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

Comparing libev/Changes (file contents):
Revision 1.75 by root, Mon Oct 27 11:08:29 2008 UTC vs.
Revision 1.170 by root, Tue Mar 16 20:39:29 2010 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
3WISH? monotonic clocks times/GetTickCount for coarse corrections? 3TODO: ABI??? API????? Changes???
4TODO: win32 write() to socket for signal handling
5 - ev_embed_stop did not correctly stop the watcher (very good
6 testcase by Vladimir Timofeev).
7 - applied win32 fixes by Michael Lenaghan (also James Mansion).
8 - replace EV_MINIMAL by EV_FEATURES.
9 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
10 - configure now prepends -O3, not appends it, so one can still
11 override it.
12 - disable poll backend on AIX, the poll header spams the namespace
13 and it's not worth working around dead platforms (reported
14 and analyzed by Aivars Kalvans).
15 - improve header file compatibility of the standalone eventfd code
16 in an obscure case.
17 - implement EV_AVOID_STDIO option.
18 - do not use sscanf to parse linux version number (smaller, faster,
19 no sscanf dependency).
20 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
21 - update libev.m4 HAVE_CLOCK_SYSCALL test for newwer glibcs.
22 - add section on accept() problems to the manpage.
23 - rename EV_TIMEOUT to EV_TIMER.
24 - remove loop_ name component from ev_loop_count/depth/verify.
25 - switch to two-digit minor version.
4 26
273.9 Thu Dec 31 07:59:59 CET 2009
28 - signalfd is no longer used by default and has to be requested
29 explicitly - this means that easy to catch bugs become hard to
30 catch race conditions, but the users have spoken.
31 - point out the unspecified signal mask in the documentation, and
32 that this is a race condition regardless of EV_SIGNALFD.
33 - backport inotify code to C89.
34 - inotify file descriptors could leak into child processes.
35 - ev_stat watchers could keep an errornous extra ref on the loop,
36 preventing exit when unregistering all watchers (testcases
37 provided by ry@tinyclouds.org).
38 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
39 symbols to make it easier for apps to do their own fd management.
40 - support EV_IDLE_ENABLE being disabled in ev++.h
41 (patch by Didier Spezia).
42 - take advantage of inotify_init1, if available, to set cloexec/nonblock
43 on fd creation, to avoid races.
44 - the signal handling pipe wasn't always initialised under windows
45 (analysed by lekma).
46 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
47 signalfd.
48 - add missing string.h include (Denis F. Latypoff).
49 - only replace ev_stat.prev when we detect an actual difference,
50 so prev is (almost) always different to attr. this might
51 have caused the probems with 04_stat.t.
52 - add ev::timer->remaining () method to C++ API.
53
543.8 Sun Aug 9 14:30:45 CEST 2009
55 - incompatible change: do not necessarily reset signal handler
56 to SIG_DFL when a sighandler is stopped.
57 - ev_default_destroy did not properly free or zero some members,
58 potentially causing crashes and memory corruption on repated
59 ev_default_destroy/ev_default_loop calls.
60 - take advantage of signalfd on GNU/Linux systems.
61 - document that the signal mask might be in an unspecified
62 state when using libev's signal handling.
63 - take advantage of some GNU/Linux calls to set cloexec/nonblock
64 on fd creation, to avoid race conditions.
65
663.7 Fri Jul 17 16:36:32 CEST 2009
67 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
68 instead of using a per-loop variable (bug caught by accident...).
69 - the ev_set_io_collect_interval interpretation has changed.
70 - add new functionality: ev_set_userdata, ev_userdata,
71 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
72 ev_invoke_pending, ev_pending_count, together with a long example
73 about thread locking.
74 - add ev_timer_remaining (as requested by Denis F. Latypoff).
75 - add ev_loop_depth.
76 - calling ev_unloop in fork/prepare watchers will no longer poll
77 for new events.
78 - Denis F. Latypoff corrected many typos in example code snippets.
79 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
80 check that the syscall number is available before trying to
81 use it (reported by ry@tinyclouds).
82 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
83 slightly higher accuracy.
84 - properly declare ev_loop_verify and ev_now_update even when
85 !EV_MULTIPLICITY.
86 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
87 - support EV_MINIMAL==2 for a reduced API.
88 - actually 0-initialise struct sigaction when installing signals.
89 - add section on hibernate and stopped processes to ev_timer docs.
90
913.6 Tue Apr 28 02:49:30 CEST 2009
92 - multiple timers becoming ready within an event loop iteration
93 will be invoked in the "correct" order now.
94 - do not leave the event loop early just because we have no active
95 watchers, fixing a problem when embedding a kqueue loop
96 that has active kernel events but no registered watchers
97 (reported by blacksand blacksand).
98 - correctly zero the idx values for arrays, so destroying and
99 reinitialising the default loop actually works (patch by
100 Malek Hadj-Ali).
101 - implement ev_suspend and ev_resume.
102 - new EV_CUSTOM revents flag for use by applications.
103 - add documentation section about priorites.
104 - add a glossary to the dcoumentation.
