ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/Changes
(Generate patch)

Comparing libev/Changes (file contents):
Revision 1.257 by root, Mon Apr 2 23:46:27 2012 UTC vs.
Revision 1.355 by root, Fri Dec 27 16:12:37 2019 UTC

1Revision history for libev, a high-performance and full-featured event loop. 1Revision history for libev, a high-performance and full-featured event loop.
2 2
3TODO: ev_loop_wakeup 3TODO: revisit 59.x timer in the light of modern powersaving
4TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone) 4TODO: ^ maybe just increase it when timerfd, modern linux only thing that counts?
5TODO: document EV_TSTAMP_T
5 6
6TODO: document WSA_EV_USE_SOCKET in win32 part 7 - the 4.31 timerfd code wrongly changes the priority of the signal
7TODO: ^ OR use WSASend/WSARecv on the handle, which always works 8 fd watcher, which is usually harmless unless signal fds are
9 also used (found via cpan tester service).
10 - support a lot more "uncommon" compile time configurations,
11 such as ev_embed enabled but ev_timer disabled.
12 - the linux aio backend is no longer compiled in by default.
8 13
144.31 Fri Dec 20 21:58:29 CET 2019
15 - handle backends with minimum wait time a bit better by not
16 waiting in the presence of already-expired timers
17 (behaviour reported by Felipe Gasper).
18 - new feature: use timerfd to detect timejumps quickly,
19 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
20 - document EV_USE_SIGNALFD feature macro.
21
224.30 (EV only)
23 - change non-autoconf test for __kernel_rwf_t by testing
24 LINUX_VERSION_CODE, the most direct test I could find.
25 - fix a bug in the io_uring backend that polled the wrong
26 backend fd, causing it to not work in many cases.
27
284.29 (EV only)
29 - add io uring autoconf and non-autoconf detection.
30 - disable io_uring when some header files are too old.
31
324.28 (EV only)
33 - linuxaio backend resulted in random memory corruption
34 when loop is forked.
35 - linuxaio backend might have tried to cancel an iocb
36 multiple times (was unable to trigger this).
37 - linuxaio backend now employs a generation counter to
38 avoid handling spurious events from cancelled requests.
39 - io_cancel can return EINTR, deal with it. also, assume
40 io_submit also returns EINTR.
41 - fix some other minor bugs in linuxaio backend.
42 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
43 - cleanup: replace expect_true/false and noinline by their
44 libecb counterparts.
45 - move syscall infrastructure from ev_linuxaio.c to ev.c.
46 - prepare io_uring integration.
47 - tweak ev_floor.
48 - epoll, poll, win32 Sleep and other places that use millisecond
49 reslution now all try to round up times.
50 - solaris port backend didn't compile.
51 - abstract time constants into their macros, for more flexibility.
52
534.27 Thu Jun 27 22:43:44 CEST 2019
54 - linux aio backend almost completely rewritten to work around its
55 limitations.
56 - linux aio backend now requires linux 4.19+.
57 - epoll backend now mandatory for linux aio backend.
58 - fail assertions more aggressively on invalid fd's detected
59 in the event loop, do not just silently fd_kill in case of
60 user error.
61 - ev_io_start/ev_io_stop now verify the watcher fd using
62 a syscall when EV_VERIFY is 2 or higher.
63
644.26 (EV only)
65 - update to libecb 0x00010006.
66 - new experimental linux aio backend (linux 4.18+).
67 - removed redundant 0-ptr check in ev_once.
68 - updated/extended ev_set_allocator documentation.
69 - replaced EMPTY2 macro by array_needsize_noinit.
70 - minor code cleanups.
71 - epoll backend now uses epoll_create1 also after fork.
72
734.25 Fri Dec 21 07:49:20 CET 2018
74 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
75 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
76 - move the darwin select workaround higher in ev.c, as newer versions of
77 darwin managed to break their broken select even more.
78 - ANDROID => __ANDROID__ (reported by enh@google.com).
79 - disable epoll_create1 on android because it has broken header files
80 and google is unwilling to fix them (reported by enh@google.com).
81 - avoid a minor compilation warning on win32.
82 - c++: remove deprecated dynamic throw() specifications.
83 - c++: improve the (unsupported) bad_loop exception class.
84 - backport perl ev_periodic example to C, untested.
85 - update libecb, biggets change is to include a memory fence
86 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
87 - minor autoconf/automake modernisation.
88
894.24 Wed Dec 28 05:19:55 CET 2016
90 - bump version to 4.24, as the release tarball inexplicably
91 didn't have the right version in ev.h, even though the cvs-tagged
92 version did have the right one (reported by Ales Teska).
93
944.23 Wed Nov 16 18:23:41 CET 2016
95 - move some declarations at the beginning to help certain retarded
96 microsoft compilers, even though their documentation claims
97 otherwise (reported by Ruslan Osmanov).
98
994.22 Sun Dec 20 22:11:50 CET 2015
100 - when epoll detects unremovable fds in the fd set, rebuild
101 only the epoll descriptor, not the signal pipe, to avoid
102 SIGPIPE in ev_async_send. This doesn't solve it on fork,
103 so document what needs to be done in ev_loop_fork
104 (analyzed by Benjamin Mahler).
