Mirror of :pserver:anonymous@cvs.schmorp.de/schmorpforge libev http://software.schmorp.de/pkg/libev.html
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

493 lines
26 KiB

14 years ago
9 years ago
9 years ago
9 years ago
9 years ago
9 years ago
10 years ago
10 years ago
10 years ago
11 years ago
11 years ago
11 years ago
11 years ago
11 years ago
11 years ago
12 years ago
11 years ago
12 years ago
10 years ago
12 years ago
11 years ago
12 years ago
11 years ago
12 years ago
12 years ago
12 years ago
12 years ago
12 years ago
12 years ago
12 years ago
10 years ago
9 years ago
13 years ago
13 years ago
13 years ago
13 years ago
13 years ago
13 years ago
13 years ago
13 years ago
11 years ago
13 years ago
13 years ago
13 years ago
11 years ago
11 years ago
14 years ago
  1. Revision history for libev, a high-performance and full-featured event loop.
  2. TODO: ev_loop_wakeup
  3. TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone)
  4. TODO: faq, process a thing in each iteration
  5. TODO: dbeugging tips, ev_verify, ev_init twice
  6. TODO: ev_break for immediate exit (EVBREAK_NOW?)
  7. TODO: ev_feed_child_event
  8. TODO: document the special problem of signals around fork.
  9. TODO: store pid for each signal
  10. TODO: document file descriptor usage per loop
  11. TODO: store loop pid_t and compare isndie signal handler,store 1 for same, 2 for differign pid, clean up in loop_fork
  12. TODO: embed watchers need updating when fd changes
  13. TODO: document portbaility requirements for atomic pointer access
  14. TODO: possible cb aliasing?
  15. TODO: document requirements for function piinters and calling conventions.
  16. - events on files were not always generated properly with the
  17. epoll backend (testcase by Assaf Inbal).
  18. - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
  19. - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
  20. - use a reasonable fallback for EV_NSIG instead of erroring out
  21. when we can't detect the signal set size.
  22. - in the absence of autoconf, do not use the clock syscall
  23. on glibc >= 2.17 (avoids the syscall AND -lrt on systems
  24. doing clock_gettime in userspace).
  25. - ensure extern "C" function pointers are used for externally-visible
  26. loop callbacks (not watcher callbacks yet).
  27. - (ecb) work around memory barriers and volatile apparently both being
  28. broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
  29. 4.15 Fri Mar 1 12:04:50 CET 2013
  30. - destroying a non-default loop would stop the global waitpid
  31. watcher (Denis Bilenko).
  32. - queueing pending watchers of higher priority from a watcher now invokes
  33. them in a timely fashion (reported by Denis Bilenko).
  34. - add throw() to all libev functions that cannot throw exceptions, for
  35. further code size decrease when compiling for C++.
  36. - add throw () to callbacks that must not throw exceptions (allocator,
  37. syserr, loop acquire/release, periodic reschedule cbs).
  38. - fix event_base_loop return code, add event_get_callback, event_base_new,
  39. event_base_get_method calls to improve libevent 1.x emulation and add
  40. some libevent 2.x functionality (based on a patch by Jeff Davey).
  41. - add more memory fences to fix a bug reported by Jeff Davey. Better
  42. be overfenced than underprotected.
  43. - ev_run now returns a boolean status (true meaning watchers are
  44. still active).
  45. - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
  46. libev's EV_ERROR (reported by 191919).
  47. - (ecb) add memory fence support for xlC (Darin McBride).
  48. - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
  49. - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
  50. - work around some kernels losing file descriptors by leaking
  51. the kqueue descriptor in the child.
  52. - work around linux inotify not reporting IN_ATTRIB changes for directories
  53. in many cases.
  54. - include sys/syscall.h instead of plain syscall.h.
  55. - check for io watcher loops in ev_verify, check for the most
  56. common reported usage bug in ev_io_start.
  57. - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
  58. - always use WSASend/WSARecv directly on windows, hoping that this
  59. works in all cases (unlike read/write/send/recv...).
  60. - try to detect signals around a fork faster (test program by
  61. Denis Bilenko).
  62. - work around recent glibc versions that leak memory in realloc.
  63. - rename ev::embed::set to ev::embed::set_embed to avoid clashing
  64. the watcher base set (loop) method.
  65. - rewrite the async/signal pipe logic to always keep a valid fd, which
  66. simplifies (and hopefully correctifies :) the race checking
  67. on fork, at the cost of one extra fd.
  68. - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
  69. inotify-supporting filesystems.
  70. - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
  71. versions ignore it before
  72. (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
  73. - add some untested android support.
  74. - enum expressions must be of type int (reported by Juan Pablo L).
  75. 4.11 Sat Feb 4 19:52:39 CET 2012
  76. - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
  77. was documented already, but not implemented in the repeating case.
  78. - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
  79. - fix a race where the workaround against the epoll fork bugs
  80. caused signals to not be handled anymore.
  81. - correct backend_fudge for most backends, and implement a windows
  82. specific workaround to avoid looping because we call both
  83. select and Sleep, both with different time resolutions.
  84. - document range and guarantees of ev_sleep.
