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

Comparing libev/Changes (file contents):
Revision 1.273 by root, Tue May 29 19:10:39 2012 UTC vs.
Revision 1.347 by root, Fri Nov 22 19:54:38 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 use timerfd to detect time jumps on linux
5TODO: faq, process a thing in each iteration 5TODO: document EV_TSTAMP_T
6TODO: dbeugging tips, ev_verify, ev_init twice
7 6
8TODO: ev_feed_child_event 7 - change non-autoconf test for __kernel_rwf_t by testing
8 LINUX_VERSION_CODE, the most direct test I could find.
9
104.29 (EV only)
11 - add io uring autoconf and non-autoconf detection.
12 - disable io_uring when some header files are too old.
13
144.28 (EV only)
15 - linuxaio backend resulted in random memory corruption
16 when loop is forked.
17 - linuxaio backend might have tried to cancel an iocb
18 multiple times (was unable to trigger this).
19 - linuxaio backend now employs a generation counter to
20 avoid handling spurious events from cancelled requests.
21 - io_cancel can return EINTR, deal with it. also, assume
22 io_submit also returns EINTR.
23 - fix some other minor bugs in linuxaio backend.
24 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
25 - cleanup: replace expect_true/false and noinline by their
26 libecb counterparts.
27 - move syscall infrastructure from ev_linuxaio.c to ev.c.
28 - prepare io_uring integration.
29 - tweak ev_floor.
30 - epoll, poll, win32 Sleep and other places that use millisecond
31 reslution now all try to round up times.
32 - solaris port backend didn't compile.
33 - abstract time constants into their macros, for more flexibility.
34
354.27 Thu Jun 27 22:43:44 CEST 2019
36 - linux aio backend almost completely rewritten to work around its
37 limitations.
38 - linux aio backend now requires linux 4.19+.
39 - epoll backend now mandatory for linux aio backend.
40 - fail assertions more aggressively on invalid fd's detected
41 in the event loop, do not just silently fd_kill in case of
42 user error.
43 - ev_io_start/ev_io_stop now verify the watcher fd using
44 a syscall when EV_VERIFY is 2 or higher.
45
464.26 (EV only)
47 - update to libecb 0x00010006.
48 - new experimental linux aio backend (linux 4.18+).
49 - removed redundant 0-ptr check in ev_once.
50 - updated/extended ev_set_allocator documentation.
51 - replaced EMPTY2 macro by array_needsize_noinit.
52 - minor code cleanups.
53 - epoll backend now uses epoll_create1 also after fork.
54
554.25 Fri Dec 21 07:49:20 CET 2018
56 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
57 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
58 - move the darwin select workaround higher in ev.c, as newer versions of
59 darwin managed to break their broken select even more.
60 - ANDROID => __ANDROID__ (reported by enh@google.com).
61 - disable epoll_create1 on android because it has broken header files
62 and google is unwilling to fix them (reported by enh@google.com).
63 - avoid a minor compilation warning on win32.
64 - c++: remove deprecated dynamic throw() specifications.
65 - c++: improve the (unsupported) bad_loop exception class.
66 - backport perl ev_periodic example to C, untested.
67 - update libecb, biggets change is to include a memory fence
68 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
69 - minor autoconf/automake modernisation.
70
714.24 Wed Dec 28 05:19:55 CET 2016
72 - bump version to 4.24, as the release tarball inexplicably
73 didn't have the right version in ev.h, even though the cvs-tagged
74 version did have the right one (reported by Ales Teska).
75
764.23 Wed Nov 16 18:23:41 CET 2016
77 - move some declarations at the beginning to help certain retarded
78 microsoft compilers, even though their documentation claims
79 otherwise (reported by Ruslan Osmanov).
80
814.22 Sun Dec 20 22:11:50 CET 2015
82 - when epoll detects unremovable fds in the fd set, rebuild
83 only the epoll descriptor, not the signal pipe, to avoid
84 SIGPIPE in ev_async_send. This doesn't solve it on fork,
85 so document what needs to be done in ev_loop_fork
86 (analyzed by Benjamin Mahler).
87 - remove superfluous sys/timeb.h include on win32
88 (analyzed by Jason Madden).
89 - updated libecb.
90
914.20 Sat Jun 20 13:01:43 CEST 2015
92 - prefer noexcept over throw () with C++ 11.
93 - update ecb.h due to incompatibilities with c11.
94 - fix a potential aliasing issue when reading and writing
95 watcher callbacks.
