| 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: for next ABI/API change, consider moving EV__IOFDSSET into io->fd instead and provide a getter. |
| 4 | TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone) |
4 | TODO: document EV_TSTAMP_T |
| 5 | TODO: faq, process a thing in each iteration |
5 | TODO: https://paste.debian.net/1268417/, stdfun |
| 6 | TODO: dbeugging tips, ev_verify, ev_init twice |
6 | |
| 7 | TODO: ev_break for immediate exit (EVBREAK_NOW?) |
7 | - libev did not compile on systems lacking clock_gettime. |
| 8 | TODO: ev_feed_child_event |
8 | - due to a logic bug, timerfd availability did not always translate |
| 9 | TODO: document the special problem of signals around fork. |
9 | into fewer wakeups - libev is now more aggressive in using |
| 10 | TODO: store pid for each signal |
10 | longer sleeps, even when timerfd is not in use. |
| 11 | TODO: document file descriptor usage per loop |
11 | |
| 12 | TODO: store loop pid_t and compare isndie signal handler,store 1 for same, 2 for differign pid, clean up in loop_fork |
12 | 4.33 Wed Mar 18 13:22:29 CET 2020 |
| 13 | TODO: embed watchers need updating when fd changes |
13 | - no changes w.r.t. 4.32. |
| 14 | TODO: document portbaility requirements for atomic pointer access |
14 | |
|
|
15 | 4.32 (EV only) |
|
|
16 | - the 4.31 timerfd code wrongly changed the priority of the signal |
|
|
17 | fd watcher, which is usually harmless unless signal fds are |
|
|
18 | also used (found via cpan tester service). |
|
|
19 | - the documentation wrongly claimed that user may modify fd and events |
|
|
20 | members in io watchers when the watcher was stopped |
|
|
21 | (found by b_jonas). |
|
|
22 | - new ev_io_modify mutator which changes only the events member, |
|
|
23 | which can be faster. also added ev::io::set (int events) method |
|
|
24 | to ev++.h. |
|
|
25 | - officially allow a zero events mask for io watchers. this should |
|
|
26 | work with older libev versions as well but was not officially |
|
|
27 | allowed before. |
|
|
28 | - do not wake up every minute when timerfd is used to detect timejumps. |
|
|
29 | - do not wake up every minute when periodics are disabled and we have |
|
|
30 | a monotonic clock. |
|
|
31 | - support a lot more "uncommon" compile time configurations, |
|
|
32 | such as ev_embed enabled but ev_timer disabled. |
|
|
33 | - use a start/stop wrapper class to reduce code duplication in |
|
|
34 | ev++.h and make it needlessly more c++-y. |
|
|
35 | - the linux aio backend is no longer compiled in by default. |
|
|
36 | - update to libecb version 0x00010008. |
|
|
37 | |
|
|
38 | 4.31 Fri Dec 20 21:58:29 CET 2019 |
|
|
39 | - handle backends with minimum wait time a bit better by not |
|
|
40 | waiting in the presence of already-expired timers |
|
|
41 | (behaviour reported by Felipe Gasper). |
|
|
42 | - new feature: use timerfd to detect timejumps quickly, |
|
|
43 | can be disabled with the new EVFLAG_NOTIMERFD loop flag. |
|
|
44 | - document EV_USE_SIGNALFD feature macro. |
|
|
45 | |
|
|
46 | 4.30 (EV only) |
|
|
47 | - change non-autoconf test for __kernel_rwf_t by testing |
|
|
48 | LINUX_VERSION_CODE, the most direct test I could find. |
|
|
49 | - fix a bug in the io_uring backend that polled the wrong |
|
|
50 | backend fd, causing it to not work in many cases. |
|
|
51 | |
|
|
52 | 4.29 (EV only) |
|
|
53 | - add io uring autoconf and non-autoconf detection. |
|
|
54 | - disable io_uring when some header files are too old. |
|
|
55 | |
|
|
56 | 4.28 (EV only) |
|
|
57 | - linuxaio backend resulted in random memory corruption |
|
|
58 | when loop is forked. |
|
|
59 | - linuxaio backend might have tried to cancel an iocb |
|
|
60 | multiple times (was unable to trigger this). |
|
|
61 | - linuxaio backend now employs a generation counter to |
|
|
62 | avoid handling spurious events from cancelled requests. |
|
|
63 | - io_cancel can return EINTR, deal with it. also, assume |
|
|
64 | io_submit also returns EINTR. |
|
|
65 | - fix some other minor bugs in linuxaio backend. |
|
|
66 | - ev_tstamp type can now be overriden by defining EV_TSTAMP_T. |
|
|
67 | - cleanup: replace expect_true/false and noinline by their |
|
|
68 | libecb counterparts. |
|
|
69 | - move syscall infrastructure from ev_linuxaio.c to ev.c. |
|
|
70 | - prepare io_uring integration. |
|
|
71 | - tweak ev_floor. |
|
|
72 | - epoll, poll, win32 Sleep and other places that use millisecond |
|
|
73 | reslution now all try to round up times. |
|
|
74 | - solaris port backend didn't compile. |
|
|
75 | - abstract time constants into their macros, for more flexibility. |
|
|
76 | |
|
|
77 | 4.27 Thu Jun 27 22:43:44 CEST 2019 |
|
|
78 | - linux aio backend almost completely rewritten to work around its |
|
|
79 | limitations. |
|
|
80 | - linux aio backend now requires linux 4.19+. |
|
|
81 | - epoll backend now mandatory for linux aio backend. |
|
|
