraft.util.Mth.floor(entity.getX()) >> 4, net.minecraft.util.Mth.floor(entity.getZ()) >> 4));
+ }
+ // Paper end
+
public boolean isPositionTicking(long pos) {
ChunkHolder playerchunk = this.getVisibleChunkIfPresent(pos);
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/ServerLevel.java b/src/main/java/net/minecraft/server/level/ServerLevel.java
index 8309eb14140ebcf62ae7be1dfa0177ac7fcf83d7..8784b0702f10eb7582c309c15f18fe13eebd9c8e 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/ServerLevel.java
+++ b/src/main/java/net/minecraft/server/level/ServerLevel.java
2022-12-07 18:08:55 +01:00
@@ -171,6 +171,7 @@ import org.bukkit.event.weather.LightningStrikeEvent;
2022-06-07 20:12:34 +02:00
import org.bukkit.event.world.GenericGameEvent;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
import org.bukkit.event.world.TimeSkipEvent;
// CraftBukkit end
+import it.unimi.dsi.fastutil.ints.IntArrayList; // Paper
public class ServerLevel extends Level implements WorldGenLevel {
2022-12-07 18:08:55 +01:00
@@ -226,6 +227,98 @@ public class ServerLevel extends Level implements WorldGenLevel {
2021-11-23 11:51:25 +01:00
return convertable.dimensionType;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
}
+ // Paper start
+ public final boolean areChunksLoadedForMove(AABB axisalignedbb) {
+ // copied code from collision methods, so that we can guarantee that they wont load chunks (we don't override
+ // ICollisionAccess methods for VoxelShapes)
+ // be more strict too, add a block (dumb plugins in move events?)
+ int minBlockX = Mth.floor(axisalignedbb.minX - 1.0E-7D) - 3;
+ int maxBlockX = Mth.floor(axisalignedbb.maxX + 1.0E-7D) + 3;
+
+ int minBlockZ = Mth.floor(axisalignedbb.minZ - 1.0E-7D) - 3;
+ int maxBlockZ = Mth.floor(axisalignedbb.maxZ + 1.0E-7D) + 3;
+
+ int minChunkX = minBlockX >> 4;
+ int maxChunkX = maxBlockX >> 4;
+
+ int minChunkZ = minBlockZ >> 4;
+ int maxChunkZ = maxBlockZ >> 4;
+
+ ServerChunkCache chunkProvider = this.getChunkSource();
+
+ for (int cx = minChunkX; cx <= maxChunkX; ++cx) {
+ for (int cz = minChunkZ; cz <= maxChunkZ; ++cz) {
+ if (chunkProvider.getChunkAtIfLoadedImmediately(cx, cz) == null) {
+ return false;
+ }
+ }
+ }
+
+ return true;
+ }
+
+ public final void loadChunksForMoveAsync(AABB axisalignedbb, ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority priority,
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ java.util.function.Consumer<List<net.minecraft.world.level.chunk.ChunkAccess>> onLoad) {
+ if (Thread.currentThread() != this.thread) {
+ this.getChunkSource().mainThreadProcessor.execute(() -> {
+ this.loadChunksForMoveAsync(axisalignedbb, priority, onLoad);
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ });
+ return;
+ }
+ List<net.minecraft.world.level.chunk.ChunkAccess> ret = new java.util.ArrayList<>();
+ IntArrayList ticketLevels = new IntArrayList();
+
+ int minBlockX = Mth.floor(axisalignedbb.minX - 1.0E-7D) - 3;
+ int maxBlockX = Mth.floor(axisalignedbb.maxX + 1.0E-7D) + 3;
+
+ int minBlockZ = Mth.floor(axisalignedbb.minZ - 1.0E-7D) - 3;
+ int maxBlockZ = Mth.floor(axisalignedbb.maxZ + 1.0E-7D) + 3;
+
+ int minChunkX = minBlockX >> 4;
+ int maxChunkX = maxBlockX >> 4;
+
+ int minChunkZ = minBlockZ >> 4;
+ int maxChunkZ = maxBlockZ >> 4;
+
+ ServerChunkCache chunkProvider = this.getChunkSource();
+
+ int requiredChunks = (maxChunkX - minChunkX + 1) * (maxChunkZ - minChunkZ + 1);
+ int[] loadedChunks = new int[1];
+
+ Long holderIdentifier = Long.valueOf(chunkProvider.chunkFutureAwaitCounter++);
+
+ java.util.function.Consumer<net.minecraft.world.level.chunk.ChunkAccess> consumer = (net.minecraft.world.level.chunk.ChunkAccess chunk) -> {
+ if (chunk != null) {
+ int ticketLevel = Math.max(33, chunkProvider.chunkMap.getUpdatingChunkIfPresent(chunk.getPos().toLong()).getTicketLevel());
+ ret.add(chunk);
+ ticketLevels.add(ticketLevel);
+ chunkProvider.addTicketAtLevel(TicketType.FUTURE_AWAIT, chunk.getPos(), ticketLevel, holderIdentifier);
+ }
+ if (++loadedChunks[0] == requiredChunks) {
+ try {
+ onLoad.accept(java.util.Collections.unmodifiableList(ret));
+ } finally {
+ for (int i = 0, len = ret.size(); i < len; ++i) {
+ ChunkPos chunkPos = ret.get(i).getPos();
+ int ticketLevel = ticketLevels.getInt(i);
+
+ chunkProvider.addTicketAtLevel(TicketType.UNKNOWN, chunkPos, ticketLevel, chunkPos);
+ chunkProvider.removeTicketAtLevel(TicketType.FUTURE_AWAIT, chunkPos, ticketLevel, holderIdentifier);
+ }
+ }
+ }
+ };
+
+ for (int cx = minChunkX; cx <= maxChunkX; ++cx) {
+ for (int cz = minChunkZ; cz <= maxChunkZ; ++cz) {
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleChunkLoad(
+ this, cx, cz, net.minecraft.world.level.chunk.ChunkStatus.FULL, true, priority, consumer
+ );
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
+ }
+ }
+ }
+ // Paper end
+
2022-06-07 20:12:34 +02:00
// Add env and gen to constructor, IWorldDataServer -> WorldDataServer
public ServerLevel(MinecraftServer minecraftserver, Executor executor, LevelStorageSource.LevelStorageAccess convertable_conversionsession, PrimaryLevelData iworlddataserver, ResourceKey<Level> resourcekey, LevelStem worlddimension, ChunkProgressListener worldloadlistener, boolean flag, long i, List<CustomSpawner> list, boolean flag1, org.bukkit.World.Environment env, org.bukkit.generator.ChunkGenerator gen, org.bukkit.generator.BiomeProvider biomeProvider) {
2022-12-07 18:08:55 +01:00
// Holder holder = worlddimension.type(); // CraftBukkit - decompile error
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/ServerPlayer.java b/src/main/java/net/minecraft/server/level/ServerPlayer.java
index f7c5f0cd9d467305c123d3e86c5723055424600e..56016d59d66d90de2ac4326c16b09f86e4ac390b 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/ServerPlayer.java
+++ b/src/main/java/net/minecraft/server/level/ServerPlayer.java
2022-12-07 18:08:55 +01:00
@@ -247,6 +247,8 @@ public class ServerPlayer extends Player {
public Integer clientViewDistance;
public String kickLeaveMessage = null; // SPIGOT-3034: Forward leave message to PlayerQuitEvent
2021-06-11 14:02:28 +02:00
// CraftBukkit end
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
+ public boolean isRealPlayer; // Paper
2021-06-11 14:02:28 +02:00
+ public final com.destroystokyo.paper.util.misc.PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<ServerPlayer> cachedSingleHashSet; // Paper
2022-12-07 18:08:55 +01:00
public ServerPlayer(MinecraftServer server, ServerLevel world, GameProfile profile) {
super(world, world.getSharedSpawnPos(), world.getSharedSpawnAngle(), profile);
@@ -312,6 +314,8 @@ public class ServerPlayer extends Player {
this.maxUpStep = 1.0F;
2021-06-11 14:02:28 +02:00
this.fudgeSpawnLocation(world);
+ this.cachedSingleHashSet = new com.destroystokyo.paper.util.misc.PooledLinkedHashSets.PooledObjectLinkedOpenHashSet<>(this); // Paper
+
// CraftBukkit start
this.displayName = this.getScoreboardName();
this.bukkitPickUpLoot = true;
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/TicketType.java b/src/main/java/net/minecraft/server/level/TicketType.java
index 3a4f026c73cdd22d30bdadabbcf24bef969b73e4..0d536d72ac918fbd403397ff369d10143ee9c204 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/TicketType.java
+++ b/src/main/java/net/minecraft/server/level/TicketType.java
@@ -7,6 +7,7 @@ import net.minecraft.util.Unit;
import net.minecraft.world.level.ChunkPos;
public class TicketType<T> {
2021-06-11 14:02:28 +02:00
+ public static final TicketType<Long> FUTURE_AWAIT = create("future_await", Long::compareTo); // Paper
private final String name;
private final Comparator<T> comparator;
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/server/level/WorldGenRegion.java b/src/main/java/net/minecraft/server/level/WorldGenRegion.java
2022-12-07 18:08:55 +01:00
index a63d5ba706a5b8e430aedc045bdeb3a410bd0eef..e96a0ca47e4701ba187555bd92c968345bc85677 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/server/level/WorldGenRegion.java
+++ b/src/main/java/net/minecraft/server/level/WorldGenRegion.java
2022-12-07 18:08:55 +01:00
@@ -160,6 +160,26 @@ public class WorldGenRegion implements WorldGenLevel {
2021-06-11 14:02:28 +02:00
return chunkX >= this.firstPos.x && chunkX <= this.lastPos.x && chunkZ >= this.firstPos.z && chunkZ <= this.lastPos.z;
}
+ // Paper start - if loaded util
+ @Nullable
+ @Override
+ public ChunkAccess getChunkIfLoadedImmediately(int x, int z) {
+ return this.getChunk(x, z, ChunkStatus.FULL, false);
+ }
+
+ @Override
+ public final BlockState getBlockStateIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ ChunkAccess chunk = this.getChunkIfLoadedImmediately(blockposition.getX() >> 4, blockposition.getZ() >> 4);
