Closed MachineBreaker closed 1 year ago
You can report it to Paper too https://github.com/PaperMC/Paper/blob/master/patches/server/0710-Optimise-non-flush-packet-sending.patch
I didn't find the related patch in Paper so i reported it here, gonna do it later
You didn't report it?
This may be really useful to fix, because with the latest netty I think it removes LazyRunnable from working completely.
LazyRunnable shouldn't be added again unless you workaround the JDK-8180450 bug. This indeed makes packet writing slow (from 880ms to 1450ms)
LazyRunnable shouldn't be added again unless you workaround the JDK-8180450 bug. This indeed makes packet writing slow (from 880ms to 1450ms)
I think you should report it to netty then, this feature is still available, but in a different method.
LazyRunnable shouldn't be added again unless you workaround the JDK-8180450 bug. This indeed makes packet writing slow (from 880ms to 1450ms)
I think you should report it to netty then, this feature is still available, but in a different method.
Netty already provided a solution, not my job to implement it
What solution?
What solution?
Override SingleThreadEventExecutor#wakesUpForTask I also fixed it directly on my netty fork without getting the always-on performance hit but yeah, there's no way to use it here
Type check vs LazyRunnable receive an always-on performance hit for a common operation that users shouldn't pay if they won't use it.
https://github.com/netty/netty/commit/de876d197a945e2b3a4a96b90a65c00611b77806