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

Comparing libev/Changes (file contents):
Revision 1.270 by root, Sun May 6 19:47:38 2012 UTC vs.
Revision 1.368 by root, Thu Jul 30 00:59:36 2020 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: for next ABI/API change, consider moving EV__IOFDSSET into io->fd instead and provide a getter.
4TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone) 4TODO: document EV_TSTAMP_T
5TODO: faq, process a thing in each iteration
6TODO: dbeugging tips, ev_verify, ev_init twice
7 5
6 - due to a logic bug, timerfd availability did not always translate
7 into fewer wakeups - libev is now more aggressive in using
8 longer sleeps, even when timerfd is not in use.
9
104.33 Wed Mar 18 13:22:29 CET 2020
11 - no changes w.r.t. 4.32.
12
134.32 (EV only)
14 - the 4.31 timerfd code wrongly changed the priority of the signal
15 fd watcher, which is usually harmless unless signal fds are
16 also used (found via cpan tester service).
17 - the documentation wrongly claimed that user may modify fd and events
18 members in io watchers when the watcher was stopped
19 (found by b_jonas).
20 - new ev_io_modify mutator which changes only the events member,
21 which can be faster. also added ev::io::set (int events) method
22 to ev++.h.
23 - officially allow a zero events mask for io watchers. this should
24 work with older libev versions as well but was not officially
25 allowed before.
26 - do not wake up every minute when timerfd is used to detect timejumps.
27 - do not wake up every minute when periodics are disabled and we have
28 a monotonic clock.
29 - support a lot more "uncommon" compile time configurations,
30 such as ev_embed enabled but ev_timer disabled.
31 - use a start/stop wrapper class to reduce code duplication in
32 ev++.h and make it needlessly more c++-y.
33 - the linux aio backend is no longer compiled in by default.
34 - update to libecb version 0x00010008.
35
364.31 Fri Dec 20 21:58:29 CET 2019
37 - handle backends with minimum wait time a bit better by not
38 waiting in the presence of already-expired timers
39 (behaviour reported by Felipe Gasper).
40 - new feature: use timerfd to detect timejumps quickly,
41 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
42 - document EV_USE_SIGNALFD feature macro.
43
444.30 (EV only)
45 - change non-autoconf test for __kernel_rwf_t by testing
46 LINUX_VERSION_CODE, the most direct test I could find.
47 - fix a bug in the io_uring backend that polled the wrong
48 backend fd, causing it to not work in many cases.
49
504.29 (EV only)
51 - add io uring autoconf and non-autoconf detection.
52 - disable io_uring when some header files are too old.
53
544.28 (EV only)
55 - linuxaio backend resulted in random memory corruption
56 when loop is forked.
57 - linuxaio backend might have tried to cancel an iocb
58 multiple times (was unable to trigger this).
59 - linuxaio backend now employs a generation counter to
60 avoid handling spurious events from cancelled requests.
61 - io_cancel can return EINTR, deal with it. also, assume
62 io_submit also returns EINTR.
63 - fix some other minor bugs in linuxaio backend.
64 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
65 - cleanup: replace expect_true/false and noinline by their
66 libecb counterparts.
67 - move syscall infrastructure from ev_linuxaio.c to ev.c.
68 - prepare io_uring integration.
69 - tweak ev_floor.
70 - epoll, poll, win32 Sleep and other places that use millisecond
71 reslution now all try to round up times.
72 - solaris port backend didn't compile.
73 - abstract time constants into their macros, for more flexibility.
74
754.27 Thu Jun 27 22:43:44 CEST 2019
76 - linux aio backend almost completely rewritten to work around its
77 limitations.
78 - linux aio backend now requires linux 4.19+.
79 - epoll backend now mandatory for linux aio backend.
80 - fail assertions more aggressively on invalid fd's detected
81 in the event loop, do not just silently fd_kill in case of
82 user error.
83 - ev_io_start/ev_io_stop now verify the watcher fd using
84 a syscall when EV_VERIFY is 2 or higher.