+ return chunk == null ? null : chunk.getBlockState(blockposition);
+ }
+
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ ChunkAccess chunk = this.getChunkIfLoadedImmediately(blockposition.getX() >> 4, blockposition.getZ() >> 4);
+ return chunk == null ? null : chunk.getFluidState(blockposition);
+ }
+ // Paper end
+
@Override
public BlockState getBlockState(BlockPos pos) {
return this.getChunk(SectionPos.blockToSectionCoord(pos.getX()), SectionPos.blockToSectionCoord(pos.getZ())).getBlockState(pos);
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
diff --git a/src/main/java/net/minecraft/server/players/PlayerList.java b/src/main/java/net/minecraft/server/players/PlayerList.java
index 029d0197863a6975a7b42132bd0b734025299a8f..25227822aa0e1dd7fbbe98a0925ccd023af8d0dc 100644
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
--- a/src/main/java/net/minecraft/server/players/PlayerList.java
+++ b/src/main/java/net/minecraft/server/players/PlayerList.java
2022-12-07 18:08:55 +01:00
@@ -181,6 +181,7 @@ public abstract class PlayerList {
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
}
public void placeNewPlayer(Connection connection, ServerPlayer player) {
+ player.isRealPlayer = true; // Paper
GameProfile gameprofile = player.getGameProfile();
GameProfileCache usercache = this.server.getProfileCache();
Optional<GameProfile> optional = usercache.get(gameprofile.getId());
2021-06-12 18:56:13 +02:00
diff --git a/src/main/java/net/minecraft/util/thread/BlockableEventLoop.java b/src/main/java/net/minecraft/util/thread/BlockableEventLoop.java
index 337e0a7b3c14e1b1a28744920e0dc0a69e0c5a87..f5829ae484d93b547a5437b85a9621346384a11b 100644
2021-06-12 18:56:13 +02:00
--- a/src/main/java/net/minecraft/util/thread/BlockableEventLoop.java
+++ b/src/main/java/net/minecraft/util/thread/BlockableEventLoop.java
@@ -78,6 +78,13 @@ public abstract class BlockableEventLoop<R extends Runnable> implements Profiler
}
}
+ // Paper start
+ public void scheduleOnMain(Runnable r0) {
+ // postToMainThread does not work the same as older versions of mc
+ // This method is actually used to create a TickTask, which can then be posted onto main
+ this.tell(this.wrapRunnable(r0));
+ }
+ // Paper end
@Override
public void tell(R runnable) {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
diff --git a/src/main/java/net/minecraft/world/entity/Entity.java b/src/main/java/net/minecraft/world/entity/Entity.java
2022-12-07 18:08:55 +01:00
index df88a36981524654fb2905782b4b1337ea261a03..742897a534dac2bbbacaca0f6c8196e8d2bc03dd 100644
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- a/src/main/java/net/minecraft/world/entity/Entity.java
+++ b/src/main/java/net/minecraft/world/entity/Entity.java
@@ -315,6 +315,11 @@ public abstract class Entity implements Nameable, EntityAccess, CommandSource {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
return this.level.hasChunk((int) Math.floor(this.getX()) >> 4, (int) Math.floor(this.getZ()) >> 4);
}
// CraftBukkit end
+ // Paper start
+ public final AABB getBoundingBoxAt(double x, double y, double z) {
+ return this.dimensions.makeBoundingBox(x, y, z);
+ }
+ // Paper end
public Entity(EntityType<?> type, Level world) {
this.id = Entity.ENTITY_COUNTER.incrementAndGet();
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/world/entity/LivingEntity.java b/src/main/java/net/minecraft/world/entity/LivingEntity.java
2022-12-07 18:08:55 +01:00
index 087cd12b08732a4c62ac4be417b17b1db830097a..f29b109fdef494927f5f894ac962c50c33b1f0b6 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/entity/LivingEntity.java
+++ b/src/main/java/net/minecraft/world/entity/LivingEntity.java
2022-12-07 18:08:55 +01:00
@@ -261,6 +261,7 @@ public abstract class LivingEntity extends Entity {
2021-06-11 14:02:28 +02:00
public boolean collides = true;
public Set<UUID> collidableExemptions = new HashSet<>();
public boolean bukkitPickUpLoot;
2021-06-11 14:02:28 +02:00
+ public org.bukkit.craftbukkit.entity.CraftLivingEntity getBukkitLivingEntity() { return (org.bukkit.craftbukkit.entity.CraftLivingEntity) super.getBukkitEntity(); } // Paper
@Override
public float getBukkitYaw() {
diff --git a/src/main/java/net/minecraft/world/entity/Mob.java b/src/main/java/net/minecraft/world/entity/Mob.java
2022-12-07 18:08:55 +01:00
index ffac9a8678989decc276b0fa42e87da1d97d3666..10949455cc92ed6cb537fd7f148714c7ade4e360 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/entity/Mob.java
+++ b/src/main/java/net/minecraft/world/entity/Mob.java
2022-06-07 20:12:34 +02:00
@@ -242,6 +242,7 @@ public abstract class Mob extends LivingEntity {
2021-06-11 14:02:28 +02:00
return this.target;
}
+ public org.bukkit.craftbukkit.entity.CraftMob getBukkitMob() { return (org.bukkit.craftbukkit.entity.CraftMob) super.getBukkitEntity(); } // Paper
public void setTarget(@Nullable LivingEntity target) {
// CraftBukkit start - fire event
2021-11-23 11:51:25 +01:00
this.setTarget(target, EntityTargetEvent.TargetReason.UNKNOWN, true);
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/world/entity/PathfinderMob.java b/src/main/java/net/minecraft/world/entity/PathfinderMob.java
2022-12-07 18:08:55 +01:00
index fbe69ec9cded766c9b76ed4a1bcba6d4f49d6165..6ae3f5cd42dfd424fc3741957995f47ad5ec8941 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/entity/PathfinderMob.java
+++ b/src/main/java/net/minecraft/world/entity/PathfinderMob.java
2021-11-23 11:51:25 +01:00
@@ -18,6 +18,8 @@ public abstract class PathfinderMob extends Mob {
super(type, world);
}
2021-06-11 14:02:28 +02:00
+ public org.bukkit.craftbukkit.entity.CraftCreature getBukkitCreature() { return (org.bukkit.craftbukkit.entity.CraftCreature) super.getBukkitEntity(); } // Paper
+
2021-11-23 11:51:25 +01:00
public float getWalkTargetValue(BlockPos pos) {
return this.getWalkTargetValue(pos, this.level);
2021-06-11 14:02:28 +02:00
}
diff --git a/src/main/java/net/minecraft/world/entity/monster/Monster.java b/src/main/java/net/minecraft/world/entity/monster/Monster.java
2022-06-07 20:12:34 +02:00
index a0b5895abc88d297045e05f25bb09527991d43f0..6e0bd0eab0b06a4ac3042496bbb91292544e9f3c 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/entity/monster/Monster.java
+++ b/src/main/java/net/minecraft/world/entity/monster/Monster.java
2022-06-07 20:12:34 +02:00
@@ -27,6 +27,7 @@ import net.minecraft.world.level.ServerLevelAccessor;
import net.minecraft.world.level.dimension.DimensionType;
2021-06-11 14:02:28 +02:00
public abstract class Monster extends PathfinderMob implements Enemy {
+ public org.bukkit.craftbukkit.entity.CraftMonster getBukkitMonster() { return (org.bukkit.craftbukkit.entity.CraftMonster) super.getBukkitEntity(); } // Paper
protected Monster(EntityType<? extends Monster> type, Level world) {
super(type, world);
this.xpReward = 5;
diff --git a/src/main/java/net/minecraft/world/item/ItemStack.java b/src/main/java/net/minecraft/world/item/ItemStack.java
index 6439fcbbd31158ce7a8ccf191b47d95911c8c3c2..cf810a1d645dd63595702a5f07456cd48393ee79 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/item/ItemStack.java
+++ b/src/main/java/net/minecraft/world/item/ItemStack.java
@@ -759,6 +759,25 @@ public final class ItemStack {
2021-06-11 14:02:28 +02:00
return this.tag != null ? this.tag.getList("Enchantments", 10) : new ListTag();
}
+ // Paper start - (this is just a good no conflict location)
+ public org.bukkit.inventory.ItemStack asBukkitMirror() {
+ return CraftItemStack.asCraftMirror(this);
+ }
+ public org.bukkit.inventory.ItemStack asBukkitCopy() {
+ return CraftItemStack.asCraftMirror(this.copy());
+ }
+ public static ItemStack fromBukkitCopy(org.bukkit.inventory.ItemStack itemstack) {
+ return CraftItemStack.asNMSCopy(itemstack);
+ }
+ private org.bukkit.craftbukkit.inventory.CraftItemStack bukkitStack;
+ public org.bukkit.inventory.ItemStack getBukkitStack() {
2021-06-16 00:24:12 +02:00
+ if (bukkitStack == null || bukkitStack.handle != this) {
2021-06-11 14:02:28 +02:00
+ bukkitStack = org.bukkit.craftbukkit.inventory.CraftItemStack.asCraftMirror(this);
+ }
+ return bukkitStack;
+ }
+ // Paper end
2021-11-23 11:51:25 +01:00
+
public void setTag(@Nullable CompoundTag nbt) {
this.tag = nbt;
2021-06-11 14:02:28 +02:00
if (this.getItem().canBeDepleted()) {
@@ -1145,6 +1164,7 @@ public final class ItemStack {
2021-06-11 14:02:28 +02:00
// CraftBukkit start
@Deprecated
public void setItem(Item item) {
+ this.bukkitStack = null; // Paper
this.item = item;
}
// CraftBukkit end
diff --git a/src/main/java/net/minecraft/world/level/BlockGetter.java b/src/main/java/net/minecraft/world/level/BlockGetter.java
index 76251b5adf41f8e5bf2c07145abe3108fcde8669..8a979600b49e8a11982577fb6dd79503e2521a0f 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/BlockGetter.java
