Minecraft 1.5.2 changed mob spawning by ignoring persistent mobs from the
mob count. In CraftBukkit all mobs that previously used a separate hard
coded persistence system were ported to instead use this one. This causes
all animals to be persistent by default and thus never be counted. To
correct this issue we consider if the mob would be considered persistent
in the hard coded system as well when deciding if a mob should count.
Currently there are several cases where a player will have their inventory
screen closed client side but we will not call an event. To correct this
we call the event when the server is the cause of the inventory closing
instead of just when the client is the cause. We also ensure the server is
closing the inventory reliably so we get the events. Additionally this
commit also calls the event when a player disconnects which will handle
kicks, quits, and server shutdown.
As an optimization we don't do any movement processing on entities that
have not moved. However, players in minecarts trigger this condition when
the minecart is moving on its own. This causes issues with things that rely
on player collision like tripwires. To correct this and any future related
issues we always handle movement for passengers and their vehicles even
if one of the two hasn't moved since they are linked.
When converting things in Minecraft to use wall time instead of ticks I
realized we'd run into integer division rounding issues and could have
updates that end up counting as zero ticks. To compensate for this the
code ensures we always process at least one tick. However, every time we
end up with zero ticks the next time we have an extra tick due to rounding
the other way with the leftovers. This means we are going far too fast and
should not have this at least one tick logic at all.
On top of this some potions rely on the number of ticks they run and not
just the amount of time they last and so potions were put back to running
with ticks entirely.
If a chunk has somehow managed to save with arrays that are not 4096
entries long when reading them again we will get exceptions. Checking the
array length and resizing if needed is cheap so we should do this to help
avoid crashing servers due to this error.
When a chunk is being loaded the server checks to ensure the chunk's idea
of where it is located matches where it was located in the region file. If
these two values do not match the chunk's idea of its position is updated
and the chunk is reloaded. In vanilla minecraft this loading involves the
chunk's tile entities as well. With the change to loading player chunks
asynchronously we split loading tile entities to a separate step that takes
place after this check. Because of this tile entities are loaded with
invalid locations that result in trying to fetch block data from negative
or too large positions in the chunk's internal block storage arrays.
Because loading the tile entities is not thread safe we cannot return to
vanilla behavior here. Instead when we detect a misplaced chunk we just
edit the NBT data for the chunk to relocate the tile entities. This results
in them moving correctly with the chunk without having to actually load
them first.
If a hopper is unable to perform any action during a tick it attempts to do
so every tick from then on. Once it is able to do so it then sets a delay
before attempting to do something again. To avoid excessive CPU usage for
hoppers sitting idle we now apply this delay regardless of the success of
the action.
When using the new feature in 1.5 to drop the item in any highlighted slot,
the anvil result slot does not apply the full anvil calculation that picking
up the item does, including the experience calculation.
Currently furnace smelting and the item pickup delay timer use wall time
(aka actual time passed) to emulate a constant tick rate so run at the
same speed regardless of the server's actual tick rate. There are several
other places this makes sense so this commit converts them.
The item despawn timer is converted so now always takes 5 minutes. Users
know this 5 minute number well so keeping this constant helps to avoid
confusion. This also helps alleviate lag because if a large number of item
drops is the reason your server is running slowly having them stay around
longer just means your server is slow longer.
Potion brewing and the zombie villager conversion timer are now constant.
These match the furnace criteria of being useful for hiding lag and not
having a detrimental effect on gameplay.
Potion effects are now also using wall time. The client is told about effect
times in ticks and displays this information to the user as minutes and
seconds assuming a solid 20 ticks per second. The server does have
code for updating the client with the current time remaining to help
avoid skew due to differing tick rates but making this a constant makes
sense due to this display.
Add a check to avoid doing movement work if an entity doesn't move. This
usually will not ever happen in the current server but is useful when it
does and will be more useful in the future.
Only process mob on mob (non-player) collisions every other tick. Players
tend to pack a lot of mobs into a small space (sheep farm, mob grinder, etc)
so they do a lot of work processing collisions. To help alleviate some of
this we only run these calculations every other tick. This has no visible
effect on the client but can be a huge win on the server depending on
circumstances.
Use generic entity inWater checking for squids. Squids have their own logic
currently for determining if they are in water. This check is almost
identical to the generic entity checking which is run anyway. To avoid
doing duplicate work we just remove the squid version. This does not
have any noticeable effect on gameplay since the checks are so similar.
Use HashSet for tile entities instead of ArrayList. Using an ArrayList for
storing tile entities in a world means we have very expensive inserts and
removes that aren't at the end of the array due to the array copy this
causes. This is most noticeable during chunk unload when a large number of
tile entities are removed from the world at once. Using a HashSet here uses
a little more memory but is O(1) for all operations so removes this
bottleneck.
When a player comes into range of an entity in a vehicle they will often be
sent the spawn packet for the rider before receiving one for the vehicle.
This causes the attachment packet to fail client side because it attempts
to attach the rider to a vehicle that doesn't exist on the client. To
correct this we account for both possible orderings and send the
attachment packet appropriately.
Vanilla also sends an new attach packet every 60 ticks even if the vehicle
has not changed. As this packet is a toggle this resulting in players
teleporting around randomly. Since we handle attachments properly now we
simply revert this section to use the 1.4 logic.
When looking up tile entities for a chunk to send to a player we currently
loop through every tile entity in the world checking if it is within the
bounds of the relevant chunk. Instead of doing this we can just use the
tile entities list stored in the chunk to avoid this costly searching.
As a further optimization, we also modify the generic range-based lookup
to use chunks as well. For most lookups this will give a smaller search
pool which will result in faster lookups.
Thanks to @mikeprimm for the idea and most of the implementation.
In certain scenarios a boat can be killed multiple ways in a single tick.
Due to improper guards this can cause the death code to run multiple times
creating item drops. To correct this we insert the appropriate death check.
The method's return value was being incorrectly calculated from the
perspective of this.canHurt(that), where it's actually used from the
this.canBeHurtBy(that) perspective.
When a world is created using our API, it does not use secondary world
server and will maintain a reference to its own scoreboard. In vanilla,
this is not an issue as there is only ever one world.
Similarly to maps, an overwrite to the scoreboard reference has been
added for when another world has been created.
This should also address BUKKIT-3982 and BUKKIT-3985
In commit 7710efc5f9 we corrected the handling of large chests as the
destination for hoppers moving items but did not apply the same fix for
large chests being the source or for droppers. This commit updates these
to have the same fix.
This implementation facilitates the correspondence of the Bukkit Scoreboard
API to the internal minecraft implementation.
When the first scoreboard is loaded, the scoreboard manager will be created.
It uses the newly added WeakCollection for handling plugin scoreboard
references to update the respective objectives. When a scoreboard contains no
more active references, it should be garbage collected.
An active reference can be held by a still registered objective, team, and
transitively a score for a still registered objective. An internal reference
will also be kept if a player's specific scoreboard has been set, and will
remain persistent until that player logs out.
A player's specific scoreboard becomes the scoreboard used when determining
team structure for the player's attacking damage and the player's vision.
This adds calls to BlockRedstoneEvent for the new daylight sensor and
trapped chest blocks. Note that the redstone level for trapped chests
cannot be modified, as it is based on the number of players currently
viewing the chest's inventory.
When a block placement happens we currently update physics on the
attempted placement and update again if the placement is cancelled.
To correct the first one we simply set the block without applying
physics. To correct the second we have to add a new method to
BlockState that lets us update without applying physics and use
this method method when putting the block back.
Large chests work in a different fashion as they are a combination of
two other inventories. This causes their getOwner method to always return
null as their is no correct return. To compensate for this for the hopper
events we special case them to use their CraftBukkit counterpart that has
the information we need for the event.
When a splash potion has no applicable effects we currently do not call
PotionSplashEvent. This means plugins are unable to make custom
potions with reliable splash handling as they have to relicate the
functionality themselves.
With this commit we simply make the event fire regardless of the effects
on the potion.
I should try to compile before I say "this change is okay".
I should try to compile before I say "this change is okay".
I should try to compile before I say "this change is okay".
I should try to compile before I say "this change is okay".
for i in range(100)
This causes the server to generate PrepareItemEnchantEvent even in the
case that an item is already enchanted or otherwise would normally not
be enchantable.
The client resets all formatting after a color code is received, but currently the ANSI codes do not, and so the console does not accurately reflect the appearance of the formatted text. Instead, the ANSI color codes are now set to reset all text attributes.
