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_walk |
3 | TODO: move some other examples to common idioms? combining watchers, |
4 | TODO: signal handling per loop |
4 | thread usage, coroutine switch? |
|
|
5 | |
|
|
6 | 4.01 Fri Nov 5 21:51:29 CET 2010 |
|
|
7 | - automake fucked it up, apparently, --add-missing -f is not quite enough |
|
|
8 | to make it update its files, so 4.00 didn't install ev++.h and |
|
|
9 | event.h on make install. grrr. |
|
|
10 | - ev_loop(count|depth) didn't return anything (Robin Haberkorn). |
|
|
11 | - change EV_UNDEF to 0xffffffff to silence some overzealous compilers. |
|
|
12 | - use "(libev) " prefix for all libev error messages now. |
|
|
13 | |
|
|
14 | 4.00 Mon Oct 25 12:32:12 CEST 2010 |
|
|
15 | - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading. |
|
|
16 | - ev_embed_stop did not correctly stop the watcher (very good |
|
|
17 | testcase by Vladimir Timofeev). |
|
|
18 | - ev_run will now always update the current loop time - it erroneously |
|
|
19 | didn't when idle watchers were active, causing timers not to fire. |
|
|
20 | - fix a bug where a timeout of zero caused the timer not to fire |
|
|
21 | in the libevent emulation (testcase by Péter Szabó). |
|
|
22 | - applied win32 fixes by Michael Lenaghan (also James Mansion). |
|
|
23 | - replace EV_MINIMAL by EV_FEATURES. |
|
|
24 | - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it |
|
|
25 | seems the former is *much* faster than the latter. |
|
|
26 | - linux kernel version detection (for inotify bug workarounds) |
|
|
27 | did not work properly. |
|
|
28 | - reduce the number of spurious wake-ups with the ports backend. |
|
|
29 | - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu). |
|
|
30 | - do async init within ev_async_start, not ev_async_set, which avoids |
|
|
31 | an API quirk where the set function must be called in the C++ API |
|
|
32 | even when there is nothing to set. |
|
|
33 | - add (undocumented) EV_ENABLE when adding events with kqueue, |
|
|
34 | this might help with OS X, which seems to need it despite documenting |
|
|
35 | not to need it (helpfully pointed out by Tilghman Lesher). |
|
|
36 | - do not use poll by default on freebsd, it's broken (what isn't |
|
|
37 | on freebsd...). |
|
|
38 | - allow to embed epoll on kernels >= 2.6.32. |
|
|
39 | - configure now prepends -O3, not appends it, so one can still |
|
|
40 | override it. |
|
|
41 | - ev.pod: greatly expanded the portability section, added a porting |
|
|
42 | section, a description of watcher states and made lots of minor fixes. |
|
|
43 | - disable poll backend on AIX, the poll header spams the namespace |
|
|
44 | and it's not worth working around dead platforms (reported |
|
|
45 | and analyzed by Aivars Kalvans). |
|
|
46 | - improve header file compatibility of the standalone eventfd code |
|
|
47 | in an obscure case. |
|
|
48 | - implement EV_AVOID_STDIO option. |
|
|
49 | - do not use sscanf to parse linux version number (smaller, faster, |
|
|
50 | no sscanf dependency). |
|
|
51 | - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings. |
|
|
52 | - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs. |
|
|
53 | - add section on accept() problems to the manpage. |
|
|
54 | - rename EV_TIMEOUT to EV_TIMER. |
|
|
55 | - rename ev_loop_count/depth/verify/loop/unloop. |
|
|
56 | - remove ev_default_destroy and ev_default_fork. |
|
|
57 | - switch to two-digit minor version. |
|
|
58 | - work around an apparent gentoo compiler bug. |
|
|
59 | - define _DARWIN_UNLIMITED_SELECT. just so. |
|
|
60 | - use enum instead of #define for most constants. |
|
|
61 | - improve compatibility to older C++ compilers. |
|
|
62 | - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now |
|
|
63 | default arguments when compiled as C++. |
|
|
64 | - enable automake dependency tracking. |
|
|
65 | - ev_loop_new no longer leaks memory when loop creation failed. |
|
|
66 | - new ev_cleanup watcher type. |
|
|
67 | |
|
|
68 | 3.9 Thu Dec 31 07:59:59 CET 2009 |
|
|
69 | - signalfd is no longer used by default and has to be requested |
|
|
70 | explicitly - this means that easy to catch bugs become hard to |
|
|
71 | catch race conditions, but the users have spoken. |
|
|
72 | - point out the unspecified signal mask in the documentation, and |
|
|
73 | that this is a race condition regardless of EV_SIGNALFD. |
|
|
74 | - backport inotify code to C89. |
|
|
75 | - inotify file descriptors could leak into child processes. |
|
|
76 | - ev_stat watchers could keep an errornous extra ref on the loop, |
|
|
77 | preventing exit when unregistering all watchers (testcases |
|
|
78 | provided by ry@tinyclouds.org). |
|
|