+++ b/src/main/java/net/minecraft/world/level/BlockGetter.java
@@ -9,10 +9,12 @@ import javax.annotation.Nullable;
2021-06-11 14:02:28 +02:00
import net.minecraft.core.BlockPos;
import net.minecraft.core.Direction;
import net.minecraft.util.Mth;
+import net.minecraft.world.level.block.Block;
import net.minecraft.world.level.block.entity.BlockEntity;
import net.minecraft.world.level.block.entity.BlockEntityType;
2021-06-11 14:02:28 +02:00
import net.minecraft.world.level.block.state.BlockState;
import net.minecraft.world.level.material.FluidState;
+import net.minecraft.world.level.material.Material;
import net.minecraft.world.phys.AABB;
import net.minecraft.world.phys.BlockHitResult;
import net.minecraft.world.phys.Vec3;
@@ -30,6 +32,19 @@ public interface BlockGetter extends LevelHeightAccessor {
}
2021-06-11 14:02:28 +02:00
BlockState getBlockState(BlockPos pos);
+ // Paper start - if loaded util
+ @Nullable BlockState getBlockStateIfLoaded(BlockPos blockposition);
+ default @Nullable Material getMaterialIfLoaded(BlockPos blockposition) {
+ BlockState type = this.getBlockStateIfLoaded(blockposition);
2021-06-11 14:02:28 +02:00
+ return type == null ? null : type.getMaterial();
+ }
+
+ default @Nullable Block getBlockIfLoaded(BlockPos blockposition) {
+ BlockState type = this.getBlockStateIfLoaded(blockposition);
2021-06-11 14:02:28 +02:00
+ return type == null ? null : type.getBlock();
+ }
+ @Nullable FluidState getFluidIfLoaded(BlockPos blockposition);
2021-06-11 14:02:28 +02:00
+ // Paper end
FluidState getFluidState(BlockPos pos);
diff --git a/src/main/java/net/minecraft/world/level/ChunkPos.java b/src/main/java/net/minecraft/world/level/ChunkPos.java
2022-06-07 20:12:34 +02:00
index a3040440ed34a1c2adca2d57d458504a4a48282f..2d41f619577b41d6420159668bbab70fb6c762eb 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/ChunkPos.java
+++ b/src/main/java/net/minecraft/world/level/ChunkPos.java
2022-06-07 20:12:34 +02:00
@@ -20,6 +20,7 @@ public class ChunkPos {
public static final int REGION_MAX_INDEX = 31;
2021-06-11 14:02:28 +02:00
public final int x;
public final int z;
+ public final long longKey; // Paper
private static final int HASH_A = 1664525;
private static final int HASH_C = 1013904223;
private static final int HASH_Z_XOR = -559038737;
2022-06-07 20:12:34 +02:00
@@ -27,16 +28,19 @@ public class ChunkPos {
2021-06-11 14:02:28 +02:00
public ChunkPos(int x, int z) {
this.x = x;
this.z = z;
+ this.longKey = asLong(this.x, this.z); // Paper
}
public ChunkPos(BlockPos pos) {
this.x = SectionPos.blockToSectionCoord(pos.getX());
this.z = SectionPos.blockToSectionCoord(pos.getZ());
2021-06-11 14:02:28 +02:00
+ this.longKey = asLong(this.x, this.z); // Paper
}
public ChunkPos(long pos) {
this.x = (int)pos;
this.z = (int)(pos >> 32);
2021-06-11 14:02:28 +02:00
+ this.longKey = asLong(this.x, this.z); // Paper
}
2022-06-07 20:12:34 +02:00
public static ChunkPos minFromRegion(int x, int z) {
@@ -48,10 +52,10 @@ public class ChunkPos {
}
2021-06-11 14:02:28 +02:00
public long toLong() {
- return asLong(this.x, this.z);
+ return longKey; // Paper
}
- public static long asLong(int chunkX, int chunkZ) {
+ public static long asLong(int chunkX, int chunkZ) {
return (long)chunkX & 4294967295L | ((long)chunkZ & 4294967295L) << 32;
2021-06-11 14:02:28 +02:00
}
diff --git a/src/main/java/net/minecraft/world/level/EmptyBlockGetter.java b/src/main/java/net/minecraft/world/level/EmptyBlockGetter.java
index 3c707d6674b2594b09503b959a31c1f4ad3981e6..db61b6b0158a9bcc0e1d735e34fe3671f8c89e21 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/EmptyBlockGetter.java
+++ b/src/main/java/net/minecraft/world/level/EmptyBlockGetter.java
@@ -17,6 +17,18 @@ public enum EmptyBlockGetter implements BlockGetter {
2021-06-11 14:02:28 +02:00
return null;
}
+ // Paper start - If loaded util
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
+ return Fluids.EMPTY.defaultFluidState();
2021-06-11 14:02:28 +02:00
+ }
+
+ @Override
+ public final BlockState getBlockStateIfLoaded(BlockPos blockposition) {
+ return Blocks.AIR.defaultBlockState();
2021-06-11 14:02:28 +02:00
+ }
+ // Paper end
+
@Override
public BlockState getBlockState(BlockPos pos) {
return Blocks.AIR.defaultBlockState();
diff --git a/src/main/java/net/minecraft/world/level/Level.java b/src/main/java/net/minecraft/world/level/Level.java
index ca5cbd7c43d7e263f76d8719a259514c0748a430..a0bd6c4df76fac867e1862273ce31dbe067de624 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/Level.java
+++ b/src/main/java/net/minecraft/world/level/Level.java
2022-12-07 18:08:55 +01:00
@@ -88,6 +88,7 @@ import org.bukkit.craftbukkit.CraftServer;
2021-06-11 14:02:28 +02:00
import org.bukkit.craftbukkit.CraftWorld;
import org.bukkit.craftbukkit.SpigotTimings; // Spigot
2021-06-11 14:02:28 +02:00
import org.bukkit.craftbukkit.block.CapturedBlockState;
+import org.bukkit.craftbukkit.block.CraftBlockState;
import org.bukkit.craftbukkit.block.data.CraftBlockData;
import org.bukkit.craftbukkit.util.CraftSpawnCategory;
import org.bukkit.craftbukkit.util.CraftNamespacedKey;
@@ -287,18 +288,52 @@ public abstract class Level implements LevelAccessor, AutoCloseable {
return y < -20000000 || y >= 20000000;
2021-06-11 14:02:28 +02:00
}
- public LevelChunk getChunkAt(BlockPos pos) {
+ public final LevelChunk getChunkAt(BlockPos pos) { // Paper - help inline
return this.getChunk(SectionPos.blockToSectionCoord(pos.getX()), SectionPos.blockToSectionCoord(pos.getZ()));
2021-06-11 14:02:28 +02:00
}
@Override
- public LevelChunk getChunk(int chunkX, int chunkZ) {
- return (LevelChunk) this.getChunk(chunkX, chunkZ, ChunkStatus.FULL);
+ public final LevelChunk getChunk(int chunkX, int chunkZ) { // Paper - final to help inline
+ return (LevelChunk) this.getChunk(chunkX, chunkZ, ChunkStatus.FULL, true); // Paper - avoid a method jump
}
+ // Paper start - if loaded
@Nullable
2021-06-11 14:02:28 +02:00
@Override
- public ChunkAccess getChunk(int chunkX, int chunkZ, ChunkStatus leastStatus, boolean create) {
+ public final ChunkAccess getChunkIfLoadedImmediately(int x, int z) {
+ return ((ServerLevel)this).chunkSource.getChunkAtIfLoadedImmediately(x, z);
+ }
+
+ @Override
+ @Nullable
+ public final BlockState getBlockStateIfLoaded(BlockPos pos) {
2021-06-11 14:02:28 +02:00
+ // CraftBukkit start - tree generation
+ if (this.captureTreeGeneration) {
+ CraftBlockState previous = this.capturedBlockStates.get(pos);
2021-06-11 14:02:28 +02:00
+ if (previous != null) {
+ return previous.getHandle();
+ }
+ }
+ // CraftBukkit end
+ if (this.isOutsideBuildHeight(pos)) {
+ return Blocks.VOID_AIR.defaultBlockState();
+ } else {
+ ChunkAccess chunk = this.getChunkIfLoadedImmediately(pos.getX() >> 4, pos.getZ() >> 4);
2021-06-11 14:02:28 +02:00
+
+ return chunk == null ? null : chunk.getBlockState(pos);
+ }
2021-06-11 14:02:28 +02:00
+ }
+
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ ChunkAccess chunk = this.getChunkIfLoadedImmediately(blockposition.getX() >> 4, blockposition.getZ() >> 4);
+
+ return chunk == null ? null : chunk.getFluidState(blockposition);
+ }
+
+ @Override
+ public final ChunkAccess getChunk(int chunkX, int chunkZ, ChunkStatus leastStatus, boolean create) { // Paper - final for inline
+ // Paper end
2021-06-11 14:02:28 +02:00
ChunkAccess ichunkaccess = this.getChunkSource().getChunk(chunkX, chunkZ, leastStatus, create);
if (ichunkaccess == null && create) {
@@ -309,7 +344,7 @@ public abstract class Level implements LevelAccessor, AutoCloseable {
2021-06-11 14:02:28 +02:00
}
@Override
- public boolean setBlock(BlockPos pos, BlockState state, int flags) {
+ public final boolean setBlock(BlockPos pos, BlockState state, int flags) { // Paper - final for inline
return this.setBlock(pos, state, flags, 512);
}
@@ -553,7 +588,7 @@ public abstract class Level implements LevelAccessor, AutoCloseable {
if (this.isOutsideBuildHeight(pos)) {
2021-06-11 14:02:28 +02:00
return Blocks.VOID_AIR.defaultBlockState();
} else {
- LevelChunk chunk = this.getChunk(SectionPos.blockToSectionCoord(pos.getX()), SectionPos.blockToSectionCoord(pos.getZ()));
+ ChunkAccess chunk = this.getChunk(pos.getX() >> 4, pos.getZ() >> 4, ChunkStatus.FULL, true); // Paper - manually inline to reduce hops and avoid unnecessary null check to reduce total byte code size, this should never return null and if it does we will see it the next line but the real stack trace will matter in the chunk engine
2021-06-11 14:02:28 +02:00
return chunk.getBlockState(pos);
}
diff --git a/src/main/java/net/minecraft/world/level/LevelReader.java b/src/main/java/net/minecraft/world/level/LevelReader.java
2022-12-07 18:08:55 +01:00