Currently when dealing with physical interactions with pressure plates
and tripwires we immediately block their activation as soon as a single
entity involved has their event cancelled. We also fire events whenever
an entity intersects the block a wooden button is in even if they aren't
actually pressing it. To correct this we move the button interaction to
the correct place and modify all three to only block the activation if
every entity is blocked from using them instead of just one of them.
If the server changes the weather it will set the per-player weather
variable and future changes will not apply. We should only set this
variable when a plugin is requesting per-player weather and not when
the server it doing it.
We used to fall Item.filterData() for this but that method is meant for
converting item data to block data during placement and does the wrong
thing for this case. Instead we just see if the item should have data and
if not set it to zero. We also have to filter wool data explicitly because
clients crash when given invalid wool data.
In Minecraft 1.5 saplings do not grow with a single use of bonemeal anymore.
Our code assumes they will and only takes away bonemeal from the player
when the tree grows successfully (not cancelled by a plugin). Instead we
now always remove a bonemeal even if a plugin is the reason a tree didn't
grow as this matches the vanilla logic more closely.
If a custom TravelAgent is used and returns null for findOrCreate method
a NullPointerException will occur.
Conflicts:
src/main/java/net/minecraft/server/PlayerList.java
Recent changes caused PlayerPortalEvent to suddenly return null
unexpectedly and could end up in NPEs resulting that did not before.
This commit addresses that situation by always ensuring a TravelAgent
instance is returned.
The TravelAgent for world 0 is returned arbitrarily in an effort to
compensate for plugins that are implementation dependent and expect some
form of a TravelAgent to be accessible in the event at all times.
Vanilla does not check for blocks in which the player could
suffocate when changing dimension, so portals will happily spawn
players in blocks when using a portal under certain
circumstances. However, we currently check for these instances
and move the player up until they will not suffocate. This means
that players can sometimes be taken to above the target portal,
making it seem as if a portal was not created. Instead, we now
disable this suffocation check when moveToWorld is called from
changeDimension, mirroring vanilla behavior more accurately.
Due to the having to generate new logic to avoid using the customized
PlayerConnection.moveToWorld, entities returning from The End were not
properly calculating their exit target. This commit corrects that
logic.
By having a single function to process BlockPlacement logic, we make
it so that there is consistent behavior throughout all BlockPlace
events. This should allow for easier troubleshooting and less diffs
in source.
This also fixes BUKKIT-3463 by including the correct coordinates that
were clicked to the event.
Also fixes: BUKKIT-3477 and BUKKIT-3488
Minecraft likes to double check that tile entities get set after they
are placed, however we didn't set tile entities until after our event
was called. This caused the world to have multiple tile entities in a
single block location; to fix this we now set tile entities before
the event.
When the skull BlockPlaceEvent was added it was made so the event
would be called after all the data has been set, however this is a
behavior change that is inconsistent with other BlockPlaceEvents.
Instead, if people wish to get the block data they should schedule
a task.
Relates to: BUKKIT-3438
When either of those settings are false, the worlds are not loaded and
therefore will not be targeted for portal exits. Existing worlds are
iterated directly to avoid defaulting to the first world if a direct
dimension match is not found.
Plugins must also specify exit from custom Bukkit worlds to comply with
original commit: https://github.com/Bukkit/CraftBukkit/commit/2dc2af0
This commit introduces a constant to clarify the dependency on the
CraftBukkit implementation of custom worlds having a dimension offset.
By returning the following value (7) we remove the need to special
case pistons in any way (other than the original purpose of this
check, which is to ensure pistons have valid data)
The previous logic was faulty since it lost the logic of "placing" the
block. It was also taking into account data that could have been
changed outside of the processing of this event, which is irrelevant
to the processing of this event.
By using return 0, we exit the loop prematurely preventing other
creature types from being spawned if one type is set to 0. By using
continue we move on to the other types and allow them to spawn
properly.
This makes it so animals (tame or not) will sit properly and not move
around.
Wild animals that are sitting may override the sitting position if
they are attacking.
The 'tag' NBTTagCompound field of the ItemStack assumes that it is OK to
save a reference to an NBT supplied via load() and assumes it is OK to
supply a reference to the internal field during a save(). Neither is true,
as Chunk NBT structures are required to be read-only once created (due to
being written asynchronously off the server thread AND due to the potential
to be passed to a new Chunk if the same chunk is reloaded before the
writing of the NBT is completed by the File I/O thread). Keeping a live
reference to the NBT copy passed in, or to the NBT value passed back
during saving, creates serious thread safety issues which can result in
corrupted data being written to the world data files.
