Revision history for IO-Async-Loop-Epoll
0.13 CHANGES:
* Rewrite to use Linux::Epoll instead, which has nicer scaling than
IO::Epoll
0.12 BUGFIXES:
* Only declare known support for on_hangup on Linux just on the
offchance that some non-Linux OS ever gains epoll
* Fix timing-related bugs by rounding epoll_pwait() timeout up to the
next milisecond (RT73498)
* on_hangup also needs to test for EPOLLERR condition
0.11 CHANGES:
* Dynamically scale epoll_pwait() maxevents parameter to avoid
stalling higher-numbered FDs during high load
* Emulate non-epoll'able filehandles (e.g. regular files) as always
read/write-ready
0.10 BUGFIXES:
* Try to workaround perl blobk/unblock signals bug by using an unsafe
POSIX::SigAction handler
0.09 CHANGES:
* Updated API version for IO::Async 0.33
0.08 CHANGES:
* Full example in SYNOPSIS section
BUGFIXES:
* Handle EPOLLERR as on_read_ready
0.07 CHANGES:
* Updated for IO-Async 0.24
* Use shared acceptance testing provided by IO::Async::LoopTests
0.06 CHANGES:
* Various small updates to keep CPANTS happy
0.05 CHANGES:
* Added 'use warnings'
BUGFIXES:
* Don't rely on writability of STDOUT during test scripts
0.04 CHANGES:
* Updated for IO-Async 0.20
0.03 BUGFIXES:
* Better handling of write-only Notifiers
0.02 CHANGES:
* Use low-level IO::Epoll interface (epoll_ctl() etc..) instead of
high-level IO::Poll-alike object wrapper. Should be better
performance, and less buggy - doesn't depend on poor IO::Poll-like
emulation details.
0.01 First version, released on an unsuspecting world.