index f5dbac0a13d23413dbdc48cfacc247ef25ef9444..7fe1b8856bf916796fa6d2a984f0a07a2331e23b 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/LevelReader.java
+++ b/src/main/java/net/minecraft/world/level/LevelReader.java
2022-12-07 18:08:55 +01:00
@@ -24,6 +24,7 @@ import net.minecraft.world.level.levelgen.Heightmap;
import net.minecraft.world.phys.AABB;
2021-06-11 14:02:28 +02:00
public interface LevelReader extends BlockAndTintGetter, CollisionGetter, BiomeManager.NoiseBiomeSource {
+ @Nullable ChunkAccess getChunkIfLoadedImmediately(int x, int z); // Paper - ifLoaded api (we need this since current impl blocks if the chunk is loading)
@Nullable
ChunkAccess getChunk(int chunkX, int chunkZ, ChunkStatus leastStatus, boolean create);
diff --git a/src/main/java/net/minecraft/world/level/PathNavigationRegion.java b/src/main/java/net/minecraft/world/level/PathNavigationRegion.java
2022-12-07 18:08:55 +01:00
index 249b3ed33672a9a9529bd14de978722b62019314..0f1025495237aebe30132ace0832aa5718d6f9bb 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/PathNavigationRegion.java
+++ b/src/main/java/net/minecraft/world/level/PathNavigationRegion.java
2022-03-01 06:43:03 +01:00
@@ -9,6 +9,7 @@ import net.minecraft.core.Holder;
import net.minecraft.core.SectionPos;
2022-12-07 18:08:55 +01:00
import net.minecraft.core.registries.Registries;
import net.minecraft.util.profiling.ProfilerFiller;
2021-06-11 14:02:28 +02:00
+import net.minecraft.server.level.ServerLevel;
import net.minecraft.world.entity.Entity;
2022-03-01 06:43:03 +01:00
import net.minecraft.world.level.biome.Biome;
import net.minecraft.world.level.biome.Biomes;
@@ -70,7 +71,7 @@ public class PathNavigationRegion implements BlockGetter, CollisionGetter {
private ChunkAccess getChunk(int chunkX, int chunkZ) {
int i = chunkX - this.centerX;
int j = chunkZ - this.centerZ;
- if (i >= 0 && i < this.chunks.length && j >= 0 && j < this.chunks[i].length) {
+ if (i >= 0 && i < this.chunks.length && j >= 0 && j < this.chunks[i].length) { // Paper - if this changes, update getChunkIfLoaded below
ChunkAccess chunkAccess = this.chunks[i][j];
2022-03-01 06:43:03 +01:00
return (ChunkAccess)(chunkAccess != null ? chunkAccess : new EmptyLevelChunk(this.level, new ChunkPos(chunkX, chunkZ), this.plains.get()));
} else {
2022-03-01 06:43:03 +01:00
@@ -78,6 +79,30 @@ public class PathNavigationRegion implements BlockGetter, CollisionGetter {
}
2021-06-11 14:02:28 +02:00
}
+ // Paper start - if loaded util
+ private @Nullable ChunkAccess getChunkIfLoaded(int x, int z) {
+ // Based on getChunk(int, int)
+ int xx = x - this.centerX;
+ int zz = z - this.centerZ;
2021-06-11 14:02:28 +02:00
+
+ if (xx >= 0 && xx < this.chunks.length && zz >= 0 && zz < this.chunks[xx].length) {
+ return this.chunks[xx][zz];
2021-06-11 14:02:28 +02:00
+ }
+ return null;
+ }
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ ChunkAccess chunk = getChunkIfLoaded(blockposition.getX() >> 4, blockposition.getZ() >> 4);
+ return chunk == null ? null : chunk.getFluidState(blockposition);
+ }
+
+ @Override
+ public final BlockState getBlockStateIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ ChunkAccess chunk = getChunkIfLoaded(blockposition.getX() >> 4, blockposition.getZ() >> 4);
+ return chunk == null ? null : chunk.getBlockState(blockposition);
+ }
+ // Paper end
+
@Override
public WorldBorder getWorldBorder() {
return this.level.getWorldBorder();
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
diff --git a/src/main/java/net/minecraft/world/level/block/state/BlockBehaviour.java b/src/main/java/net/minecraft/world/level/block/state/BlockBehaviour.java
2022-12-07 18:08:55 +01:00
index ec86c786f793da00fdcee7b72f989068cbd750e4..aa2f76ff47699f23474f2e409ef9b773e48d50b8 100644
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
--- a/src/main/java/net/minecraft/world/level/block/state/BlockBehaviour.java
+++ b/src/main/java/net/minecraft/world/level/block/state/BlockBehaviour.java
2022-12-07 18:08:55 +01:00
@@ -686,14 +686,14 @@ public abstract class BlockBehaviour implements FeatureElement {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
public abstract static class BlockStateBase extends StateHolder<Block, BlockState> {
- private final int lightEmission;
- private final boolean useShapeForLightOcclusion;
+ private final int lightEmission; public final int getEmittedLight() { return this.lightEmission; } // Paper - OBFHELPER
+ private final boolean useShapeForLightOcclusion; public final boolean isTransparentOnSomeFaces() { return this.useShapeForLightOcclusion; } // Paper - OBFHELPER
private final boolean isAir;
private final Material material;
private final MaterialColor materialColor;
public final float destroySpeed;
private final boolean requiresCorrectToolForDrops;
- private final boolean canOcclude;
+ private final boolean canOcclude; public final boolean isOpaque() { return this.canOcclude; } // Paper - OBFHELPER
private final BlockBehaviour.StatePredicate isRedstoneConductor;
private final BlockBehaviour.StatePredicate isSuffocating;
private final BlockBehaviour.StatePredicate isViewBlocking;
2022-12-07 18:08:55 +01:00
@@ -728,12 +728,20 @@ public abstract class BlockBehaviour implements FeatureElement {
this.spawnParticlesOnBreak = blockbase_info.spawnParticlesOnBreak;
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
}
+ // Paper start
+ protected boolean shapeExceedsCube = true;
+ public final boolean shapeExceedsCube() {
+ return this.shapeExceedsCube;
+ }
+ // Paper end
+
public void initCache() {
2022-12-07 18:08:55 +01:00
this.fluidState = ((Block) this.owner).getFluidState(this.asState());
this.isRandomlyTicking = ((Block) this.owner).isRandomlyTicking(this.asState());
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
if (!this.getBlock().hasDynamicShape()) {
this.cache = new BlockBehaviour.BlockStateBase.Cache(this.asState());
}
+ this.shapeExceedsCube = this.cache == null || this.cache.largeCollisionShape; // Paper - moved from actual method to here
}
2022-12-07 18:08:55 +01:00
@@ -769,8 +777,8 @@ public abstract class BlockBehaviour implements FeatureElement {
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
return this.getBlock().getOcclusionShape(this.asState(), world, pos);
}
- public boolean hasLargeCollisionShape() {
- return this.cache == null || this.cache.largeCollisionShape;
+ public final boolean hasLargeCollisionShape() { // Paper
+ return this.shapeExceedsCube; // Paper - moved into shape cache init
}
public boolean useShapeForLightOcclusion() {
2021-11-23 11:51:25 +01:00
diff --git a/src/main/java/net/minecraft/world/level/chunk/ChunkAccess.java b/src/main/java/net/minecraft/world/level/chunk/ChunkAccess.java
2022-12-07 18:08:55 +01:00
index db6a64ae4437b76c39e7ddb02adbea27c95fde78..3fdbb777d4722596cc4df79b2d4d7b9c553580fd 100644
2021-11-23 11:51:25 +01:00
--- a/src/main/java/net/minecraft/world/level/chunk/ChunkAccess.java
+++ b/src/main/java/net/minecraft/world/level/chunk/ChunkAccess.java
2022-06-07 20:12:34 +02:00
@@ -58,7 +58,7 @@ public abstract class ChunkAccess implements BlockGetter, BiomeManager.NoiseBiom
2021-11-23 11:51:25 +01:00
protected final ShortList[] postProcessing;
protected volatile boolean unsaved;
private volatile boolean isLightCorrect;
- protected final ChunkPos chunkPos;
+ protected final ChunkPos chunkPos; public final long coordinateKey; public final int locX; public final int locZ; // Paper - cache coordinate key
private long inhabitedTime;
/** @deprecated */
2021-11-23 11:51:25 +01:00
@Nullable
2022-06-07 20:12:34 +02:00
@@ -83,7 +83,8 @@ public abstract class ChunkAccess implements BlockGetter, BiomeManager.NoiseBiom
// CraftBukkit end
2021-11-23 11:51:25 +01:00
public ChunkAccess(ChunkPos pos, UpgradeData upgradeData, LevelHeightAccessor heightLimitView, Registry<Biome> biome, long inhabitedTime, @Nullable LevelChunkSection[] sectionArrayInitializer, @Nullable BlendingData blendingData) {
- this.chunkPos = pos;
+ this.locX = pos.x; this.locZ = pos.z; // Paper - reduce need for field lookups
+ this.chunkPos = pos; this.coordinateKey = ChunkPos.asLong(locX, locZ); // Paper - cache long key
this.upgradeData = upgradeData;
this.levelHeightAccessor = heightLimitView;
this.sections = new LevelChunkSection[heightLimitView.getSectionsCount()];
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/world/level/chunk/LevelChunk.java b/src/main/java/net/minecraft/world/level/chunk/LevelChunk.java
2022-12-07 18:08:55 +01:00
index 51032b7ba79693882b64ef0fea095a7e94d215c1..272203f71dc9bb525a91f835fd73ca1d03aa982e 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/chunk/LevelChunk.java
+++ b/src/main/java/net/minecraft/world/level/chunk/LevelChunk.java
2022-03-01 06:43:03 +01:00
@@ -25,6 +25,7 @@ import net.minecraft.nbt.CompoundTag;