79 | - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration |
|
|
80 | symbols to make it easier for apps to do their own fd management. |
|
|
81 | - support EV_IDLE_ENABLE being disabled in ev++.h |
|
|
82 | (patch by Didier Spezia). |
|
|
83 | - take advantage of inotify_init1, if available, to set cloexec/nonblock |
|
|
84 | on fd creation, to avoid races. |
|
|
85 | - the signal handling pipe wasn't always initialised under windows |
|
|
86 | (analysed by lekma). |
|
|
87 | - changed minimum glibc requirement from glibc 2.9 to 2.7, for |
|
|
88 | signalfd. |
|
|
89 | - add missing string.h include (Denis F. Latypoff). |
|
|
90 | - only replace ev_stat.prev when we detect an actual difference, |
|
|
91 | so prev is (almost) always different to attr. this might |
|
|
92 | have caused the problems with 04_stat.t. |
|
|
93 | - add ev::timer->remaining () method to C++ API. |
|
|
94 | |
|
|
95 | 3.8 Sun Aug 9 14:30:45 CEST 2009 |
|
|
96 | - incompatible change: do not necessarily reset signal handler |
|
|
97 | to SIG_DFL when a sighandler is stopped. |
|
|
98 | - ev_default_destroy did not properly free or zero some members, |
|
|
99 | potentially causing crashes and memory corruption on repeated |
|
|
100 | ev_default_destroy/ev_default_loop calls. |
|
|
101 | - take advantage of signalfd on GNU/Linux systems. |
|
|
102 | - document that the signal mask might be in an unspecified |
|
|
103 | state when using libev's signal handling. |
|
|
104 | - take advantage of some GNU/Linux calls to set cloexec/nonblock |
|
|
105 | on fd creation, to avoid race conditions. |
5 | |
106 | |
6 | 3.7 Fri Jul 17 16:36:32 CEST 2009 |
107 | 3.7 Fri Jul 17 16:36:32 CEST 2009 |
7 | - ev_unloop and ev_loop wrongly used a global variable to exit loops, |
108 | - ev_unloop and ev_loop wrongly used a global variable to exit loops, |
8 | instead of using a per-loop variable (bug caught by accident...). |
109 | instead of using a per-loop variable (bug caught by accident...). |
9 | - the ev_set_io_collect_interval interpretation has changed. |
110 | - the ev_set_io_collect_interval interpretation has changed. |
… | |
… | |
125 | do similar changes for event.c. |
226 | do similar changes for event.c. |
126 | |
227 | |
127 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
228 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
128 | - disable inotify usage on linux <2.6.25, as it is broken |
229 | - disable inotify usage on linux <2.6.25, as it is broken |
129 | (reported by Yoann Vandoorselaere). |
230 | (reported by Yoann Vandoorselaere). |
130 | - ev_stat errornously would try to add inotify watchers |
231 | - ev_stat erroneously would try to add inotify watchers |
131 | even when inotify wasn't available (this should only |
232 | even when inotify wasn't available (this should only |
132 | have a performance impact). |
233 | have a performance impact). |
133 | - ev_once now passes both timeout and io to the callback if both |
234 | - ev_once now passes both timeout and io to the callback if both |
134 | occur concurrently, instead of giving timeouts precedence. |
235 | occur concurrently, instead of giving timeouts precedence. |
135 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
236 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
… | |
… | |
203 | - enable inlining with gcc versions 2 and 3. |
304 | - enable inlining with gcc versions 2 and 3. |
204 | - work around broken poll implementations potentially |
305 | - work around broken poll implementations potentially |
205 | not clearing revents field in ev_poll (Brandon Black) |
306 | not clearing revents field in ev_poll (Brandon Black) |
206 | (no such systems are known at this time). |
307 | (no such systems are known at this time). |
207 | - work around a bug in realloc on openbsd and darwin, |
308 | - work around a bug in realloc on openbsd and darwin, |
208 | also makes the errornous valgrind complaints |
309 | also makes the erroneous valgrind complaints |
209 | go away (noted by various people). |
310 | go away (noted by various people). |
210 | - fix ev_async_pending, add c++ wrapper for ev_async |
311 | - fix ev_async_pending, add c++ wrapper for ev_async |
211 | (based on patch sent by Johannes Deisenhofer. |
312 | (based on patch sent by Johannes Deisenhofer). |
212 | - add sensible set method to ev::embed. |
313 | - add sensible set method to ev::embed. |
213 | - made integer constants type int in ev.h. |
314 | - made integer constants type int in ev.h. |
214 | |
315 | |
215 | 3.2 Wed Apr 2 17:11:19 CEST 2008 |
316 | 3.2 Wed Apr 2 17:11:19 CEST 2008 |
216 | - fix a 64 bit overflow issue in the select backend, |
317 | - fix a 64 bit overflow issue in the select backend, |