105 - remove superfluous sys/timeb.h include on win32
106 (analyzed by Jason Madden).
107 - updated libecb.
108
1094.20 Sat Jun 20 13:01:43 CEST 2015
110 - prefer noexcept over throw () with C++ 11.
111 - update ecb.h due to incompatibilities with c11.
112 - fix a potential aliasing issue when reading and writing
113 watcher callbacks.
114
1154.19 Thu Sep 25 08:18:25 CEST 2014
116 - ev.h wasn't valid C++ anymore, which tripped compilers other than
117 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
118 C++ doesn't support typedefs for function pointers fully, so the affected
119 declarations have to spell out the types each time.
120 - when not using autoconf, tighten the check for clock_gettime and related
121 functionality.
122
1234.18 Fri Sep 5 17:55:26 CEST 2014
124 - events on files were not always generated properly with the
125 epoll backend (testcase by Assaf Inbal).
126 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
127 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
128 - use a reasonable fallback for EV_NSIG instead of erroring out
129 when we can't detect the signal set size.
130 - in the absence of autoconf, do not use the clock syscall
131 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
132 doing clock_gettime in userspace).
133 - ensure extern "C" function pointers are used for externally-visible
134 loop callbacks (not watcher callbacks yet).
135 - (ecb) work around memory barriers and volatile apparently both being
136 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
137
1384.15 Fri Mar 1 12:04:50 CET 2013
139 - destroying a non-default loop would stop the global waitpid
140 watcher (Denis Bilenko).
141 - queueing pending watchers of higher priority from a watcher now invokes
142 them in a timely fashion (reported by Denis Bilenko).
143 - add throw() to all libev functions that cannot throw exceptions, for
144 further code size decrease when compiling for C++.
145 - add throw () to callbacks that must not throw exceptions (allocator,
146 syserr, loop acquire/release, periodic reschedule cbs).
9 - fix event_base_loop return code, add event_get_callback, event_base_new, 147 - fix event_base_loop return code, add event_get_callback, event_base_new,
10 event_base_get_method calls to improve libevent 1.x emulation and add 148 event_base_get_method calls to improve libevent 1.x emulation and add
11 some libevent 2.x funcitonality (based on a patch by Jeff Davey). 149 some libevent 2.x functionality (based on a patch by Jeff Davey).
150 - add more memory fences to fix a bug reported by Jeff Davey. Better
151 be overfenced than underprotected.
12 - ev_run now returns a boolean status (true meaning watchers are 152 - ev_run now returns a boolean status (true meaning watchers are
13 still active). 153 still active).
14 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with 154 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
15 libev's EV_ERROR (reported by 191919). 155 libev's EV_ERROR (reported by 191919).
16 - (ecb) add memory fence support for xlC (Darin McBride). 156 - (ecb) add memory fence support for xlC (Darin McBride).
17 - (ecb) add memory fence support for gcc-mips (Anton Kirilov). 157 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
158 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
18 - work around some kernels losing file descriptors by leaking 159 - work around some kernels losing file descriptors by leaking
19 the kqueue descriptor in the child. 160 the kqueue descriptor in the child.
161 - work around linux inotify not reporting IN_ATTRIB changes for directories
162 in many cases.
163 - include sys/syscall.h instead of plain syscall.h.
164 - check for io watcher loops in ev_verify, check for the most
165 common reported usage bug in ev_io_start.
166 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
167 - always use WSASend/WSARecv directly on windows, hoping that this
168 works in all cases (unlike read/write/send/recv...).
169 - try to detect signals around a fork faster (test program by
170 Denis Bilenko).
171 - work around recent glibc versions that leak memory in realloc.
172 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
173 the watcher base set (loop) method.
174 - rewrite the async/signal pipe logic to always keep a valid fd, which
175 simplifies (and hopefully correctifies :) the race checking
176 on fork, at the cost of one extra fd.
177 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
178 inotify-supporting filesystems.
179 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
180 versions ignore it before
181 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
182 - add some untested android support.
183 - enum expressions must be of type int (reported by Juan Pablo L).
20 184
214.11 Sat Feb 4 19:52:39 CET 2012 1854.11 Sat Feb 4 19:52:39 CET 2012
22 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as 186 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
23 was documented already, but not implemented in the repeating case. 187 was documented already, but not implemented in the repeating case.
24 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS. 188 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
… …
208 reinitialising the default loop actually works (patch by 372 reinitialising the default loop actually works (patch by
209 Malek Hadj-Ali). 373 Malek Hadj-Ali).
210 - implement ev_suspend and ev_resume. 374 - implement ev_suspend and ev_resume.
211 - new EV_CUSTOM revents flag for use by applications. 375 - new EV_CUSTOM revents flag for use by applications.
212 - add documentation section about priorities. 376 - add documentation section about priorities.
213 - add a glossary to the dcoumentation. 377 - add a glossary to the documentation.
214 - extend the ev_fork description slightly. 378 - extend the ev_fork description slightly.
215 - optimize a jump out of call_pending. 379 - optimize a jump out of call_pending.
216 380
2173.53 Sun Feb 15 02:38:20 CET 2009 3813.53 Sun Feb 15 02:38:20 CET 2009
218 - fix a bug in event pipe creation on win32 that would cause a 382 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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