85
864.26 (EV only)
87 - update to libecb 0x00010006.
88 - new experimental linux aio backend (linux 4.18+).
89 - removed redundant 0-ptr check in ev_once.
90 - updated/extended ev_set_allocator documentation.
91 - replaced EMPTY2 macro by array_needsize_noinit.
92 - minor code cleanups.
93 - epoll backend now uses epoll_create1 also after fork.
94
954.25 Fri Dec 21 07:49:20 CET 2018
96 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
97 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
98 - move the darwin select workaround higher in ev.c, as newer versions of
99 darwin managed to break their broken select even more.
100 - ANDROID => __ANDROID__ (reported by enh@google.com).
101 - disable epoll_create1 on android because it has broken header files
102 and google is unwilling to fix them (reported by enh@google.com).
103 - avoid a minor compilation warning on win32.
104 - c++: remove deprecated dynamic throw() specifications.
105 - c++: improve the (unsupported) bad_loop exception class.
106 - backport perl ev_periodic example to C, untested.
107 - update libecb, biggets change is to include a memory fence
108 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
109 - minor autoconf/automake modernisation.
110
1114.24 Wed Dec 28 05:19:55 CET 2016
112 - bump version to 4.24, as the release tarball inexplicably
113 didn't have the right version in ev.h, even though the cvs-tagged
114 version did have the right one (reported by Ales Teska).
115
1164.23 Wed Nov 16 18:23:41 CET 2016
117 - move some declarations at the beginning to help certain retarded
118 microsoft compilers, even though their documentation claims
119 otherwise (reported by Ruslan Osmanov).
120
1214.22 Sun Dec 20 22:11:50 CET 2015
122 - when epoll detects unremovable fds in the fd set, rebuild
123 only the epoll descriptor, not the signal pipe, to avoid
124 SIGPIPE in ev_async_send. This doesn't solve it on fork,
125 so document what needs to be done in ev_loop_fork
126 (analyzed by Benjamin Mahler).
127 - remove superfluous sys/timeb.h include on win32
128 (analyzed by Jason Madden).
129 - updated libecb.
130
1314.20 Sat Jun 20 13:01:43 CEST 2015
132 - prefer noexcept over throw () with C++ 11.
133 - update ecb.h due to incompatibilities with c11.
134 - fix a potential aliasing issue when reading and writing
135 watcher callbacks.
136
1374.19 Thu Sep 25 08:18:25 CEST 2014
138 - ev.h wasn't valid C++ anymore, which tripped compilers other than
139 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
140 C++ doesn't support typedefs for function pointers fully, so the affected
141 declarations have to spell out the types each time.
142 - when not using autoconf, tighten the check for clock_gettime and related
143 functionality.
144
1454.18 Fri Sep 5 17:55:26 CEST 2014
146 - events on files were not always generated properly with the
147 epoll backend (testcase by Assaf Inbal).
148 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
149 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
150 - use a reasonable fallback for EV_NSIG instead of erroring out
151 when we can't detect the signal set size.
152 - in the absence of autoconf, do not use the clock syscall
153 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
154 doing clock_gettime in userspace).
155 - ensure extern "C" function pointers are used for externally-visible
156 loop callbacks (not watcher callbacks yet).
157 - (ecb) work around memory barriers and volatile apparently both being
158 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
159
1604.15 Fri Mar 1 12:04:50 CET 2013
161 - destroying a non-default loop would stop the global waitpid
162 watcher (Denis Bilenko).
8 - queueing pending watchers of higher priority from a watcher now invokes 163 - queueing pending watchers of higher priority from a watcher now invokes
9 them in a timely fashion (reported by Denis Bilenko). 164 them in a timely fashion (reported by Denis Bilenko).
10 - add throw() to all libev functions that cannot throw exceptions, for 165 - add throw() to all libev functions that cannot throw exceptions, for
11 further code size decrease when compiling for C++. 166 further code size decrease when compiling for C++.
