When the connection between a minecraft client and a server
(e.g. me and the spigot test server) is bad sometimes two placement
packets will arrive at the same time. When the extra one gets canceled
any blocks placed against it get rejected by the server causing the
whole chain to disappear. This fixes this by allowing more packets
to occur whilst making the window wider (since the client seems to be
capped at about one placement every 100ms on a normal connection)
This code was added in 1c84a5dd8c to alleviate some crashes. Since then we have improved our thread safety and detection of unsafe thread behaviour and as such there is no need to incur thhis performance hit anymore.
When on Java 7 we can register the classloader as parallel capable to prevent deadlocks caused by certain scenarios. Due to the nature of PluginClassLoader this isn't completely safe, but we can make it safer by switching to concurrency focused collections. Either way this is far better than crashing the server.
If you must use a plugin which relies on unsafe versioned code, please pick a library such as ProtocolLib by @aadnk which other plugins can depend on, instead of having each plugin implement their own risky behaviour.
This flag contains the same message as eula.txt (albeit with proper grammar for 3rd person plural). By setting this flag you are agreeing to the terms listed in the message printed upon setting this flag, which are no different from those agreed to by any other method.
This feature only exists for hosting companies and people who manage servers in bulk who wish to agree to the terms by setting a flag in a script.
Once and for all.
Thanks to Thinkofdeath & the users who put up with these changes, it was for a good cause, however at this point in time we do not feel it is possible to correct the large number of exploits present in player movement without the cooperation of Mojang. We will be seeking their cooperation, however in the mean time it is recommended that users install an anticheat plugin such as NoCheat+ to ensure some level of protection.
This will fix some strange behaviour users have been seeing in relation to player movements.
* Now Tracks Commands
* Resolved Performance issue with Bukkit HandlerList method. Now all timings simply uses the Spigot System
* Performance Improvement simply not using TimedRegisteredListener too
* Bug with SyncChunkLoad tracking resolved. Now properly tracks many aspects of sync chunk load.
* Reset/On/Off accuracy - should no longer have any issues turning it on/off during runtime, so this has been re-enabled.
* Paste command on RCON now works
* Now tracks everything related to plugins too, so you can easily see total plugin cost
* Now tracks Tasks better and where they came from
* Now tracks plugins event handlers to the Listener/Method name too.
* Merged some Bukkit Patches so all timings changes are in 1 patch.
* Moved back to a CLQ for CustomTimingsHandler for thread safety for when tasks are created Async but then executed sync.
Hopefully no plugins are somehow leaking buffers, but disabling the ResourceLeakDetector is a good thing for performance of the Netty IO subsytem (it's been disabled in BungeeCord for a while now).