import net.minecraft.network.FriendlyByteBuf;
2021-11-23 11:51:25 +01:00
import net.minecraft.network.protocol.game.ClientboundLevelChunkPacketData;
import net.minecraft.server.level.ChunkHolder;
+import net.minecraft.server.level.ServerChunkCache;
import net.minecraft.server.level.ServerLevel;
import net.minecraft.util.profiling.ProfilerFiller;
import net.minecraft.world.entity.Entity;
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -124,6 +125,109 @@ public class LevelChunk extends ChunkAccess {
2021-06-11 14:02:28 +02:00
// CraftBukkit end
+ // Paper start
+ public @Nullable ChunkHolder playerChunk;
2021-06-11 14:02:28 +02:00
+
+ static final int NEIGHBOUR_CACHE_RADIUS = 3;
+ public static int getNeighbourCacheRadius() {
+ return NEIGHBOUR_CACHE_RADIUS;
+ }
+
+ boolean loadedTicketLevel;
+ private long neighbourChunksLoadedBitset;
+ private final LevelChunk[] loadedNeighbourChunks = new LevelChunk[(NEIGHBOUR_CACHE_RADIUS * 2 + 1) * (NEIGHBOUR_CACHE_RADIUS * 2 + 1)];
+
+ private static int getNeighbourIndex(final int relativeX, final int relativeZ) {
+ // index = (relativeX + NEIGHBOUR_CACHE_RADIUS) + (relativeZ + NEIGHBOUR_CACHE_RADIUS) * (NEIGHBOUR_CACHE_RADIUS * 2 + 1)
+ // optimised variant of the above by moving some of the ops to compile time
+ return relativeX + (relativeZ * (NEIGHBOUR_CACHE_RADIUS * 2 + 1)) + (NEIGHBOUR_CACHE_RADIUS + NEIGHBOUR_CACHE_RADIUS * ((NEIGHBOUR_CACHE_RADIUS * 2 + 1)));
+ }
+
+ public final LevelChunk getRelativeNeighbourIfLoaded(final int relativeX, final int relativeZ) {
+ return this.loadedNeighbourChunks[getNeighbourIndex(relativeX, relativeZ)];
+ }
+
+ public final boolean isNeighbourLoaded(final int relativeX, final int relativeZ) {
+ return (this.neighbourChunksLoadedBitset & (1L << getNeighbourIndex(relativeX, relativeZ))) != 0;
+ }
+
+ public final void setNeighbourLoaded(final int relativeX, final int relativeZ, final LevelChunk chunk) {
+ if (chunk == null) {
+ throw new IllegalArgumentException("Chunk must be non-null, neighbour: (" + relativeX + "," + relativeZ + "), chunk: " + this.chunkPos);
+ }
+ final long before = this.neighbourChunksLoadedBitset;
+ final int index = getNeighbourIndex(relativeX, relativeZ);
+ this.loadedNeighbourChunks[index] = chunk;
+ this.neighbourChunksLoadedBitset |= (1L << index);
+ this.onNeighbourChange(before, this.neighbourChunksLoadedBitset);
+ }
+
+ public final void setNeighbourUnloaded(final int relativeX, final int relativeZ) {
+ final long before = this.neighbourChunksLoadedBitset;
+ final int index = getNeighbourIndex(relativeX, relativeZ);
+ this.loadedNeighbourChunks[index] = null;
+ this.neighbourChunksLoadedBitset &= ~(1L << index);
+ this.onNeighbourChange(before, this.neighbourChunksLoadedBitset);
+ }
+
+ public final void resetNeighbours() {
+ final long before = this.neighbourChunksLoadedBitset;
+ this.neighbourChunksLoadedBitset = 0L;
+ java.util.Arrays.fill(this.loadedNeighbourChunks, null);
+ this.onNeighbourChange(before, 0L);
+ }
+
+ protected void onNeighbourChange(final long bitsetBefore, final long bitsetAfter) {
+
+ }
+
+ public final boolean isAnyNeighborsLoaded() {
+ return neighbourChunksLoadedBitset != 0;
+ }
+ public final boolean areNeighboursLoaded(final int radius) {
+ return LevelChunk.areNeighboursLoaded(this.neighbourChunksLoadedBitset, radius);
+ }
+
+ public static boolean areNeighboursLoaded(final long bitset, final int radius) {
+ // index = relativeX + (relativeZ * (NEIGHBOUR_CACHE_RADIUS * 2 + 1)) + (NEIGHBOUR_CACHE_RADIUS + NEIGHBOUR_CACHE_RADIUS * ((NEIGHBOUR_CACHE_RADIUS * 2 + 1)))
+ switch (radius) {
+ case 0: {
+ return (bitset & (1L << getNeighbourIndex(0, 0))) != 0;
+ }
+ case 1: {
+ long mask = 0L;
+ for (int dx = -1; dx <= 1; ++dx) {
+ for (int dz = -1; dz <= 1; ++dz) {
+ mask |= (1L << getNeighbourIndex(dx, dz));
+ }
+ }
+ return (bitset & mask) == mask;
+ }
+ case 2: {
+ long mask = 0L;
+ for (int dx = -2; dx <= 2; ++dx) {
+ for (int dz = -2; dz <= 2; ++dz) {
+ mask |= (1L << getNeighbourIndex(dx, dz));
+ }
+ }
+ return (bitset & mask) == mask;
+ }
+ case 3: {
+ long mask = 0L;
+ for (int dx = -3; dx <= 3; ++dx) {
+ for (int dz = -3; dz <= 3; ++dz) {
+ mask |= (1L << getNeighbourIndex(dx, dz));
+ }
+ }
+ return (bitset & mask) == mask;
+ }
+
+ default:
+ throw new IllegalArgumentException("Radius not recognized: " + radius);
+ }
+ }
+ // Paper end
+
public LevelChunk(ServerLevel world, ProtoChunk protoChunk, @Nullable LevelChunk.PostLoadProcessor entityLoader) {
this(world, protoChunk.getPos(), protoChunk.getUpgradeData(), protoChunk.unpackBlockTicks(), protoChunk.unpackFluidTicks(), protoChunk.getInhabitedTime(), protoChunk.getSections(), entityLoader, protoChunk.getBlendingData());
Iterator iterator = protoChunk.getBlockEntities().values().iterator();
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -233,6 +337,18 @@ public class LevelChunk extends ChunkAccess {
2021-06-11 14:02:28 +02:00
}
}
+ // Paper start - If loaded util
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ return this.getFluidState(blockposition);
+ }
+
+ @Override
+ public final BlockState getBlockStateIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ return this.getBlockState(blockposition);
+ }
+ // Paper end
+
@Override
public FluidState getFluidState(BlockPos pos) {
return this.getFluidState(pos.getX(), pos.getY(), pos.getZ());
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -354,6 +470,7 @@ public class LevelChunk extends ChunkAccess {
2021-06-11 14:02:28 +02:00
return this.getBlockEntity(pos, LevelChunk.EntityCreationType.CHECK);
}
2021-06-17 21:52:26 +02:00
+ @Deprecated @Nullable public final BlockEntity getTileEntityImmediately(BlockPos pos) { return this.getBlockEntity(pos, EntityCreationType.IMMEDIATE); } // Paper - OBFHELPER
2021-06-11 14:02:28 +02:00
@Nullable
public BlockEntity getBlockEntity(BlockPos pos, LevelChunk.EntityCreationType creationType) {
// CraftBukkit start
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -535,7 +652,25 @@ public class LevelChunk extends ChunkAccess {
2021-06-11 14:02:28 +02:00
// CraftBukkit start
public void loadCallback() {
+ // Paper start - neighbour cache
+ int chunkX = this.chunkPos.x;
+ int chunkZ = this.chunkPos.z;
+ ServerChunkCache chunkProvider = this.level.getChunkSource();
2021-06-11 14:02:28 +02:00
+ for (int dx = -NEIGHBOUR_CACHE_RADIUS; dx <= NEIGHBOUR_CACHE_RADIUS; ++dx) {
+ for (int dz = -NEIGHBOUR_CACHE_RADIUS; dz <= NEIGHBOUR_CACHE_RADIUS; ++dz) {
+ LevelChunk neighbour = chunkProvider.getChunkAtIfLoadedMainThreadNoCache(chunkX + dx, chunkZ + dz);
+ if (neighbour != null) {
+ neighbour.setNeighbourLoaded(-dx, -dz, this);
+ // should be in cached already
+ this.setNeighbourLoaded(dx, dz, neighbour);
+ }
+ }
+ }
+ this.setNeighbourLoaded(0, 0, this);
+ this.loadedTicketLevel = true;
+ // Paper end - neighbour cache
org.bukkit.Server server = this.level.getCraftServer();
+ this.level.getChunkSource().addLoadedChunk(this); // Paper
2021-06-11 14:02:28 +02:00
if (server != null) {
/*
* If it's a new world, the first few chunks are generated inside
Rewrite chunk system (#8177) Patch documentation to come Issues with the old system that are fixed now: - World generation does not scale with cpu cores effectively. - Relies on the main thread for scheduling and maintaining chunk state, dropping chunk load/generate rates at lower tps. - Unreliable prioritisation of chunk gen/load calls that block the main thread. - Shutdown logic is utterly unreliable, as it has to wait for all chunks to unload - is it guaranteed that the chunk system is in a state on shutdown that it can reliably do this? Watchdog shutdown also typically failed due to thread checks, which is now resolved. - Saving of data is not unified (i.e can save chunk data without saving entity data, poses problems for desync if shutdown is really abnormal. - Entities are not loaded with chunks. This caused quite a bit of headache for Chunk#getEntities API, but now the new chunk system loads entities with chunks so that they are ready whenever the chunk loads in. Effectively brings the behavior back to 1.16 era, but still storing entities in their own separate regionfiles. The above list is not complete. The patch documentation will complete it. New chunk system hard relies on starlight and dataconverter, and most importantly the new concurrent utilities in ConcurrentUtil. Some of the old async chunk i/o interface (i.e the old file io thread reroutes _some_ calls to the new file io thread) is kept for plugin compat reasons. It will be removed in the next major version of minecraft. The old legacy chunk system patches have been moved to the removed folder in case we need them again.
