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: somehow unblock procmask? |
3 | TODO: include ev_xyz_start in each example? |
|
|
4 | TODO: which supports signals and child events |
|
|
5 | TODO: pointer races |
|
|
6 | TODO: ev_default_destroy, ev_default_fork |
|
|
7 | - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading. |
|
|
8 | - ev_embed_stop did not correctly stop the watcher (very good |
|
|
9 | testcase by Vladimir Timofeev). |
|
|
10 | - ev_run will now always update the current loop time - it erroneously |
|
|
11 | didn't when idle watchers were active, causing timers not to fire. |
|
|
12 | - fix a bug where a timeout of zero caused the timer not to fire |
|
|
13 | in the libevent emulation (testcase by Péter Szabó). |
|
|
14 | - applied win32 fixes by Michael Lenaghan (also James Mansion). |
|
|
15 | - replace EV_MINIMAL by EV_FEATURES. |
|
|
16 | - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it |
|
|
17 | seems the former is *much* faster than the latter. |
|
|
18 | - linux kernel version detection (for inotify bug workarounds) |
|
|
19 | did not work properly. |
|
|
20 | - reduce the number of spurious wake-ups with the ports backend. |
|
|
21 | - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu). |
|
|
22 | - do async init within ev_async_start, not ev_async_set, which avoids |
|
|
23 | an API quirk where the set function must be called in the C++ API |
|
|
24 | even when there is nothing to set. |
|
|
25 | - add (undocumented) EV_ENABLE when adding events with kqueue, |
|
|
26 | this might help with OS X, which seems to need it despite documenting |
|
|
27 | not to need it (helpfully pointed out by Tilghman Lesher). |
|
|
28 | - do not use poll by default on freebsd, it's broken (what isn't |
|
|
29 | on freebsd...). |
|
|
30 | - allow to embed epoll on kernels >= 2.6.32. |
|
|
31 | - configure now prepends -O3, not appends it, so one can still |
|
|
32 | override it. |
|
|
33 | - ev.pod: greatly expanded the portability section, added a porting |
|
|
34 | section, a description of watcher states and made lots of minor fixes. |
|
|
35 | - disable poll backend on AIX, the poll header spams the namespace |
|
|
36 | and it's not worth working around dead platforms (reported |
|
|
37 | and analyzed by Aivars Kalvans). |
|
|
38 | - improve header file compatibility of the standalone eventfd code |
|
|
39 | in an obscure case. |
|
|
40 | - implement EV_AVOID_STDIO option. |
|
|
41 | - do not use sscanf to parse linux version number (smaller, faster, |
|
|
42 | no sscanf dependency). |
|
|
43 | - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings. |
|
|
44 | - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs. |
|
|
45 | - add section on accept() problems to the manpage. |
|
|
46 | - rename EV_TIMEOUT to EV_TIMER. |
|
|
47 | - rename ev_loop_count/depth/verify/loop/unloop. |
|
|
48 | - remove ev_default_destroy and ev_default_fork. |
|
|
49 | - switch to two-digit minor version. |
|
|
50 | - work around an apparent gentoo compiler bug. |
|
|
51 | - define _DARWIN_UNLIMITED_SELECT. just so. |
|
|
52 | - use enum instead of #define for most constants. |
|
|
53 | - improve compatibility to older C++ compilers. |
|
|
54 | - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now |
|
|
55 | default arguments when compiled as C++. |
|
|
56 | - enable automake dependency tracking. |
|
|
57 | - ev_loop_new no longer leaks memory when loop creation failed. |
|
|
58 | - new ev_cleanup watcher type. |
|
|
59 | |
|
|
60 | 3.9 Thu Dec 31 07:59:59 CET 2009 |
|
|
61 | - signalfd is no longer used by default and has to be requested |
|
|
62 | explicitly - this means that easy to catch bugs become hard to |
|
|
63 | catch race conditions, but the users have spoken. |
|
|
64 | - point out the unspecified signal mask in the documentation, and |
|
|
65 | that this is a race condition regardless of EV_SIGNALFD. |
4 | - backport inotify code to C89. |
66 | - backport inotify code to C89. |
5 | - inotify file descriptors could leak into child processes. |
67 | - inotify file descriptors could leak into child processes. |
6 | - ev_stat watchers could keep an errornous extra ref on the loop, |
68 | - ev_stat watchers could keep an errornous extra ref on the loop, |
7 | preventing exit when unregistering all watchers (testcases |
69 | preventing exit when unregistering all watchers (testcases |
8 | provided by ry@tinyclouds.org). |
70 | provided by ry@tinyclouds.org). |
9 | - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration |
71 | - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration |
10 | symbols to make it easier for apps to do their own fd management. |
72 | symbols to make it easier for apps to do their own fd management. |
11 | - support EV_IDLE_ENABLE being disabled in ev++.h |
73 | - support EV_IDLE_ENABLE being disabled in ev++.h |
12 | (patch by Didier Spezia). |
74 | (patch by Didier Spezia). |
13 | - point out the unspecified signal mask in the documentation. |
|
