| 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 | |
|
|
3 | TODO: revisit 59.x timer in the light of modern powersaving |
|
|
4 | TODO: maybe use timerfd to detect time jumps on linux |
|
|
5 | TODO: document EV_TSTAMP_T |
|
|
6 | |
|
|
7 | 4.29 (EV only) |
|
|
8 | - add io uring autoconf and non-autoconf detection. |
|
|
9 | - disable io_uring when some header files are too old. |
|
|
10 | |
|
|
11 | 4.28 (EV only) |
|
|
12 | - linuxaio backend resulted in random memory corruption |
|
|
13 | when loop is forked. |
|
|
14 | - linuxaio backend might have tried to cancel an iocb |
|
|
15 | multiple times (was unable to trigger this). |
|
|
16 | - linuxaio backend now employs a generation counter to |
|
|
17 | avoid handling spurious events from cancelled requests. |
|
|
18 | - io_cancel can return EINTR, deal with it. also, assume |
|
|
19 | io_submit also returns EINTR. |
|
|
20 | - fix some other minor bugs in linuxaio backend. |
|
|
21 | - ev_tstamp type can now be overriden by defining EV_TSTAMP_T. |
|
|
22 | - cleanup: replace expect_true/false and noinline by their |
|
|
23 | libecb counterparts. |
|
|
24 | - move syscall infrastructure from ev_linuxaio.c to ev.c. |
|
|
25 | - prepare io_uring integration. |
|
|
26 | - tweak ev_floor. |
|
|
27 | - epoll, poll, win32 Sleep and other places that use millisecond |
|
|
28 | reslution now all try to round up times. |
|
|
29 | - solaris port backend didn't compile. |
|
|
30 | - abstract time constants into their macros, for more flexibility. |
|
|
31 | |
|
|
32 | 4.27 Thu Jun 27 22:43:44 CEST 2019 |
|
|
33 | - linux aio backend almost completely rewritten to work around its |
|
|
34 | limitations. |
|
|
35 | - linux aio backend now requires linux 4.19+. |
|
|
36 | - epoll backend now mandatory for linux aio backend. |
|
|
37 | - fail assertions more aggressively on invalid fd's detected |
|
|
38 | in the event loop, do not just silently fd_kill in case of |
|
|
39 | user error. |
|
|
40 | - ev_io_start/ev_io_stop now verify the watcher fd using |
|
|
41 | a syscall when EV_VERIFY is 2 or higher. |
|
|
42 | |
|
|
43 | 4.26 (EV only) |
|
|
44 | - update to libecb 0x00010006. |
|
|
45 | - new experimental linux aio backend (linux 4.18+). |
|
|
46 | - removed redundant 0-ptr check in ev_once. |
|
|
47 | - updated/extended ev_set_allocator documentation. |
|
|
48 | - replaced EMPTY2 macro by array_needsize_noinit. |
|
|
49 | - minor code cleanups. |
|
|
50 | - epoll backend now uses epoll_create1 also after fork. |
|
|
51 | |
|
|
52 | 4.25 Fri Dec 21 07:49:20 CET 2018 |
|
|
53 | - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT |
|
|
54 | (EV_THROW still provided) and now uses noexcept on C++11 or newer. |
|
|
55 | - move the darwin select workaround higher in ev.c, as newer versions of |
|
|
56 | darwin managed to break their broken select even more. |
|
|
57 | - ANDROID => __ANDROID__ (reported by enh@google.com). |
|
|
58 | - disable epoll_create1 on android because it has broken header files |
|
|
59 | and google is unwilling to fix them (reported by enh@google.com). |
|
|
60 | - avoid a minor compilation warning on win32. |
|
|
61 | - c++: remove deprecated dynamic throw() specifications. |
|
|
62 | - c++: improve the (unsupported) bad_loop exception class. |
|
|
63 | - backport perl ev_periodic example to C, untested. |
|
|
64 | - update libecb, biggets change is to include a memory fence |
|
|
65 | in ECB_MEMORY_FENCE_RELEASE on x86/amd64. |
|
|
66 | - minor autoconf/automake modernisation. |
| 2 | |
67 | |
| 3 | 4.24 Wed Dec 28 05:19:55 CET 2016 |
68 | 4.24 Wed Dec 28 05:19:55 CET 2016 |
| 4 | - bump version to 4.24, as the release tarball inexplicably |
69 | - bump version to 4.24, as the release tarball inexplicably |
| 5 | didn't have the right version in ev.h, even though the cvs-tagged |
70 | didn't have the right version in ev.h, even though the cvs-tagged |
| 6 | version did have the right one (reported by Ales Teska). |
71 | version did have the right one (reported by Ales Teska). |