82 | - fail assertions more aggressively on invalid fd's detected |
|
|
83 | in the event loop, do not just silently fd_kill in case of |
|
|
84 | user error. |
|
|
85 | - ev_io_start/ev_io_stop now verify the watcher fd using |
|
|
86 | a syscall when EV_VERIFY is 2 or higher. |
|
|
87 | |
|
|
88 | 4.26 (EV only) |
|
|
89 | - update to libecb 0x00010006. |
|
|
90 | - new experimental linux aio backend (linux 4.18+). |
|
|
91 | - removed redundant 0-ptr check in ev_once. |
|
|
92 | - updated/extended ev_set_allocator documentation. |
|
|
93 | - replaced EMPTY2 macro by array_needsize_noinit. |
|
|
94 | - minor code cleanups. |
|
|
95 | - epoll backend now uses epoll_create1 also after fork. |
|
|
96 | |
|
|
97 | 4.25 Fri Dec 21 07:49:20 CET 2018 |
|
|
98 | - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT |
|
|
99 | (EV_THROW still provided) and now uses noexcept on C++11 or newer. |
|
|
100 | - move the darwin select workaround higher in ev.c, as newer versions of |
|
|
101 | darwin managed to break their broken select even more. |
|
|
102 | - ANDROID => __ANDROID__ (reported by enh@google.com). |
|
|
103 | - disable epoll_create1 on android because it has broken header files |
|
|
104 | and google is unwilling to fix them (reported by enh@google.com). |
|
|
105 | - avoid a minor compilation warning on win32. |
|
|
106 | - c++: remove deprecated dynamic throw() specifications. |
|
|
107 | - c++: improve the (unsupported) bad_loop exception class. |
|
|
108 | - backport perl ev_periodic example to C, untested. |
|
|
109 | - update libecb, biggets change is to include a memory fence |
|
|
110 | in ECB_MEMORY_FENCE_RELEASE on x86/amd64. |
|
|
111 | - minor autoconf/automake modernisation. |
|
|
112 | |
|
|
113 | 4.24 Wed Dec 28 05:19:55 CET 2016 |
|
|
114 | - bump version to 4.24, as the release tarball inexplicably |
|
|
115 | didn't have the right version in ev.h, even though the cvs-tagged |
|
|
116 | version did have the right one (reported by Ales Teska). |
|
|
117 | |
|
|
118 | 4.23 Wed Nov 16 18:23:41 CET 2016 |
|
|
119 | - move some declarations at the beginning to help certain retarded |
|
|
120 | microsoft compilers, even though their documentation claims |
|
|
121 | otherwise (reported by Ruslan Osmanov). |
|
|
122 | |
|
|
123 | 4.22 Sun Dec 20 22:11:50 CET 2015 |
|
|
124 | - when epoll detects unremovable fds in the fd set, rebuild |
|
|
125 | only the epoll descriptor, not the signal pipe, to avoid |
|
|
126 | SIGPIPE in ev_async_send. This doesn't solve it on fork, |
|
|
127 | so document what needs to be done in ev_loop_fork |
|
|
128 | (analyzed by Benjamin Mahler). |
|
|
129 | - remove superfluous sys/timeb.h include on win32 |
|
|
130 | (analyzed by Jason Madden). |
|
|
131 | - updated libecb. |
|
|
132 | |
|
|
133 | 4.20 Sat Jun 20 13:01:43 CEST 2015 |
|
|
134 | - prefer noexcept over throw () with C++ 11. |
|
|
135 | - update ecb.h due to incompatibilities with c11. |
|
|
136 | - fix a potential aliasing issue when reading and writing |
|
|
137 | watcher callbacks. |
|
|
138 | |
|
|
139 | 4.19 Thu Sep 25 08:18:25 CEST 2014 |
|
|
140 | - ev.h wasn't valid C++ anymore, which tripped compilers other than |
|
|
141 | clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately, |
|
|
142 | C++ doesn't support typedefs for function pointers fully, so the affected |
|
|
143 | declarations have to spell out the types each time. |
|
|
144 | - when not using autoconf, tighten the check for clock_gettime and related |
|
|
145 | functionality. |
|
|
146 | |
|
|
147 | 4.18 Fri Sep 5 17:55:26 CEST 2014 |
|
|
148 | - events on files were not always generated properly with the |
|
|
149 | epoll backend (testcase by Assaf Inbal). |
|
|
150 | - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin). |
|
|
151 | - (ecb) support m68k, m88k and sh (patch by Miod Vallat). |
|
|
152 | - use a reasonable fallback for EV_NSIG instead of erroring out |
|
|
153 | when we can't detect the signal set size. |
|
|
154 | - in the absence of autoconf, do not use the clock syscall |
|
|
155 | on glibc >= 2.17 (avoids the syscall AND -lrt on systems |
|
|
156 | doing clock_gettime in userspace). |
|
|
157 | - ensure extern "C" function pointers are used for externally-visible |
|
|
158 | loop callbacks (not watcher callbacks yet). |
|
|
159 | - (ecb) work around memory barriers and volatile apparently both being |
|
|
160 | broken in visual studio 2008 and later (analysed and patch by Nicolas Noble). |
| 15 | |
161 | |
| 16 | 4.15 Fri Mar 1 12:04:50 CET 2013 |
162 | 4.15 Fri Mar 1 12:04:50 CET 2013 |
| 17 | - destroying a non-default loop would stop the global waitpid |
163 | - destroying a non-default loop would stop the global waitpid |
| 18 | watcher (Denis Bilenko). |
164 | watcher (Denis Bilenko). |
| 19 | - queueing pending watchers of higher priority from a watcher now invokes |
165 | - queueing pending watchers of higher priority from a watcher now invokes |
