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

Comparing libev/Changes (file contents):
Revision 1.285 by root, Thu Feb 28 02:17:20 2013 UTC vs.
Revision 1.363 by root, Wed Jan 22 14:14:41 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
7TODO: ev_break for immediate exit (EVBREAK_NOW?)
8TODO: ev_feed_child_event
9 5
10TODO: document the special problem of signals around fork. 6 - the 4.31 timerfd code wrongly changes the priority of the signal
11TODO: store pid for each signal 7 fd watcher, which is usually harmless unless signal fds are
12TODO: document file descriptor usage per loop 8 also used (found via cpan tester service).
13TODO: store loop pid_t and compare isndie signal handler,store 1 for same, 2 for differign pid, clean up in loop_fork 9 - the documentation wrongly claimed that user may modify fd and events
14TODO: embed watchers need updating when fd changes 10 members in io watchers when the wqatcher was stopped
15TODO: document portbaility requirements for atomic pointer access 11 (found by b_jonas).
12 - new ev_io_modify mutator which changes only the events member,
13 which can be faster. also added ev::io::set (int events) method
14 to ev++.h.
15 - officially allow a zero events mask for io watchers. this should
16 work with older libev versions as well but was not officially
17 allowed before.
18 - do not wake up every minute when timerfd is used to detect timejumps.
19 - do not wake up every minute when periodics are disabled and we have
20 a monotonic clock.
21 - support a lot more "uncommon" compile time configurations,
22 such as ev_embed enabled but ev_timer disabled.
23 - use a start/stop wrapper class to reduce code duplication in
24 ev++.h and make it needlessly more c++-y.
25 - the linux aio backend is no longer compiled in by default.
26 - update to libecb version 0x00010008.
27
284.31 Fri Dec 20 21:58:29 CET 2019
29 - handle backends with minimum wait time a bit better by not
30 waiting in the presence of already-expired timers
31 (behaviour reported by Felipe Gasper).
32 - new feature: use timerfd to detect timejumps quickly,
33 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
34 - document EV_USE_SIGNALFD feature macro.
35
364.30 (EV only)
37 - change non-autoconf test for __kernel_rwf_t by testing
38 LINUX_VERSION_CODE, the most direct test I could find.
39 - fix a bug in the io_uring backend that polled the wrong
40 backend fd, causing it to not work in many cases.
41
424.29 (EV only)
43 - add io uring autoconf and non-autoconf detection.
44 - disable io_uring when some header files are too old.
45
464.28 (EV only)
47 - linuxaio backend resulted in random memory corruption
48 when loop is forked.
49 - linuxaio backend might have tried to cancel an iocb
50 multiple times (was unable to trigger this).
51 - linuxaio backend now employs a generation counter to
52 avoid handling spurious events from cancelled requests.
53 - io_cancel can return EINTR, deal with it. also, assume
54 io_submit also returns EINTR.
55 - fix some other minor bugs in linuxaio backend.
56 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
57 - cleanup: replace expect_true/false and noinline by their
58 libecb counterparts.
59 - move syscall infrastructure from ev_linuxaio.c to ev.c.
60 - prepare io_uring integration.
61 - tweak ev_floor.
62 - epoll, poll, win32 Sleep and other places that use millisecond
63 reslution now all try to round up times.
64 - solaris port backend didn't compile.
65 - abstract time constants into their macros, for more flexibility.
66
674.27 Thu Jun 27 22:43:44 CEST 2019
68 - linux aio backend almost completely rewritten to work around its
69 limitations.
70 - linux aio backend now requires linux 4.19+.
71 - epoll backend now mandatory for linux aio backend.
72 - fail assertions more aggressively on invalid fd's detected
73 in the event loop, do not just silently fd_kill in case of
74 user error.
75 - ev_io_start/ev_io_stop now verify the watcher fd using
76 a syscall when EV_VERIFY is 2 or higher.
77
784.26 (EV only)
79 - update to libecb 0x00010006.
80 - new experimental linux aio backend (linux 4.18+).
81 - removed redundant 0-ptr check in ev_once.
82 - updated/extended ev_set_allocator documentation.
83 - replaced EMPTY2 macro by array_needsize_noinit.
84 - minor code cleanups.
85 - epoll backend now uses epoll_create1 also after fork.
86
874.25 Fri Dec 21 07:49:20 CET 2018
88 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
89 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
90 - move the darwin select workaround higher in ev.c, as newer versions of
91 darwin managed to break their broken select even more.
92 - ANDROID => __ANDROID__ (reported by enh@google.com).
93 - disable epoll_create1 on android because it has broken header files
94 and google is unwilling to fix them (reported by enh@google.com).
95 - avoid a minor compilation warning on win32.
96 - c++: remove deprecated dynamic throw() specifications.
97 - c++: improve the (unsupported) bad_loop exception class.
98 - backport perl ev_periodic example to C, untested.
99 - update libecb, biggets change is to include a memory fence
100 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
101 - minor autoconf/automake modernisation.
102
1034.24 Wed Dec 28 05:19:55 CET 2016
104 - bump version to 4.24, as the release tarball inexplicably
105 didn't have the right version in ev.h, even though the cvs-tagged
106 version did have the right one (reported by Ales Teska).