2022-09-26 10:02:51 +02:00
@@ -574,6 +709,22 @@ public class LevelChunk extends ChunkAccess {
2021-06-11 14:02:28 +02:00
server.getPluginManager().callEvent(unloadEvent);
// note: saving can be prevented, but not forced if no saving is actually required
this.mustNotSave = !unloadEvent.isSaveChunk();
+ this.level.getChunkSource().removeLoadedChunk(this); // Paper
2021-06-11 14:02:28 +02:00
+ // Paper start - neighbour cache
+ int chunkX = this.chunkPos.x;
+ int chunkZ = this.chunkPos.z;
+ ServerChunkCache chunkProvider = this.level.getChunkSource();
2021-06-11 14:02:28 +02:00
+ for (int dx = -NEIGHBOUR_CACHE_RADIUS; dx <= NEIGHBOUR_CACHE_RADIUS; ++dx) {
+ for (int dz = -NEIGHBOUR_CACHE_RADIUS; dz <= NEIGHBOUR_CACHE_RADIUS; ++dz) {
+ LevelChunk neighbour = chunkProvider.getChunkAtIfLoadedMainThreadNoCache(chunkX + dx, chunkZ + dz);
+ if (neighbour != null) {
+ neighbour.setNeighbourUnloaded(-dx, -dz);
+ }
+ }
+ }
+ this.loadedTicketLevel = false;
+ this.resetNeighbours();
+ // Paper end
}
2021-11-23 11:51:25 +01:00
@Override
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/net/minecraft/world/level/chunk/ProtoChunk.java b/src/main/java/net/minecraft/world/level/chunk/ProtoChunk.java
index cb4e43177a735442fa2adda8640263bca8cdcb64..92a64c49b1c7227a5b34488ea15d3d8adb0f9c80 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/net/minecraft/world/level/chunk/ProtoChunk.java
+++ b/src/main/java/net/minecraft/world/level/chunk/ProtoChunk.java
2022-03-01 06:43:03 +01:00
@@ -74,6 +74,18 @@ public class ProtoChunk extends ChunkAccess {
2021-11-23 11:51:25 +01:00
return new ChunkAccess.TicksToSave(this.blockTicks, this.fluidTicks);
2021-06-11 14:02:28 +02:00
}
+ // Paper start - If loaded util
+ @Override
+ public final FluidState getFluidIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ return this.getFluidState(blockposition);
+ }
+
+ @Override
+ public final BlockState getBlockStateIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ return this.getBlockState(blockposition);
+ }
+ // Paper end
+
@Override
public BlockState getBlockState(BlockPos pos) {
int i = pos.getY();
diff --git a/src/main/java/net/minecraft/world/level/entity/PersistentEntitySectionManager.java b/src/main/java/net/minecraft/world/level/entity/PersistentEntitySectionManager.java
index f66369ddaeab5c5ac643c0979dac3ed21337ff71..038abf2ac104ceecaab11b10d466ea69ec86623e 100644
--- a/src/main/java/net/minecraft/world/level/entity/PersistentEntitySectionManager.java
+++ b/src/main/java/net/minecraft/world/level/entity/PersistentEntitySectionManager.java
@@ -90,6 +90,18 @@ public class PersistentEntitySectionManager<T extends EntityAccess> implements A
}
private boolean addEntity(T entity, boolean existing) {
+ // Paper start - chunk system hooks
+ if (existing) {
+ // I don't want to know why this is a generic type.
+ Entity entityCasted = (Entity)entity;
+ boolean wasRemoved = entityCasted.isRemoved();
+ io.papermc.paper.chunk.system.ChunkSystem.onEntityPreAdd((net.minecraft.server.level.ServerLevel)entityCasted.level, entityCasted);
+ if (!wasRemoved && entityCasted.isRemoved()) {
+ // removed by callback
+ return false;
+ }
+ }
+ // Paper end - chunk system hooks
if (!this.addEntityUuid(entity)) {
return false;
} else {
diff --git a/src/main/java/org/bukkit/craftbukkit/CraftWorld.java b/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
index dc8f094dea2fd46fe4d860536908a42e4b5806a1..57ed87689c7cb709d54e611bfbdbb70890021df2 100644
--- a/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
+++ b/src/main/java/org/bukkit/craftbukkit/CraftWorld.java
2022-12-07 18:08:55 +01:00
@@ -233,8 +233,8 @@ public class CraftWorld extends CraftRegionAccessor implements World {
@Override
public Chunk[] getLoadedChunks() {
- Long2ObjectLinkedOpenHashMap<ChunkHolder> chunks = this.world.getChunkSource().chunkMap.visibleChunkMap;
- return chunks.values().stream().map(ChunkHolder::getFullChunkNow).filter(Objects::nonNull).map(net.minecraft.world.level.chunk.LevelChunk::getBukkitChunk).toArray(Chunk[]::new);
+ List<ChunkHolder> chunks = io.papermc.paper.chunk.system.ChunkSystem.getVisibleChunkHolders(this.world); // Paper
+ return chunks.stream().map(ChunkHolder::getFullChunkNow).filter(Objects::nonNull).map(net.minecraft.world.level.chunk.LevelChunk::getBukkitChunk).toArray(Chunk[]::new);
}
@Override
2022-12-07 18:08:55 +01:00
@@ -309,7 +309,7 @@ public class CraftWorld extends CraftRegionAccessor implements World {
@Override
public boolean refreshChunk(int x, int z) {
- ChunkHolder playerChunk = this.world.getChunkSource().chunkMap.visibleChunkMap.get(ChunkPos.asLong(x, z));
+ ChunkHolder playerChunk = this.world.getChunkSource().chunkMap.getVisibleChunkIfPresent(ChunkPos.asLong(x, z));
if (playerChunk == null) return false;
playerChunk.getTickingChunkFuture().thenAccept(either -> {
@@ -1959,4 +1959,32 @@ public class CraftWorld extends CraftRegionAccessor implements World {
return this.spigot;
}
// Spigot end
+ // Paper start
+ public java.util.concurrent.CompletableFuture<Chunk> getChunkAtAsync(int x, int z, boolean gen, boolean urgent) {
+ if (Bukkit.isPrimaryThread()) {
+ net.minecraft.world.level.chunk.LevelChunk immediate = this.world.getChunkSource().getChunkAtIfLoadedImmediately(x, z);
+ if (immediate != null) {
+ return java.util.concurrent.CompletableFuture.completedFuture(immediate.getBukkitChunk());
+ }
+ }
+
+ ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority priority;
+ if (urgent) {
+ priority = ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.HIGHER;
+ } else {
+ priority = ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL;
+ }
+
+ java.util.concurrent.CompletableFuture<Chunk> ret = new java.util.concurrent.CompletableFuture<>();
+
+ io.papermc.paper.chunk.system.ChunkSystem.scheduleChunkLoad(this.getHandle(), x, z, gen, ChunkStatus.FULL, true, priority, (c) -> {
+ net.minecraft.server.MinecraftServer.getServer().scheduleOnMain(() -> {
+ net.minecraft.world.level.chunk.LevelChunk chunk = (net.minecraft.world.level.chunk.LevelChunk)c;
+ ret.complete(chunk == null ? null : chunk.getBukkitChunk());
+ });
+ });
+
+ return ret;
+ }
+ // Paper end
}
diff --git a/src/main/java/org/bukkit/craftbukkit/entity/CraftEntity.java b/src/main/java/org/bukkit/craftbukkit/entity/CraftEntity.java
2022-12-07 18:08:55 +01:00
index c3cad46ce064c9581e68db17495dabe2dc2ca161..5f6d4f553cab8f20c2049c706e69de3d79701dfd 100644
--- a/src/main/java/org/bukkit/craftbukkit/entity/CraftEntity.java
+++ b/src/main/java/org/bukkit/craftbukkit/entity/CraftEntity.java
2022-12-07 18:08:55 +01:00
@@ -1147,4 +1147,37 @@ public abstract class CraftEntity implements org.bukkit.entity.Entity {
return this.spigot;
}
// Spigot end
+
+ // Paper start
+ @Override
+ public java.util.concurrent.CompletableFuture<Boolean> teleportAsync(Location location, TeleportCause cause) {
+ Preconditions.checkArgument(location != null, "location");
+ location.checkFinite();
+ Location locationClone = location.clone(); // clone so we don't need to worry about mutations after this call.
+
+ net.minecraft.server.level.ServerLevel world = ((CraftWorld)locationClone.getWorld()).getHandle();
+ java.util.concurrent.CompletableFuture<Boolean> ret = new java.util.concurrent.CompletableFuture<>();
+
+ world.loadChunksForMoveAsync(getHandle().getBoundingBoxAt(locationClone.getX(), locationClone.getY(), locationClone.getZ()),
+ this instanceof CraftPlayer ? ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.HIGHER : ca.spottedleaf.concurrentutil.executor.standard.PrioritisedExecutor.Priority.NORMAL, (list) -> {
+ net.minecraft.server.level.ServerChunkCache chunkProviderServer = world.getChunkSource();
+ for (net.minecraft.world.level.chunk.ChunkAccess chunk : list) {
+ chunkProviderServer.addTicketAtLevel(net.minecraft.server.level.TicketType.POST_TELEPORT, chunk.getPos(), 33, CraftEntity.this.getEntityId());
+ }
+ net.minecraft.server.MinecraftServer.getServer().scheduleOnMain(() -> {
+ try {
+ ret.complete(CraftEntity.this.teleport(locationClone, cause) ? Boolean.TRUE : Boolean.FALSE);
+ } catch (Throwable throwable) {
+ if (throwable instanceof ThreadDeath) {
+ throw (ThreadDeath)throwable;
+ }
+ net.minecraft.server.MinecraftServer.LOGGER.error("Failed to teleport entity " + CraftEntity.this, throwable);
+ ret.completeExceptionally(throwable);
+ }
+ });
+ });
+
+ return ret;
+ }
+ // Paper end
}
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/org/bukkit/craftbukkit/scheduler/CraftScheduler.java b/src/main/java/org/bukkit/craftbukkit/scheduler/CraftScheduler.java
index 2f52148f36e56c503e619634eedd3d46d9f44938..054fcc3713f02e358dfe049491c8d1689ccc750b 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/org/bukkit/craftbukkit/scheduler/CraftScheduler.java
+++ b/src/main/java/org/bukkit/craftbukkit/scheduler/CraftScheduler.java
@@ -44,6 +44,7 @@ import org.bukkit.scheduler.BukkitWorker;
2021-06-11 14:02:28 +02:00
*/
public class CraftScheduler implements BukkitScheduler {
+ static Plugin MINECRAFT = new MinecraftInternalPlugin();
/**
* The start ID for the counter.
