This is about as optimized as it can get. Thanks to @Leymooo for the idea, I simply expanded on it. We optimize for the common 1-3 byte cases, and punt more "complicated" cases to the original VarInt writing function we had before.
Inspired by the approach described at the bottom of https://richardstartin.github.io/posts/dont-use-protobuf-for-telemetry
Given that we do a lot of varint writing as well, this should provide a small performance boost for larger/complex packets whilst not regressing hard on smaller packets.
This includes a test to ensure that the behavior is as expected and fixes the initialization loop so that the correct results will be given. Much thanks to @octylFractal for acting as my duck while trying to figure this out.
Inspired by the approach described at the bottom of https://richardstartin.github.io/posts/dont-use-protobuf-for-telemetry
Given that we do a lot of varint writing as well, this should provide a small performance boost for larger/complex packets whilst not regressing hard on smaller packets.
Mitigates attacks like the one described in SpigotMC/BungeeCord#3066. This cannot be considered a full protection, only a mitigation that expects full packets. The attack described is essentially the infamous Slowloris attack.
Someone discovered this bug but decided not to tell anyone. For the record, I'm the one who was brain-damaged... well, I made that particular copy-paste error.
This reverts commit 4f80d2b261d0a90d9076474e9e5d55f63fd171d9.
Experience elsewhere (Waterfall PR) and confirmation from TCPShield means
this ought to work. Let's hope.
This commit simplifies the Metrics class by using the new base module
from bStats. It also migrates to the new bStats config file that will
be used by plugins that integrate bStats. If a user disabled bStats in
the old config file, this setting will be copied to the new config file.