The specific issue here was uncovered by the recent change to use
setName("") on the ItemStack.tag object. When a chunk is being loaded
again before its save is completed, this results in name of the field
in the NBT being set to "". This causes it to be saved as "" instead
of "tag" resulting in it not being properly reloaded in the future which
results in the itemstack losing all of its metadata.
This adds two settings to bukkit.yml, allowing activation and control of
two chunk garbage collection triggering conditions:
chunk-gc/period-in-ticks controls a periodic GC, run once every N ticks
(default is 600); chunk-gc/load-threshold causes the GC to run once
after every N calls to loadChunk() on a given world (this call is an API
call used by plugins, and is distinct from the path taken for routine
player movement-based loading). In both cases, setting to zero will
disable the given GC scheduling strategy.
In either case, the act of doing the GC is simply one of scanning the
loaded chunks, seeing which are NOT being used by one or more players
(due to view-distance) and which are not already queued for unload, and
queueing them for a normal unload. Ultimately, the unload is then
processed the same as if the chunk were unloaded due to leaving the
view-distance range of all players, so the impact on plugins should be
no different (and strategies such as handling the ChunkUnloadEvent in
order to prevent unload will still work).
The initial interval for the periodic GC is randomized on a per-world
basis, in order to avoid all world being GCed at the same time -
minimizing potential lag spikes.
With the persistence api introduced, pets did not have their
persistence flag updated to reflect their persistence. This caused
tame ocelots to not persist under specific conditions.
Slimes and wolves have health that can change based on certain
conditions. So we check if their max health should be updated, and if
it has been customized in any way.
We also scale the wolf's health for their tail
An ItemStack gains the tag name "tag" when the stack is serialized
to NBT, however items don't have a tag *until* they are serialized at
least once. So to solve this, we remove the tag name when loading the
NBT data.
Another problem with NBT are TagLists, when transferring tag lists
between the server and the client the names are lost, and so we
simply don't add a name to the tag.
If you cancel a BlockPlaceEvent for a sign the world is updated as if
the block was placed and then destroyed. To avoid this we set the block
without updating physics then apply the update after the event.
When unloading chunks we have a check to ensure we do not remove players
from the world due to the issues this would cause. However, our check
to see if the player is in this chunk is reversed and is in fact entirely
wrong. Even if the player isn't currently in this chunk we do not want
to remove them as that will still cause the same issues.
The key "direction" incorrectly mapped to variables that were already
set in the entity. In order to prevent loading incorrect data we
renamed "direction" to "power."
The player would have no permissions (other than their OP status)
when checked in the Quit event. This is because we removed permissions
before the event occurred. By calling it afterwards, we can persist
the data until the server finally removes the player.
With 1.4, entity sound tracking changed for the better.
Our previous method additions can now be removed.
All that's left is checking if the source can be seen
by the recipient of the sound packet. Thanks, Mojang!
When a player triggers a chunk load via walking around or teleporting there
is no need to stop everything and get this chunk on the main thread. The
client is used to having to wait some time for this chunk and the server
doesn't immediately do anything with it except send it to the player. At
the same time chunk loading is the last major source of file IO that still
runs on the main thread.
These two facts make it possible to offload chunks loaded for this reason
to another thread. However, not all parts of chunk loading can happen off
the main thread. For this we use the new AsynchronousExecutor system to
split chunk loading in to three pieces. The first is loading data from
disk, decompressing it, and parsing it in to an NBT structure. The second
piece is creating entities and tile entities in the chunk and adding them
to the world, this is still done on the main thread. The third piece is
informing everyone who requested a chunk load that the load is finished.
For this we register callbacks and then run them on the main thread once
the previous two stages are finished.
There are still cases where a chunk is needed immediately and these will
still trigger chunk loading entirely on the main thread. The most obvious
case is plugins using the API to request a chunk load. We also must load
the chunk immediately when something in the world tries to access it. In
these cases we ignore any possibly pending or in progress chunk loading
that is happening asynchronously as we will have the chunk loaded by the
time they are finished.
The hope is that overall this system will result in less CPU time and
pauses due to blocking file IO on the main thread thus giving more
consistent performance. Testing so far has shown that this also speeds up
chunk loading client side although some of this is likely to be because
we are sending less chunks at once for the client to process.