2021-06-11 14:02:28 +02:00
*/
@@ -192,6 +193,11 @@ public class CraftScheduler implements BukkitScheduler {
this.runTaskTimer(plugin, (Object) task, delay, period);
2021-06-11 14:02:28 +02:00
}
+ public BukkitTask scheduleInternalTask(Runnable run, int delay, String taskName) {
+ final CraftTask task = new CraftTask(run, nextId(), taskName);
+ return handle(task, delay);
+ }
+
public BukkitTask runTaskTimer(Plugin plugin, Object runnable, long delay, long period) {
CraftScheduler.validate(plugin, runnable);
2021-06-11 14:02:28 +02:00
if (delay < 0L) {
@@ -415,13 +421,20 @@ public class CraftScheduler implements BukkitScheduler {
2021-06-11 14:02:28 +02:00
task.run();
task.timings.stopTiming(); // Spigot
} catch (final Throwable throwable) {
- task.getOwner().getLogger().log(
+ // Paper start
+ String msg = String.format(
+ "Task #%s for %s generated an exception",
+ task.getTaskId(),
+ task.getOwner().getDescription().getFullName());
+ if (task.getOwner() == MINECRAFT) {
+ net.minecraft.server.MinecraftServer.LOGGER.error(msg, throwable);
+ } else {
+ task.getOwner().getLogger().log(
Level.WARNING,
- String.format(
- "Task #%s for %s generated an exception",
- task.getTaskId(),
- task.getOwner().getDescription().getFullName()),
+ msg,
throwable);
+ }
+ // Paper end
} finally {
this.currentTask = null;
2021-06-11 14:02:28 +02:00
}
diff --git a/src/main/java/org/bukkit/craftbukkit/scheduler/CraftTask.java b/src/main/java/org/bukkit/craftbukkit/scheduler/CraftTask.java
index 592a2a513ebf0002abf1255e11012ecc66adecf0..b89846e0f645c79afec018dae1d64a1bda043ed9 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/org/bukkit/craftbukkit/scheduler/CraftTask.java
+++ b/src/main/java/org/bukkit/craftbukkit/scheduler/CraftTask.java
@@ -40,6 +40,21 @@ public class CraftTask implements BukkitTask, Runnable { // Spigot
2021-06-11 14:02:28 +02:00
CraftTask(final Object task) {
this(null, task, CraftTask.NO_REPEATING, CraftTask.NO_REPEATING);
}
+ // Paper start
+ public String taskName = null;
+ boolean internal = false;
+ CraftTask(final Object task, int id, String taskName) {
+ this.rTask = (Runnable) task;
+ this.cTask = null;
+ this.plugin = CraftScheduler.MINECRAFT;
+ this.taskName = taskName;
+ this.internal = true;
+ this.id = id;
+ this.period = CraftTask.NO_REPEATING;
+ this.taskName = taskName;
+ this.timings = null; // Will be changed in later patch
+ }
+ // Paper end
CraftTask(final Plugin plugin, final Object task, final int id, final long period) {
this.plugin = plugin;
diff --git a/src/main/java/org/bukkit/craftbukkit/scheduler/MinecraftInternalPlugin.java b/src/main/java/org/bukkit/craftbukkit/scheduler/MinecraftInternalPlugin.java
new file mode 100644
index 0000000000000000000000000000000000000000..909b2c98e7a9117d2f737245e4661792ffafb744
2021-06-11 14:02:28 +02:00
--- /dev/null
+++ b/src/main/java/org/bukkit/craftbukkit/scheduler/MinecraftInternalPlugin.java
@@ -0,0 +1,140 @@
2021-06-11 14:02:28 +02:00
+package org.bukkit.craftbukkit.scheduler;
+
+
+import org.bukkit.Server;
+import org.bukkit.command.Command;
+import org.bukkit.command.CommandSender;
+import org.bukkit.configuration.file.FileConfiguration;
+import org.bukkit.generator.BiomeProvider;
2021-06-11 14:02:28 +02:00
+import org.bukkit.generator.ChunkGenerator;
+import org.bukkit.plugin.PluginBase;
+import org.bukkit.plugin.PluginDescriptionFile;
+import org.bukkit.plugin.PluginLoader;
+import org.bukkit.plugin.PluginLogger;
+import org.jetbrains.annotations.NotNull;
+import org.jetbrains.annotations.Nullable;
2021-06-11 14:02:28 +02:00
+
+import java.io.File;
+import java.io.InputStream;
+import java.util.List;
+
+public class MinecraftInternalPlugin extends PluginBase {
+ private boolean enabled = true;
+
+ private final String pluginName;
+ private PluginDescriptionFile pdf;
+
+ public MinecraftInternalPlugin() {
+ this.pluginName = "Minecraft";
+ pdf = new PluginDescriptionFile(pluginName, "1.0", "nms");
+ }
+
+ public void setEnabled(boolean enabled) {
+ this.enabled = enabled;
+ }
+
+ @Override
+ public File getDataFolder() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public PluginDescriptionFile getDescription() {
+ return pdf;
+ }
+
+ @Override
+ public FileConfiguration getConfig() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public InputStream getResource(String filename) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void saveConfig() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void saveDefaultConfig() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void saveResource(String resourcePath, boolean replace) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void reloadConfig() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public PluginLogger getLogger() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public PluginLoader getPluginLoader() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public Server getServer() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public boolean isEnabled() {
+ return enabled;
+ }
+
+ @Override
+ public void onDisable() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void onLoad() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void onEnable() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public boolean isNaggable() {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public void setNaggable(boolean canNag) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public ChunkGenerator getDefaultWorldGenerator(String worldName, String id) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public @Nullable BiomeProvider getDefaultBiomeProvider(@NotNull String worldName, @Nullable String id) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
2021-06-11 14:02:28 +02:00
+ public boolean onCommand(CommandSender sender, Command command, String label, String[] args) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+
+ @Override
+ public List<String> onTabComplete(CommandSender sender, Command command, String alias, String[] args) {
+ throw new UnsupportedOperationException("Not supported.");
+ }
+}
2021-11-24 20:02:36 +01:00
diff --git a/src/main/java/org/bukkit/craftbukkit/util/CraftMagicNumbers.java b/src/main/java/org/bukkit/craftbukkit/util/CraftMagicNumbers.java
index 8ded1f61a6088cf57847a409421e92e2d2007158..b39ded554ee86bf840ce04eba58ac19d40226c9f 100644
2021-11-24 20:02:36 +01:00
--- a/src/main/java/org/bukkit/craftbukkit/util/CraftMagicNumbers.java
+++ b/src/main/java/org/bukkit/craftbukkit/util/CraftMagicNumbers.java
@@ -102,8 +102,17 @@ public final class CraftMagicNumbers implements UnsafeValues {
private static final BiMap<net.minecraft.world.level.material.Fluid, Fluid> FLUIDTYPE_FLUID = HashBiMap.create();
2021-11-24 20:02:36 +01:00
private static final Map<Material, Item> MATERIAL_ITEM = new HashMap<>();
private static final Map<Material, Block> MATERIAL_BLOCK = new HashMap<>();
+ // Paper start
+ private static final Map<org.bukkit.entity.EntityType, net.minecraft.world.entity.EntityType<?>> ENTITY_TYPE_ENTITY_TYPES = new HashMap<>();
+ private static final Map<net.minecraft.world.entity.EntityType<?>, org.bukkit.entity.EntityType> ENTITY_TYPES_ENTITY_TYPE = new HashMap<>();
static {
+ for (org.bukkit.entity.EntityType type : org.bukkit.entity.EntityType.values()) {
+ if (type == org.bukkit.entity.EntityType.UNKNOWN) continue;
2022-12-07 18:08:55 +01:00
+ ENTITY_TYPE_ENTITY_TYPES.put(type, BuiltInRegistries.ENTITY_TYPE.get(CraftNamespacedKey.toMinecraft(type.getKey())));
+ ENTITY_TYPES_ENTITY_TYPE.put(BuiltInRegistries.ENTITY_TYPE.get(CraftNamespacedKey.toMinecraft(type.getKey())), type);
2021-11-24 20:02:36 +01:00
+ }
+ // Paper end
2022-12-07 18:08:55 +01:00
for (Block block : BuiltInRegistries.BLOCK) {
BLOCK_MATERIAL.put(block, Material.getMaterial(BuiltInRegistries.BLOCK.getKey(block).getPath().toUpperCase(Locale.ROOT)));
2021-11-24 20:02:36 +01:00
}
@@ -167,6 +176,14 @@ public final class CraftMagicNumbers implements UnsafeValues {
2021-11-24 20:02:36 +01:00
public static ResourceLocation key(Material mat) {
return CraftNamespacedKey.toMinecraft(mat.getKey());
}
+ // Paper start
+ public static net.minecraft.world.entity.EntityType<?> getEntityTypes(org.bukkit.entity.EntityType type) {
+ return ENTITY_TYPE_ENTITY_TYPES.get(type);
+ }
+ public static org.bukkit.entity.EntityType getEntityType(net.minecraft.world.entity.EntityType<?> entityTypes) {
+ return ENTITY_TYPES_ENTITY_TYPE.get(entityTypes);
+ }
+ // Paper end
// ========================================================================
public static byte toLegacyData(BlockState data) {
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/org/bukkit/craftbukkit/util/DummyGeneratorAccess.java b/src/main/java/org/bukkit/craftbukkit/util/DummyGeneratorAccess.java
2022-12-07 18:08:55 +01:00
index bbc2ab1972d9b88530282225650ec14b521d413f..9a80e0c390d13453a4a79e00d18c20b79afd3c7f 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/org/bukkit/craftbukkit/util/DummyGeneratorAccess.java
+++ b/src/main/java/org/bukkit/craftbukkit/util/DummyGeneratorAccess.java
2022-12-07 18:08:55 +01:00
@@ -213,7 +213,23 @@ public class DummyGeneratorAccess implements WorldGenLevel {
2021-06-11 14:02:28 +02:00
public FluidState getFluidState(BlockPos pos) {
return Fluids.EMPTY.defaultFluidState(); // SPIGOT-6634
2021-06-11 14:02:28 +02:00
}
+ // Paper start - if loaded util
+ @javax.annotation.Nullable
+ @Override
+ public ChunkAccess getChunkIfLoadedImmediately(int x, int z) {
+ throw new UnsupportedOperationException("Not supported yet.");
+ }
+
+ @Override
+ public BlockState getBlockStateIfLoaded(BlockPos blockposition) {
2021-06-11 14:02:28 +02:00
+ throw new UnsupportedOperationException("Not supported yet.");
+ }
+ @Override
+ public FluidState getFluidIfLoaded(BlockPos blockposition) {
+ throw new UnsupportedOperationException("Not supported yet.");
+ }
+ // Paper end
@Override