105 - extend the ev_fork description slightly.
106 - optimize a jump out of call_pending.
107
1083.53 Sun Feb 15 02:38:20 CET 2009
109 - fix a bug in event pipe creation on win32 that would cause a
110 failed assertion on event loop creation (patch by Malek Hadj-Ali).
111 - probe for CLOCK_REALTIME support at runtime as well and fall
112 back to gettimeofday if there is an error, to support older
113 operating systems with newer header files/libraries.
114 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
115 (default most everywhere), otherwise not.
116
1173.52 Wed Jan 7 21:43:02 CET 2009
118 - fix compilation of select backend in fd_set mode when NFDBITS is
119 missing (to get it to compile on QNX, reported by Rodrigo Campos).
120 - better select-nfds handling when select backend is in fd_set mode.
121 - diagnose fd_set overruns when select backend is in fd_set mode.
122 - due to a thinko, instead of disabling everything but
123 select on the borked OS X platform, everything but select was
124 allowed (reported by Emanuele Giaquinta).
125 - actually verify that local and remote port are matching in
126 libev's socketpair emulation, which makes denial-of-service
127 attacks harder (but not impossible - it's windows). Make sure
128 it even works under vista, which thinks that getpeer/sockname
129 should return fantasy port numbers.
130 - include "libev" in all assertion messages for potentially
131 clearer diagnostics.
132 - event_get_version (libevent compatibility) returned
133 a useless string instead of the expected version string
134 (patch by W.C.A. Wijngaards).
135
1363.51 Wed Dec 24 23:00:11 CET 2008
137 - fix a bug where an inotify watcher was added twice, causing
138 freezes on hash collisions (reported and analysed by Graham Leggett).
139 - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
140 a direct syscall - slower, but no dependency on librt et al.
141 - assume negative return values != -1 signals success of port_getn
142 (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
143 (no known failure reports, but it doesn't hurt).
144 - fork detection in ev_embed now stops and restarts the watcher
145 automatically.
146 - EXPERIMENTAL: default the method to operator () in ev++.h,
147 to make it nicer to use functors (requested by Benedek László).
148 - fixed const object callbacks in ev++.h.
149 - replaced loop_ref argument of watcher.set (loop) by a direct
150 ev_loop * in ev++.h, to avoid clashes with functor patch.
151 - do not try to watch the empty string via inotify.
152 - inotify watchers could be leaked under certain circumstances.
153 - OS X 10.5 is actually even more broken than earlier versions,
154 so fall back to select on that piece of garbage.
155 - fixed some weirdness in the ev_embed documentation.
156
1573.49 Wed Nov 19 11:26:53 CET 2008
158 - ev_stat watchers will now use inotify as a mere hint on
159 kernels <2.6.25, or if the filesystem is not in the
160 "known to be good" list.
161 - better mingw32 compatibility (it's not as borked as native win32)
162 (analysed by Roger Pack).
163 - include stdio.h in the example program, as too many people are
164 confused by the weird C language otherwise. I guess the next thing
165 I get told is that the "..." ellipses in the examples don't compile
166 with their C compiler.
167
1683.48 Thu Oct 30 09:02:37 CET 2008
5 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll 169 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
6 backend by assuming the kernel event mask hasn't changed if 170 backend by assuming the kernel event mask hasn't changed if
7 ADD fails with EEXIST. 171 ADD fails with EEXIST.
8 - work around spurious event notification bugs in epoll by using 172 - work around spurious event notification bugs in epoll by using
9 an 8-bit generation counter. 173 a 32-bit generation counter. recreate kernel state if we receive
174 spurious notifications or unwanted events. this is very costly,
175 but I didn't come up with this horrible design.
10 - use memset to initialise most arrays now and do away with the 176 - use memset to initialise most arrays now and do away with the
11 init functions. 177 init functions.
12 - expand time-out strategies into a "Be smart about timeouts" section. 178 - expand time-out strategies into a "Be smart about timeouts" section.
13 - drop the "struct" from all ev_watcher declarations in the 179 - drop the "struct" from all ev_watcher declarations in the
14 documentation and did other clarifications (yeah, it was a mistake 180 documentation and did other clarifications (yeah, it was a mistake
15 to have a struct AND a function called ev_loop). 181 to have a struct AND a function called ev_loop).
16 - fix a bug where ev_default would not initialise the default 182 - fix a bug where ev_default would not initialise the default
17 loop again after it was destroyed with ev_default_destroy. 183 loop again after it was destroyed with ev_default_destroy.
184 - rename syserr to ev_syserr to avoid name clashes when embedding,
185 do similar changes for event.c.
18 186
193.45 Tue Oct 21 21:59:26 CEST 2008 1873.45 Tue Oct 21 21:59:26 CEST 2008
20 - disable inotify usage on linux <2.6.25, as it is broken 188 - disable inotify usage on linux <2.6.25, as it is broken
21 (reported by Yoann Vandoorselaere). 189 (reported by Yoann Vandoorselaere).
22 - ev_stat errornously would try to add inotify watchers 190 - ev_stat errornously would try to add inotify watchers

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines