[ViewVC] Diff of: cvs/libev/Changes

Comparing libev/Changes (file contents):
Revision 1.254 by root, Mon Apr 2 18:39:54 2012 UTC vs.
Revision 1.352 by root, Tue Dec 24 13:27:28 2019 UTC

…		…
1	Revision history for libev, a high-performance and full-featured event loop.	1	Revision history for libev, a high-performance and full-featured event loop.
2		2
3	TODO: ev_loop_wakeup	3	TODO: revisit 59.x timer in the light of modern powersaving
4	TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone)	4	TODO: document EV_TSTAMP_T
5		5
6	TODO: document WSA_EV_USE_SOCKET in win32 part	6	- the new timerfd code wrongly changed the priority of ther signal
7	TODO: ^ OR use WSASend/WSARecv on the handle, which always works	7	fd watcher, which is usuallys harmless unless signal fds are
8	TODO: use __OPTIMIZE__ or __OPTIMIZE_SIZE__?	8	also used (found via cpan tester service).
		9
		10	4.31 Fri Dec 20 21:58:29 CET 2019
		11	- handle backends with minimum wait time a bit better by not
		12	waiting in the presence of already-expired timers
		13	(behaviour reported by Felipe Gasper).
		14	- new feature: use timerfd to detect timejumps quickly,
		15	can be disabled with the new EVFLAG_NOTIMERFD loop flag.
		16	- document EV_USE_SIGNALFD feature macro.
		17
		18	4.30 (EV only)
		19	- change non-autoconf test for __kernel_rwf_t by testing
		20	LINUX_VERSION_CODE, the most direct test I could find.
		21	- fix a bug in the io_uring backend that polled the wrong
		22	backend fd, causing it to not work in many cases.
		23
		24	4.29 (EV only)
		25	- add io uring autoconf and non-autoconf detection.
		26	- disable io_uring when some header files are too old.
		27
		28	4.28 (EV only)
		29	- linuxaio backend resulted in random memory corruption
		30	when loop is forked.
		31	- linuxaio backend might have tried to cancel an iocb
		32	multiple times (was unable to trigger this).
		33	- linuxaio backend now employs a generation counter to
		34	avoid handling spurious events from cancelled requests.
		35	- io_cancel can return EINTR, deal with it. also, assume
		36	io_submit also returns EINTR.
		37	- fix some other minor bugs in linuxaio backend.
		38	- ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
		39	- cleanup: replace expect_true/false and noinline by their
		40	libecb counterparts.
		41	- move syscall infrastructure from ev_linuxaio.c to ev.c.
		42	- prepare io_uring integration.
		43	- tweak ev_floor.
		44	- epoll, poll, win32 Sleep and other places that use millisecond
		45	reslution now all try to round up times.
		46	- solaris port backend didn't compile.
		47	- abstract time constants into their macros, for more flexibility.
		48
		49	4.27 Thu Jun 27 22:43:44 CEST 2019
		50	- linux aio backend almost completely rewritten to work around its
		51	limitations.
		52	- linux aio backend now requires linux 4.19+.
		53	- epoll backend now mandatory for linux aio backend.
		54	- fail assertions more aggressively on invalid fd's detected
		55	in the event loop, do not just silently fd_kill in case of
		56	user error.
		57	- ev_io_start/ev_io_stop now verify the watcher fd using
		58	a syscall when EV_VERIFY is 2 or higher.
		59
		60	4.26 (EV only)
		61	- update to libecb 0x00010006.
		62	- new experimental linux aio backend (linux 4.18+).
		63	- removed redundant 0-ptr check in ev_once.
		64	- updated/extended ev_set_allocator documentation.
		65	- replaced EMPTY2 macro by array_needsize_noinit.
		66	- minor code cleanups.
		67	- epoll backend now uses epoll_create1 also after fork.
		68
		69	4.25 Fri Dec 21 07:49:20 CET 2018
		70	- INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
		71	(EV_THROW still provided) and now uses noexcept on C++11 or newer.
		72	- move the darwin select workaround higher in ev.c, as newer versions of
		73	darwin managed to break their broken select even more.
		74	- ANDROID => __ANDROID__ (reported by enh@google.com).