public WorldBorder getWorldBorder() {
throw new UnsupportedOperationException("Not supported yet.");
Merge tuinity (#6413) This PR contains all of Tuinity's patches. Very notable ones are: - Highly optimised collisions - Optimised entity lookups by bounding box (Mojang made regressions in 1.17, this brings it back to 1.16) - Starlight https://github.com/PaperMC/Starlight - Rewritten dataconverter system https://github.com/PaperMC/DataConverter - Random block ticking optimisation (wrongly dropped from Paper 1.17) - Chunk ticking optimisations - Anything else I've forgotten in the 60 or so patches If you are a previous Tuinity user, your config will not migrate. You must do it yourself. The config options have simply been moved into paper.yml, so it will be an easy migration. However, please note that the chunk loading options in tuinity.yml are NOT compatible with the options in paper.yml. * Port tuinity, initial patchset * Update gradle to 7.2 jmp said it fixes rebuildpatches not working for me. it fucking better * Completely clean apply * Remove tuinity config, add per player api patch * Remove paper reobf mappings patch * Properly update gradlew * Force clean rebuild * Mark fixups Comments and ATs still need to be done * grep -r "Tuinity" * Fixup * Ensure gameprofile lastaccess is written only under the state lock * update URL for dataconverter * Only clean rebuild tuinity patches might fix merge conflicts * Use UTF-8 for gradlew * Clean rb patches again * Convert block ids used as item ids Neither the converters of pre 1.13 nor DFU handled these cases, as by the time they were written the game at the time didn't consider these ids valid - they would be air. Because of this, some worlds have logspam since only DataConverter (not DFU or legacy converters) will warn when an invalid id has been seen. While quite a few do need to now be considered as air, quite a lot do not. So it makes sense to add conversion for these items, instead of simply suppressing or ignoring the logs. I've now added id -> string conversion for all block ids that could be used as items that existed in the game before 1.7.10 (I have no interest in tracking down the exact version block ids stopped working) that were on https://minecraft-ids.grahamedgecombe.com/ Items that did not directly convert to new items will be instead converted to air: stems, wheat crops, piston head, tripwire wire block * Fix LightPopulated parsing in V1466 The DFU code was checking if the number existed, not if it didn't exist. I misread the original code. * Always parse protochunk light sources unless it is marked as non-lit Chunks not marked as lit will always go through the light engine, so they should always have their block sources parsed. * Update custom names to JSON for players Missed this fix from CB, as it was inside the DataFixers class. I decided to double check all of the CB changes again: DataFixers.java was the only area I missed, as I had inspected all datafixer diffs and implemented them all into DataConverter. I also checked Bootstrap.java again, and re-evaluated their changes. I had previously done this, but determined that they were all bad. The changes to make standing_sign block map to oak_sign block in V1450 is bad, because that's not the item id V1450 accepts. Only in 1.14 did oak_sign even exist, and as expected there is a converter to rename all existing sign items/blocks. The fix to register the portal block under id 1440 is useless, as the flattenning logic will default to the lowest registered id - which is the exact blockstate that CB registers into 1440. So it just doesn't do anything. The extra item ids in the id -> string converter are already added, but I found this from EMC originally. The change for the spawn egg id 23 -> Arrow is just wrong, that id DOES correspond to TippedArrow, NOT Arrow. As expected, the spawn egg already has a dedicated mapping for Arrow, which is id 10 - which was Arrow's entity id. I also ported a fix for the cooked_fished id update. This doesn't really matter since there is already a dataconverter to fix this, but the game didn't accept cooked_fished at the time. So I see no harm. * Review all converters and walkers - Refactor V99 to have helper methods for defining entity/tile entity types - Automatically namespace all ids that should be namespaced. While vanilla never saved non-namespaced data for things that are namespaced, plugins/users might have. - Synchronised the identity ensure map in HelperBlockFlatteningV1450 - Code style consistency - Add missing log warning in V102 for ITEM_NAME type conversion - Use getBoolean instead of getByte - Use ConverterAbstractEntityRename for V143 TippedArrow -> Arrow rename, as it will affect ENTITY_NAME type - Always set isVillager to false in V502 for Zombie - Register V808's converter under subversion 1 like DFU - Register a breakpoint for V1.17.1. In the future, all final versions of major releases will have a breakpoint so that the work required to determine if a converter needs a breakpoint is minimal - Validate that a dataconverter is only registered for a version that is registered - ConverterFlattenTileEntity is actually ConverterFlattenEntity It even registered the converters under TILE_ENTITY, instead of ENTITY. - Fix id comparison in V1492 STRUCTURE_FEATURE renamer - Use ConverterAbstractStatsRename for V1510 stats renamer At the time I had written that class, the abstract renamer didn't exist. - Ensure OwnerUUID is at least set to empty string in V1904 if the ocelot is converted to a cat (this is likely so that it retains a collar) - Use generic read/write for Records in V1946 Records is actually a list, not a map. So reading map was invalid. * Always set light to zero when propagating decrease This fixes an almost infinite loop where light values would be spam queued on a very small subset on blocks. This also likely fixes the memory issues people were seeing. * re-organize patches * Apply and fix conflicts * Revert some patches getChunkAt retains chunks so that plugins don't spam loads revert mc-4 fix will remain unless issues pop up * Shuffle iterated chunks if per player is not enabled Can help with some mob spawning stacking up at locations * Make per player default, migrate all configs * Adjust comments in fixups * Rework config for player chunk loader Old config is not compatible. Move all configs to be under `settings` in paper.yml The player chunk loader has been modified to less aggressively load chunks, but to send chunks at higher rates compared to tuinity. There are new config entries to tune this behavior. * Add back old constructor to CompressionEncoder/Decoder (fixes Tuinity #358) * Raise chunk loading default limits * Reduce worldgen thread workers for lower core count cpus * Raise limits for chunk loading config Also place it under `chunk-loading` * Disable max chunk send rate by default * Fix conflicts and rebuild patches * Drop default send rate again Appears to be still causing problems for no known reason * Raise chunk send limits to 100 per player While a low limit fixes ping issues for some people, most people do not suffer from this issue and thus should not suffer from an extremely slow load-in rate. * Rebase part 1 Autosquash the fixups * Move not implemented up * Fixup mc-dev fixes Missed this one * Rebase per player viewdistance api into the original api patch * Remove old light engine patch part 1 The prioritisation must be kept from it, so that part has been rebased into the priority patch. Part 2 will deal with rebasing all of the patches _after_ * Rebase remaining patches for old light patch removal * Remove other mid tick patch * Remove Optimize-PlayerChunkMap-memory-use-for-visibleChunks.patch Replaced by `Do not copy visible chunks` * Revert AT for Vec3i setX/Y/Z The class is immutable. set should not be exposed * Remove old IntegerUtil class * Replace old CraftChunk#getEntities patch * Remove import for SWMRNibbleArray in ChunkAccess * Finished merge checklist * Remove ensureTickThread impl in urgency patch Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com> Co-authored-by: Jason Penilla <11360596+jpenilla@users.noreply.github.com>
2021-08-31 13:02:11 +02:00
diff --git a/src/main/java/org/bukkit/craftbukkit/util/UnsafeList.java b/src/main/java/org/bukkit/craftbukkit/util/UnsafeList.java
index d40c0d8be1b0153d62021b8bcb6e8b37fd0acb4e..e38e57b1f9ef27020de35d7ddcb36a663140f880 100644
--- a/src/main/java/org/bukkit/craftbukkit/util/UnsafeList.java
+++ b/src/main/java/org/bukkit/craftbukkit/util/UnsafeList.java
@@ -119,6 +119,32 @@ public class UnsafeList<E> extends AbstractList<E> implements List<E>, RandomAcc
return this.indexOf(o) >= 0;
}
+ // Paper start
+ protected transient int maxSize;
+ public void setSize(int size) {
+ if (this.maxSize < this.size) {
+ this.maxSize = this.size;
+ }
+ this.size = size;
+ }
+
+ public void completeReset() {
+ if (this.data != null) {
+ Arrays.fill(this.data, 0, Math.max(this.size, this.maxSize), null);
+ }
+ this.size = 0;
+ this.maxSize = 0;
+ if (this.iterPool != null) {
+ for (Iterator temp : this.iterPool) {
+ if (temp == null) {
+ continue;
+ }
+ ((Itr)temp).valid = false;
+ }
+ }
+ }
+ // Paper end
+
@Override
public void clear() {
// Create new array to reset memory usage to initial capacity
2021-06-11 14:02:28 +02:00
diff --git a/src/main/java/org/spigotmc/SpigotConfig.java b/src/main/java/org/spigotmc/SpigotConfig.java
2022-12-07 18:08:55 +01:00
index b599e098e6d2e6f78b89114e05cb03d716b7c5f4..83eabc34d952bbb13ec4b4bdcf34f647189c0b46 100644
2021-06-11 14:02:28 +02:00
--- a/src/main/java/org/spigotmc/SpigotConfig.java
+++ b/src/main/java/org/spigotmc/SpigotConfig.java
@@ -118,7 +118,11 @@ public class SpigotConfig
}
}
}
-
+ // Paper start
+ SpigotConfig.save();
+ }
+ public static void save() {
+ // Paper end
try
{
SpigotConfig.config.save( CONFIG_FILE );