|
14 | - take advantage of inotify_init1, if available, to set cloexec/nonblock |
75 | - take advantage of inotify_init1, if available, to set cloexec/nonblock |
15 | on fd creation, to avoid races. |
76 | on fd creation, to avoid races. |
16 | - the signal handling pipe wasn't always initialised under windows |
77 | - the signal handling pipe wasn't always initialised under windows |
17 | (analysed by lekma). |
78 | (analysed by lekma). |
18 | - changed minimum glibc requirement from glibc 2.9 to 2.7, for |
79 | - changed minimum glibc requirement from glibc 2.9 to 2.7, for |
19 | signalfd. |
80 | signalfd. |
20 | - add missing string.h include (Denis F. Latypoff). |
81 | - add missing string.h include (Denis F. Latypoff). |
21 | - only replace ev_stat.prev when we detect an actual difference, |
82 | - only replace ev_stat.prev when we detect an actual difference, |
22 | so prev is (almost) always different to attr. this might |
83 | so prev is (almost) always different to attr. this might |
23 | have caused the probems with 04_stat.t. |
84 | have caused the problems with 04_stat.t. |
24 | - add ev::timer->remaining () method to C++ API. |
85 | - add ev::timer->remaining () method to C++ API. |
25 | |
86 | |
26 | 3.8 Sun Aug 9 14:30:45 CEST 2009 |
87 | 3.8 Sun Aug 9 14:30:45 CEST 2009 |
27 | - incompatible change: do not necessarily reset signal handler |
88 | - incompatible change: do not necessarily reset signal handler |
28 | to SIG_DFL when a sighandler is stopped. |
89 | to SIG_DFL when a sighandler is stopped. |
29 | - ev_default_destroy did not properly free or zero some members, |
90 | - ev_default_destroy did not properly free or zero some members, |
30 | potentially causing crashes and memory corruption on repated |
91 | potentially causing crashes and memory corruption on repeated |
31 | ev_default_destroy/ev_default_loop calls. |
92 | ev_default_destroy/ev_default_loop calls. |
32 | - take advantage of signalfd on GNU/Linux systems. |
93 | - take advantage of signalfd on GNU/Linux systems. |
33 | - document that the signal mask might be in an unspecified |
94 | - document that the signal mask might be in an unspecified |
34 | state when using libev's signal handling. |
95 | state when using libev's signal handling. |
35 | - take advantage of some GNU/Linux calls to set cloexec/nonblock |
96 | - take advantage of some GNU/Linux calls to set cloexec/nonblock |
… | |
… | |
157 | do similar changes for event.c. |
218 | do similar changes for event.c. |
158 | |
219 | |
159 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
220 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
160 | - disable inotify usage on linux <2.6.25, as it is broken |
221 | - disable inotify usage on linux <2.6.25, as it is broken |
161 | (reported by Yoann Vandoorselaere). |
222 | (reported by Yoann Vandoorselaere). |
162 | - ev_stat errornously would try to add inotify watchers |
223 | - ev_stat erroneously would try to add inotify watchers |
163 | even when inotify wasn't available (this should only |
224 | even when inotify wasn't available (this should only |
164 | have a performance impact). |
225 | have a performance impact). |
165 | - ev_once now passes both timeout and io to the callback if both |
226 | - ev_once now passes both timeout and io to the callback if both |
166 | occur concurrently, instead of giving timeouts precedence. |
227 | occur concurrently, instead of giving timeouts precedence. |
167 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
228 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
… | |
… | |
235 | - enable inlining with gcc versions 2 and 3. |
296 | - enable inlining with gcc versions 2 and 3. |
236 | - work around broken poll implementations potentially |
297 | - work around broken poll implementations potentially |
237 | not clearing revents field in ev_poll (Brandon Black) |
298 | not clearing revents field in ev_poll (Brandon Black) |
238 | (no such systems are known at this time). |
299 | (no such systems are known at this time). |
239 | - work around a bug in realloc on openbsd and darwin, |
300 | - work around a bug in realloc on openbsd and darwin, |
240 | also makes the errornous valgrind complaints |
301 | also makes the erroneous valgrind complaints |
241 | go away (noted by various people). |
302 | go away (noted by various people). |
242 | - fix ev_async_pending, add c++ wrapper for ev_async |
303 | - fix ev_async_pending, add c++ wrapper for ev_async |
243 | (based on patch sent by Johannes Deisenhofer. |
304 | (based on patch sent by Johannes Deisenhofer). |
244 | - add sensible set method to ev::embed. |
305 | - add sensible set method to ev::embed. |
245 | - made integer constants type int in ev.h. |
306 | - made integer constants type int in ev.h. |
246 | |
307 | |
247 | 3.2 Wed Apr 2 17:11:19 CEST 2008 |
308 | 3.2 Wed Apr 2 17:11:19 CEST 2008 |
248 | - fix a 64 bit overflow issue in the select backend, |
309 | - fix a 64 bit overflow issue in the select backend, |