		75	- disable epoll_create1 on android because it has broken header files
		76	and google is unwilling to fix them (reported by enh@google.com).
		77	- avoid a minor compilation warning on win32.
		78	- c++: remove deprecated dynamic throw() specifications.
		79	- c++: improve the (unsupported) bad_loop exception class.
		80	- backport perl ev_periodic example to C, untested.
		81	- update libecb, biggets change is to include a memory fence
		82	in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
		83	- minor autoconf/automake modernisation.
		84
		85	4.24 Wed Dec 28 05:19:55 CET 2016
		86	- bump version to 4.24, as the release tarball inexplicably
		87	didn't have the right version in ev.h, even though the cvs-tagged
		88	version did have the right one (reported by Ales Teska).
		89
		90	4.23 Wed Nov 16 18:23:41 CET 2016
		91	- move some declarations at the beginning to help certain retarded
		92	microsoft compilers, even though their documentation claims
		93	otherwise (reported by Ruslan Osmanov).
		94
		95	4.22 Sun Dec 20 22:11:50 CET 2015
		96	- when epoll detects unremovable fds in the fd set, rebuild
		97	only the epoll descriptor, not the signal pipe, to avoid
		98	SIGPIPE in ev_async_send. This doesn't solve it on fork,
		99	so document what needs to be done in ev_loop_fork
		100	(analyzed by Benjamin Mahler).
		101	- remove superfluous sys/timeb.h include on win32
		102	(analyzed by Jason Madden).
		103	- updated libecb.
		104
		105	4.20 Sat Jun 20 13:01:43 CEST 2015
		106	- prefer noexcept over throw () with C++ 11.
		107	- update ecb.h due to incompatibilities with c11.
		108	- fix a potential aliasing issue when reading and writing
		109	watcher callbacks.
		110
		111	4.19 Thu Sep 25 08:18:25 CEST 2014
		112	- ev.h wasn't valid C++ anymore, which tripped compilers other than
		113	clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
		114	C++ doesn't support typedefs for function pointers fully, so the affected
		115	declarations have to spell out the types each time.
		116	- when not using autoconf, tighten the check for clock_gettime and related
		117	functionality.
		118
		119	4.18 Fri Sep 5 17:55:26 CEST 2014
		120	- events on files were not always generated properly with the
		121	epoll backend (testcase by Assaf Inbal).
		122	- mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
		123	- (ecb) support m68k, m88k and sh (patch by Miod Vallat).
		124	- use a reasonable fallback for EV_NSIG instead of erroring out
		125	when we can't detect the signal set size.
		126	- in the absence of autoconf, do not use the clock syscall
		127	on glibc >= 2.17 (avoids the syscall AND -lrt on systems
		128	doing clock_gettime in userspace).
		129	- ensure extern "C" function pointers are used for externally-visible
		130	loop callbacks (not watcher callbacks yet).
		131	- (ecb) work around memory barriers and volatile apparently both being
		132	broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
		133
		134	4.15 Fri Mar 1 12:04:50 CET 2013
		135	- destroying a non-default loop would stop the global waitpid
		136	watcher (Denis Bilenko).
		137	- queueing pending watchers of higher priority from a watcher now invokes
		138	them in a timely fashion (reported by Denis Bilenko).
		139	- add throw() to all libev functions that cannot throw exceptions, for
		140	further code size decrease when compiling for C++.
		141	- add throw () to callbacks that must not throw exceptions (allocator,
		142	syserr, loop acquire/release, periodic reschedule cbs).
		143	- fix event_base_loop return code, add event_get_callback, event_base_new,
		144	event_base_get_method calls to improve libevent 1.x emulation and add
		145	some libevent 2.x functionality (based on a patch by Jeff Davey).
		146	- add more memory fences to fix a bug reported by Jeff Davey. Better
		147	be overfenced than underprotected.
		148	- ev_run now returns a boolean status (true meaning watchers are