96
974.19 Thu Sep 25 08:18:25 CEST 2014
98 - ev.h wasn't valid C++ anymore, which tripped compilers other than
99 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
100 C++ doesn't support typedefs for function pointers fully, so the affected
101 declarations have to spell out the types each time.
102 - when not using autoconf, tighten the check for clock_gettime and related
103 functionality.
104
1054.18 Fri Sep 5 17:55:26 CEST 2014
106 - events on files were not always generated properly with the
107 epoll backend (testcase by Assaf Inbal).
108 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
109 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
110 - use a reasonable fallback for EV_NSIG instead of erroring out
111 when we can't detect the signal set size.
112 - in the absence of autoconf, do not use the clock syscall
113 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
114 doing clock_gettime in userspace).
115 - ensure extern "C" function pointers are used for externally-visible
116 loop callbacks (not watcher callbacks yet).
117 - (ecb) work around memory barriers and volatile apparently both being
118 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
119
1204.15 Fri Mar 1 12:04:50 CET 2013
9 - destroying a non-default loop would stop the global waitpid 121 - destroying a non-default loop would stop the global waitpid
10 watcher (Denis Bilenko). 122 watcher (Denis Bilenko).
11 - queueing pending watchers of higher priority from a watcher now invokes 123 - queueing pending watchers of higher priority from a watcher now invokes
12 them in a timely fashion (reported by Denis Bilenko). 124 them in a timely fashion (reported by Denis Bilenko).
13 - add throw() to all libev functions that cannot throw exceptions, for 125 - add throw() to all libev functions that cannot throw exceptions, for
… …
26 - (ecb) add memory fence support for xlC (Darin McBride). 138 - (ecb) add memory fence support for xlC (Darin McBride).
27 - (ecb) add memory fence support for gcc-mips (Anton Kirilov). 139 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
28 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber). 140 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
29 - work around some kernels losing file descriptors by leaking 141 - work around some kernels losing file descriptors by leaking
30 the kqueue descriptor in the child. 142 the kqueue descriptor in the child.
143 - work around linux inotify not reporting IN_ATTRIB changes for directories
144 in many cases.
31 - include sys/syscall.h instead of plain syscall.h. 145 - include sys/syscall.h instead of plain syscall.h.
32 - check for io watcher loops in ev_verify, check for the most 146 - check for io watcher loops in ev_verify, check for the most
33 common reported usage bug in ev_io_start. 147 common reported usage bug in ev_io_start.
34 - chose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET. 148 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
35 - always use WSASend/WSARecv directly on windows, hoping that this 149 - always use WSASend/WSARecv directly on windows, hoping that this
36 works in all cases (unlike read/write/send/recv...). 150 works in all cases (unlike read/write/send/recv...).
151 - try to detect signals around a fork faster (test program by
152 Denis Bilenko).
153 - work around recent glibc versions that leak memory in realloc.
154 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
155 the watcher base set (loop) method.
156 - rewrite the async/signal pipe logic to always keep a valid fd, which
157 simplifies (and hopefully correctifies :) the race checking
158 on fork, at the cost of one extra fd.
159 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
160 inotify-supporting filesystems.
161 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
162 versions ignore it before
163 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
164 - add some untested android support.
165 - enum expressions must be of type int (reported by Juan Pablo L).
37 166
384.11 Sat Feb 4 19:52:39 CET 2012 1674.11 Sat Feb 4 19:52:39 CET 2012
39 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as 168 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
40 was documented already, but not implemented in the repeating case. 169 was documented already, but not implemented in the repeating case.
41 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS. 170 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
… …
225 reinitialising the default loop actually works (patch by 354 reinitialising the default loop actually works (patch by
226 Malek Hadj-Ali). 355 Malek Hadj-Ali).
227 - implement ev_suspend and ev_resume. 356 - implement ev_suspend and ev_resume.
228 - new EV_CUSTOM revents flag for use by applications. 357 - new EV_CUSTOM revents flag for use by applications.
229 - add documentation section about priorities. 358 - add documentation section about priorities.
230 - add a glossary to the dcoumentation. 359 - add a glossary to the documentation.
231 - extend the ev_fork description slightly. 360 - extend the ev_fork description slightly.
232 - optimize a jump out of call_pending. 361 - optimize a jump out of call_pending.
233 362
2343.53 Sun Feb 15 02:38:20 CET 2009 3633.53 Sun Feb 15 02:38:20 CET 2009
235 - fix a bug in event pipe creation on win32 that would cause a 364 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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