107
1084.23 Wed Nov 16 18:23:41 CET 2016
109 - move some declarations at the beginning to help certain retarded
110 microsoft compilers, even though their documentation claims
111 otherwise (reported by Ruslan Osmanov).
112
1134.22 Sun Dec 20 22:11:50 CET 2015
114 - when epoll detects unremovable fds in the fd set, rebuild
115 only the epoll descriptor, not the signal pipe, to avoid
116 SIGPIPE in ev_async_send. This doesn't solve it on fork,
117 so document what needs to be done in ev_loop_fork
118 (analyzed by Benjamin Mahler).
119 - remove superfluous sys/timeb.h include on win32
120 (analyzed by Jason Madden).
121 - updated libecb.
122
1234.20 Sat Jun 20 13:01:43 CEST 2015
124 - prefer noexcept over throw () with C++ 11.
125 - update ecb.h due to incompatibilities with c11.
126 - fix a potential aliasing issue when reading and writing
127 watcher callbacks.
128
1294.19 Thu Sep 25 08:18:25 CEST 2014
130 - ev.h wasn't valid C++ anymore, which tripped compilers other than
131 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
132 C++ doesn't support typedefs for function pointers fully, so the affected
133 declarations have to spell out the types each time.
134 - when not using autoconf, tighten the check for clock_gettime and related
135 functionality.
136
1374.18 Fri Sep 5 17:55:26 CEST 2014
138 - events on files were not always generated properly with the
139 epoll backend (testcase by Assaf Inbal).
140 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
141 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
142 - use a reasonable fallback for EV_NSIG instead of erroring out
143 when we can't detect the signal set size.
144 - in the absence of autoconf, do not use the clock syscall
145 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
146 doing clock_gettime in userspace).
147 - ensure extern "C" function pointers are used for externally-visible
148 loop callbacks (not watcher callbacks yet).
149 - (ecb) work around memory barriers and volatile apparently both being
150 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
151
1524.15 Fri Mar 1 12:04:50 CET 2013
16 - destroying a non-default loop would stop the global waitpid 153 - destroying a non-default loop would stop the global waitpid
17 watcher (Denis Bilenko). 154 watcher (Denis Bilenko).
18 - queueing pending watchers of higher priority from a watcher now invokes 155 - queueing pending watchers of higher priority from a watcher now invokes
19 them in a timely fashion (reported by Denis Bilenko). 156 them in a timely fashion (reported by Denis Bilenko).
20 - add throw() to all libev functions that cannot throw exceptions, for 157 - add throw() to all libev functions that cannot throw exceptions, for
… …
38 - work around linux inotify not reporting IN_ATTRIB changes for directories 175 - work around linux inotify not reporting IN_ATTRIB changes for directories
39 in many cases. 176 in many cases.
40 - include sys/syscall.h instead of plain syscall.h. 177 - include sys/syscall.h instead of plain syscall.h.
41 - check for io watcher loops in ev_verify, check for the most 178 - check for io watcher loops in ev_verify, check for the most
42 common reported usage bug in ev_io_start. 179 common reported usage bug in ev_io_start.
43 - chose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET. 180 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
44 - always use WSASend/WSARecv directly on windows, hoping that this 181 - always use WSASend/WSARecv directly on windows, hoping that this
45 works in all cases (unlike read/write/send/recv...). 182 works in all cases (unlike read/write/send/recv...).
46 - try to detect signals around a fork faster (test program by 183 - try to detect signals around a fork faster (test program by
47 Denis Bilenko). 184 Denis Bilenko).
48 - work around recent glibc versions that leak memory in realloc. 185 - work around recent glibc versions that leak memory in realloc.
49 - rename ev::embed::set to ev::embed::set_embed to avoid clashing 186 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
50 the watcher base set (loop) method. 187 the watcher base set (loop) method.
51 - rewrite the async/signal pipe logic to always keep a valid fd, which 188 - rewrite the async/signal pipe logic to always keep a valid fd, which
52 simplifies (and hopefuly correctifies :) the race checking 189 simplifies (and hopefully correctifies :) the race checking
53 on fork, at the cost of one extra fd. 190 on fork, at the cost of one extra fd.
54 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of 191 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
55 inotify-supporting filesystems. 192 inotify-supporting filesystems.
56 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf 193 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
57 versions ignore it before 194 versions ignore it before
… …
249 reinitialising the default loop actually works (patch by 386 reinitialising the default loop actually works (patch by
250 Malek Hadj-Ali). 387 Malek Hadj-Ali).
251 - implement ev_suspend and ev_resume. 388 - implement ev_suspend and ev_resume.
252 - new EV_CUSTOM revents flag for use by applications. 389 - new EV_CUSTOM revents flag for use by applications.
253 - add documentation section about priorities. 390 - add documentation section about priorities.
254 - add a glossary to the dcoumentation. 391 - add a glossary to the documentation.
255 - extend the ev_fork description slightly. 392 - extend the ev_fork description slightly.
256 - optimize a jump out of call_pending. 393 - optimize a jump out of call_pending.
257 394
2583.53 Sun Feb 15 02:38:20 CET 2009 3953.53 Sun Feb 15 02:38:20 CET 2009
259 - fix a bug in event pipe creation on win32 that would cause a 396 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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