		149	still active).
9	- ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with	150	- ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
10	libev's EV_ERROR (reported by 191919).	151	libev's EV_ERROR (reported by 191919).
11
12	TODO: Jeff Davey libev patch
13	- (ecb) add memory fence support for xlC (Darin McBride).	152	- (ecb) add memory fence support for xlC (Darin McBride).
14	- (ecb) add memory fence support for gcc-mips (Anton Kirilov).	153	- (ecb) add memory fence support for gcc-mips (Anton Kirilov).
		154	- (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
15	- work around some kernels losing file descriptors by leaking	155	- work around some kernels losing file descriptors by leaking
16	the kqueue descriptor in the child.	156	the kqueue descriptor in the child.
		157	- work around linux inotify not reporting IN_ATTRIB changes for directories
		158	in many cases.
		159	- include sys/syscall.h instead of plain syscall.h.
		160	- check for io watcher loops in ev_verify, check for the most
		161	common reported usage bug in ev_io_start.
		162	- choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
		163	- always use WSASend/WSARecv directly on windows, hoping that this
		164	works in all cases (unlike read/write/send/recv...).
		165	- try to detect signals around a fork faster (test program by
		166	Denis Bilenko).
		167	- work around recent glibc versions that leak memory in realloc.
		168	- rename ev::embed::set to ev::embed::set_embed to avoid clashing
		169	the watcher base set (loop) method.
		170	- rewrite the async/signal pipe logic to always keep a valid fd, which
		171	simplifies (and hopefully correctifies :) the race checking
		172	on fork, at the cost of one extra fd.
		173	- add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
		174	inotify-supporting filesystems.
		175	- move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
		176	versions ignore it before
		177	(https://bugzilla.redhat.com/show_bug.cgi?id=908096).
		178	- add some untested android support.
		179	- enum expressions must be of type int (reported by Juan Pablo L).
17		180
18	4.11 Sat Feb 4 19:52:39 CET 2012	181	4.11 Sat Feb 4 19:52:39 CET 2012
19	- INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as	182	- INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
20	was documented already, but not implemented in the repeating case.	183	was documented already, but not implemented in the repeating case.
21	- new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.	184	- new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
205	reinitialising the default loop actually works (patch by	368	reinitialising the default loop actually works (patch by
206	Malek Hadj-Ali).	369	Malek Hadj-Ali).
207	- implement ev_suspend and ev_resume.	370	- implement ev_suspend and ev_resume.
208	- new EV_CUSTOM revents flag for use by applications.	371	- new EV_CUSTOM revents flag for use by applications.
209	- add documentation section about priorities.	372	- add documentation section about priorities.
210	- add a glossary to the dcoumentation.	373	- add a glossary to the documentation.
211	- extend the ev_fork description slightly.	374	- extend the ev_fork description slightly.
212	- optimize a jump out of call_pending.	375	- optimize a jump out of call_pending.
213		376
214	3.53 Sun Feb 15 02:38:20 CET 2009	377	3.53 Sun Feb 15 02:38:20 CET 2009
215	- fix a bug in event pipe creation on win32 that would cause a	378	- fix a bug in event pipe creation on win32 that would cause a

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing libev/Changes (file contents): Revision 1.254 by root, Mon Apr 2 18:39:54 2012 UTC vs. Revision 1.352 by root, Tue Dec 24 13:27:28 2019 UTC

Diff Legend

Comparing libev/Changes (file contents):
Revision 1.254 by root, Mon Apr 2 18:39:54 2012 UTC vs.
Revision 1.352 by root, Tue Dec 24 13:27:28 2019 UTC