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

Comparing libev/Changes (file contents):
Revision 1.287 by root, Fri Mar 1 11:13:22 2013 UTC vs.
Revision 1.358 by root, Wed Jan 22 02:00:13 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: 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: faq, process a thing in each iteration 5TODO: document EV_TSTAMP_T
6TODO: dbeugging tips, ev_verify, ev_init twice 6TODO: for next ABI/API change, consider moving EV__IOFDSSET into io->fd instead and provide a getter.
7TODO: ev_break for immediate exit (EVBREAK_NOW?) 7
8TODO: ev_feed_child_event 8TODO: maybe allow events = 0 for io watchers
9TODO: document the special problem of signals around fork. 9 - the 4.31 timerfd code wrongly changes the priority of the signal
10TODO: store pid for each signal 10 fd watcher, which is usually harmless unless signal fds are
11TODO: document file descriptor usage per loop 11 also used (found via cpan tester service).
12TODO: store loop pid_t and compare isndie signal handler,store 1 for same, 2 for differign pid, clean up in loop_fork 12 - the documentation wrongly claimed that user may modify fd and events
13TODO: embed watchers need updating when fd changes 13 members in io watchers when the wqatcher was stopped
14TODO: document portbaility requirements for atomic pointer access 14 (found by b_jonas).
15 - new ev_io_modify mutator which changes only the events member,
16 which cna be faster.
17 - do not wake up every minute when timerfd is used to detect timejumps.
18 - support a lot more "uncommon" compile time configurations,
19 such as ev_embed enabled but ev_timer disabled.
20 - the linux aio backend is no longer compiled in by default.
21 - update to libecb version 0x00010008.
22
234.31 Fri Dec 20 21:58:29 CET 2019
24 - handle backends with minimum wait time a bit better by not
25 waiting in the presence of already-expired timers
26 (behaviour reported by Felipe Gasper).
27 - new feature: use timerfd to detect timejumps quickly,
28 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
29 - document EV_USE_SIGNALFD feature macro.
30
314.30 (EV only)
32 - change non-autoconf test for __kernel_rwf_t by testing
33 LINUX_VERSION_CODE, the most direct test I could find.
34 - fix a bug in the io_uring backend that polled the wrong
35 backend fd, causing it to not work in many cases.
36
374.29 (EV only)
38 - add io uring autoconf and non-autoconf detection.
39 - disable io_uring when some header files are too old.
40
414.28 (EV only)
42 - linuxaio backend resulted in random memory corruption
43 when loop is forked.
44 - linuxaio backend might have tried to cancel an iocb
45 multiple times (was unable to trigger this).
46 - linuxaio backend now employs a generation counter to
47 avoid handling spurious events from cancelled requests.
48 - io_cancel can return EINTR, deal with it. also, assume
49 io_submit also returns EINTR.
50 - fix some other minor bugs in linuxaio backend.
51 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
52 - cleanup: replace expect_true/false and noinline by their
53 libecb counterparts.
54 - move syscall infrastructure from ev_linuxaio.c to ev.c.
55 - prepare io_uring integration.
56 - tweak ev_floor.
57 - epoll, poll, win32 Sleep and other places that use millisecond
58 reslution now all try to round up times.
59 - solaris port backend didn't compile.
60 - abstract time constants into their macros, for more flexibility.
61
624.27 Thu Jun 27 22:43:44 CEST 2019
63 - linux aio backend almost completely rewritten to work around its
64 limitations.
65 - linux aio backend now requires linux 4.19+.
66 - epoll backend now mandatory for linux aio backend.
67 - fail assertions more aggressively on invalid fd's detected
68 in the event loop, do not just silently fd_kill in case of
69 user error.
70 - ev_io_start/ev_io_stop now verify the watcher fd using
71 a syscall when EV_VERIFY is 2 or higher.
72
734.26 (EV only)
74 - update to libecb 0x00010006.
75 - new experimental linux aio backend (linux 4.18+).
76 - removed redundant 0-ptr check in ev_once.
77 - updated/extended ev_set_allocator documentation.
78 - replaced EMPTY2 macro by array_needsize_noinit.
79 - minor code cleanups.
80 - epoll backend now uses epoll_create1 also after fork.
81
824.25 Fri Dec 21 07:49:20 CET 2018
83 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
84 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
85 - move the darwin select workaround higher in ev.c, as newer versions of
86 darwin managed to break their broken select even more.
87 - ANDROID => __ANDROID__ (reported by enh@google.com).
88 - disable epoll_create1 on android because it has broken header files
89 and google is unwilling to fix them (reported by enh@google.com).
90 - avoid a minor compilation warning on win32.
91 - c++: remove deprecated dynamic throw() specifications.
92 - c++: improve the (unsupported) bad_loop exception class.
93 - backport perl ev_periodic example to C, untested.
94 - update libecb, biggets change is to include a memory fence
95 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
96 - minor autoconf/automake modernisation.
97
984.24 Wed Dec 28 05:19:55 CET 2016
99 - bump version to 4.24, as the release tarball inexplicably
100 didn't have the right version in ev.h, even though the cvs-tagged
101 version did have the right one (reported by Ales Teska).
102
1034.23 Wed Nov 16 18:23:41 CET 2016
104 - move some declarations at the beginning to help certain retarded
105 microsoft compilers, even though their documentation claims
106 otherwise (reported by Ruslan Osmanov).
107
1084.22 Sun Dec 20 22:11:50 CET 2015
109 - when epoll detects unremovable fds in the fd set, rebuild
110 only the epoll descriptor, not the signal pipe, to avoid
111 SIGPIPE in ev_async_send. This doesn't solve it on fork,
112 so document what needs to be done in ev_loop_fork
113 (analyzed by Benjamin Mahler).
114 - remove superfluous sys/timeb.h include on win32
115 (analyzed by Jason Madden).
116 - updated libecb.
117
1184.20 Sat Jun 20 13:01:43 CEST 2015
119 - prefer noexcept over throw () with C++ 11.
120 - update ecb.h due to incompatibilities with c11.
121 - fix a potential aliasing issue when reading and writing
122 watcher callbacks.
123
1244.19 Thu Sep 25 08:18:25 CEST 2014
125 - ev.h wasn't valid C++ anymore, which tripped compilers other than
126 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
127 C++ doesn't support typedefs for function pointers fully, so the affected
128 declarations have to spell out the types each time.
129 - when not using autoconf, tighten the check for clock_gettime and related
130 functionality.
131
1324.18 Fri Sep 5 17:55:26 CEST 2014
133 - events on files were not always generated properly with the
134 epoll backend (testcase by Assaf Inbal).
135 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
136 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
137 - use a reasonable fallback for EV_NSIG instead of erroring out
138 when we can't detect the signal set size.
139 - in the absence of autoconf, do not use the clock syscall
140 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
141 doing clock_gettime in userspace).
142 - ensure extern "C" function pointers are used for externally-visible
143 loop callbacks (not watcher callbacks yet).
144 - (ecb) work around memory barriers and volatile apparently both being
145 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
15 146
164.15 Fri Mar 1 12:04:50 CET 2013 1474.15 Fri Mar 1 12:04:50 CET 2013
17 - destroying a non-default loop would stop the global waitpid 148 - destroying a non-default loop would stop the global waitpid
18 watcher (Denis Bilenko). 149 watcher (Denis Bilenko).
19 - queueing pending watchers of higher priority from a watcher now invokes 150 - queueing pending watchers of higher priority from a watcher now invokes

Diff Legend

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