12 - add throw () to callbacks that must not throw exceptions (allocator, 167 - add throw () to callbacks that must not throw exceptions (allocator,
13 syserr, loop acquire/release, periodic reschedule cbs). 168 syserr, loop acquire/release, periodic reschedule cbs).
14 - fix event_base_loop return code, add event_get_callback, event_base_new, 169 - fix event_base_loop return code, add event_get_callback, event_base_new,
15 event_base_get_method calls to improve libevent 1.x emulation and add 170 event_base_get_method calls to improve libevent 1.x emulation and add
16 some libevent 2.x funcitonality (based on a patch by Jeff Davey). 171 some libevent 2.x functionality (based on a patch by Jeff Davey).
17 - add more memory fences to fix a bug reported by Jeff Davey. Better 172 - add more memory fences to fix a bug reported by Jeff Davey. Better
18 be overfenced than underprotected. 173 be overfenced than underprotected.
19 - ev_run now returns a boolean status (true meaning watchers are 174 - ev_run now returns a boolean status (true meaning watchers are
20 still active). 175 still active).
21 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with 176 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
… …
23 - (ecb) add memory fence support for xlC (Darin McBride). 178 - (ecb) add memory fence support for xlC (Darin McBride).
24 - (ecb) add memory fence support for gcc-mips (Anton Kirilov). 179 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
25 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber). 180 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
26 - work around some kernels losing file descriptors by leaking 181 - work around some kernels losing file descriptors by leaking
27 the kqueue descriptor in the child. 182 the kqueue descriptor in the child.
183 - work around linux inotify not reporting IN_ATTRIB changes for directories
184 in many cases.
28 - include sys/syscall.h instead of plain syscall.h. 185 - include sys/syscall.h instead of plain syscall.h.
29 - check for io watcher loops in ev_verify, check for the most 186 - check for io watcher loops in ev_verify, check for the most
30 common reported usage bug in ev_io_start. 187 common reported usage bug in ev_io_start.
31 - chose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET. 188 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
32 - always use WSASend/WSARecv directly on windows, hoping that this 189 - always use WSASend/WSARecv directly on windows, hoping that this
33 works in all cases (unlike read/write/send/recv...). 190 works in all cases (unlike read/write/send/recv...).
191 - try to detect signals around a fork faster (test program by
192 Denis Bilenko).
193 - work around recent glibc versions that leak memory in realloc.
194 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
195 the watcher base set (loop) method.
196 - rewrite the async/signal pipe logic to always keep a valid fd, which
197 simplifies (and hopefully correctifies :) the race checking
198 on fork, at the cost of one extra fd.
199 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
200 inotify-supporting filesystems.
201 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
202 versions ignore it before
203 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
204 - add some untested android support.
205 - enum expressions must be of type int (reported by Juan Pablo L).
34 206
354.11 Sat Feb 4 19:52:39 CET 2012 2074.11 Sat Feb 4 19:52:39 CET 2012
36 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as 208 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
37 was documented already, but not implemented in the repeating case. 209 was documented already, but not implemented in the repeating case.
38 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS. 210 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
… …
222 reinitialising the default loop actually works (patch by 394 reinitialising the default loop actually works (patch by
223 Malek Hadj-Ali). 395 Malek Hadj-Ali).
224 - implement ev_suspend and ev_resume. 396 - implement ev_suspend and ev_resume.
225 - new EV_CUSTOM revents flag for use by applications. 397 - new EV_CUSTOM revents flag for use by applications.
226 - add documentation section about priorities. 398 - add documentation section about priorities.
227 - add a glossary to the dcoumentation. 399 - add a glossary to the documentation.
228 - extend the ev_fork description slightly. 400 - extend the ev_fork description slightly.
229 - optimize a jump out of call_pending. 401 - optimize a jump out of call_pending.
230 402
2313.53 Sun Feb 15 02:38:20 CET 2009 4033.53 Sun Feb 15 02:38:20 CET 2009
232 - fix a bug in event pipe creation on win32 that would cause a 404 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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