Thanks for @ammaraskar for help with the implementation of this feature.
When a player has canPickUpLoot set to true the code for mob pickup is
triggerd which does not know how to deal with player inventory. Since
players have their own logic for picking up items we simply disable this
code for them.
The old flag for picking up loot was default to false, making existing players not able to pickup items. We now use this flag for Players, which gives us the problem we had in 48b46f83.
To fix this, we add an incremental flag that will be cross-examined to check if the data was saved before or after the flag level was introduced.
Addresses BUKKIT-3143
As an added feature, players defaulted to being able to not pick up items if the flag was false. However, since minecraft doesn't normally use the flag on players, the flag was always false.
Adds:
- Getting/Setting equipment
- getting/setting drop rates
- getting/setting ability to pick up items
-- As an added feature, players with this flag start off with a canceled PlayerPickupItemEvent
When a mob is marked with the persistent flag (animal or anything with
setRemoveWhenFarAway(false)) the entire block of code for checking if they
should be despawned is skipped. However, one part of this code updates the
mob state if a player is close enough to them. It turns out this state is
used by the AI system to decide if the mob should move around randomly or
not. To stop mobs from being frozen in place we now update this state if
the persistent flag is set as well.
The old default for the persistent flag on mobs was false which was then
written out to their NBT data when they were saved. We now use this data
for all mobs, not just non-animal mobs. However, this means animals that
spawned before that change will now start despawning like monsters do.
To avoid this we add a new flag to the mob's saved data to mark if the
data was saved before or after we started using it and ignore it if it
was before.
As of 1.4 mobs have a flag to determine if they despawn when away from a
player or not. Unfortunately animals still use their own system to prevent
despawning instead of making use of this flag. This change modifies them
to use the new system (defaults to true) and to add API for plugins to adjust
this.
If the player is not in Creative (i.e. does not have the ability to
instantly build) we need to decrement the MonsterEgg item stack when used
on a breedable parent mob.
On join we unconditionally add the player to the world they logged out in.
If a plugin teleports a player during PlayerJoinEvent in a way that adds
them to a world (cross-world teleport) we end up with one player in two
places. To avoid this we check to see if the player has changed worlds or
is already added to the world we have we skip adding them again.
When sending chunks to a player we use their writer thread to do chunk
compression to avoid blocking the main thread with this work. However,
after a teleport or respawn there are a large number of chunk packets to
process. This causes the thread to spend a long period handling compression
while we continue dumping more chunk packets on it to handle. The result of
this is a noticable delay in getting responses to commands and chat
immediately after teleporting.
Switching to a lower compression level reduces this load and makes our
behavior more like vanilla. We do, however, still give this thread more
work to do so there will likely still be some delay when comparing to
vanilla. The only way to avoid this would be to put chunk compression back
on the main thread and give everyone on the server a poorer experience
instead.
When an event changes the item to be dispensed we check to see if the new
item has special behavior for dispensing and if so pass it on to that
behavior handler. However, we are actually checking the old itemstack and
passing the new itemstack so this check fails.
If a player travels past 32,000,000 blocks on the X or Z coordinates they
will be kicked for having an illegal position. On kick their player data
is saved which includes their (illegal) position. This means on join they
are immediately kicked again for the same reason and are stuck. Instead of
kicking at all in this case just teleport the player back to their previous
position just like the moved wrongly check does.
In order to correctly handle disconnects for invalid chat we setup a
Waitable and pass it to the main thread then wait for it to be processed.
However, commands are also chat packets and they are already on the main
thread. In this case, waiting will deadlock the server so we should just
do a normal disconnect.
End portals can only be placed in the end during the dragon's death.
Attempts to place them outside of this window causes the block to remove
itself. However, we still create the tile entity for the portal which
leads to exceptions spamming the console about a tile entity existing
without the appropriate block. In these cases we should not place the tile
entity at all.
When invalid chat is detected we currently drop the connection with no
hint as to why as anything else is not allowed while we're off the main
thread. To give valid disconnect reasons and fire proper events instead
pass these off to the main thread and wait for it to process them.
If a plugin cancels a PlayerInteractEvent when left clicking a block the
client may have removed this block if they are in creative mode or if the
block breaks in a single hit. In this case, we need to update the client's
tile entity as well as telling it the block still exists.