  85. - document reasonable ranges for periodics interval and offset.
  86. - rename backend_fudge to backend_mintime to avoid future confusion :)
  87. - change the default periodic reschedule function to hopefully be more
  88. exact and correct even in corner cases or in the far future.
  89. - do not rely on -lm anymore: use it when available but use our
  90. own floor () if it is missing. This should make it easier to embed,
  91. as no external libraries are required.
  92. - strategically import macros from libecb and mark rarely-used functions
  93. as cache-cold (saving almost 2k code size on typical amd64 setups).
  94. - add Symbols.ev and Symbols.event files, that were missing.
  95. - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
  96. - fix #3 "be smart about timeouts" to not "deadlock" when
  97. timeout == now, also improve the section overall.
  98. - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
  99. - support new EV_API_STATIC mode to make all libev symbols
  100. static.
  101. - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
  102. were empty.
  103. 4.04 Wed Feb 16 09:01:51 CET 2011
  104. - fix two problems in the native win32 backend, where reuse of fd's
  105. with different underlying handles caused handles not to be removed
  106. or added to the select set (analyzed and tested by Bert Belder).
  107. - do no rely on ceil() in ev_e?poll.c.
  108. - backport libev to HP-UX versions before 11 v3.
  109. - configure did not detect nanosleep and clock_gettime properly when
  110. they are available in the libc (as opposed to -lrt).
  111. 4.03 Tue Jan 11 14:37:25 CET 2011
  112. - officially support polling files with all backends.
  113. - support files, /dev/zero etc. the same way as select in the epoll
  114. backend, by generating events on our own.
  115. - ports backend: work around solaris bug 6874410 and many related ones
  116. (EINTR, maybe more), with no performance loss (note that the solaris
  117. bug report is actually wrong, reality is far more bizarre and broken
  118. than that).
  119. - define EV_READ/EV_WRITE as macros in event.h, as some programs use
  120. #ifdef to test for them.
  121. - new (experimental) function: ev_feed_signal.
  122. - new (to become default) EVFLAG_NOSIGMASK flag.
  123. - new EVBACKEND_MASK symbol.
  124. - updated COMMON IDIOMS SECTION.
  125. 4.01 Fri Nov 5 21:51:29 CET 2010
  126. - automake fucked it up, apparently, --add-missing -f is not quite enough
  127. to make it update its files, so 4.00 didn't install ev++.h and
  128. event.h on make install. grrr.
  129. - ev_loop(count|depth) didn't return anything (Robin Haberkorn).
  130. - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
  131. - use "(libev) " prefix for all libev error messages now.
  132. 4.00 Mon Oct 25 12:32:12 CEST 2010
  133. - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
  134. - ev_embed_stop did not correctly stop the watcher (very good
  135. testcase by Vladimir Timofeev).
  136. - ev_run will now always update the current loop time - it erroneously
  137. didn't when idle watchers were active, causing timers not to fire.
  138. - fix a bug where a timeout of zero caused the timer not to fire
  139. in the libevent emulation (testcase by Péter Szabó).
  140. - applied win32 fixes by Michael Lenaghan (also James Mansion).
  141. - replace EV_MINIMAL by EV_FEATURES.
  142. - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
  143. seems the former is *much* faster than the latter.
  144. - linux kernel version detection (for inotify bug workarounds)
  145. did not work properly.
  146. - reduce the number of spurious wake-ups with the ports backend.
  147. - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
  148. - do async init within ev_async_start, not ev_async_set, which avoids
  149. an API quirk where the set function must be called in the C++ API
  150. even when there is nothing to set.
  151. - add (undocumented) EV_ENABLE when adding events with kqueue,
  152. this might help with OS X, which seems to need it despite documenting
  153. not to need it (helpfully pointed out by Tilghman Lesher).
  154. - do not use poll by default on freebsd, it's broken (what isn't
  155. on freebsd...).
  156. - allow to embed epoll on kernels >= 2.6.32.
  157. - configure now prepends -O3, not appends it, so one can still
  158. override it.
  159. - ev.pod: greatly expanded the portability section, added a porting
  160. section, a description of watcher states and made lots of minor fixes.
  161. - disable poll backend on AIX, the poll header spams the namespace
  162. and it's not worth working around dead platforms (reported
  163. and analyzed by Aivars Kalvans).
  164. - improve header file compatibility of the standalone eventfd code
  165. in an obscure case.
  166. - implement EV_AVOID_STDIO option.
  167. - do not use sscanf to parse linux version number (smaller, faster,
  168. no sscanf dependency).
  169. - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
  170. - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
  171. - add section on accept() problems to the manpage.
  172. - rename EV_TIMEOUT to EV_TIMER.
  173. - rename ev_loop_count/depth/verify/loop/unloop.
  174. - remove ev_default_destroy and ev_default_fork.
  175. - switch to two-digit minor version.
  176. - work around an apparent gentoo compiler bug.
  177. - define _DARWIN_UNLIMITED_SELECT. just so.
  178. - use enum instead of #define for most constants.
  179. - improve compatibility to older C++ compilers.
  180. - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now
  181. default arguments when compiled as C++.
  182. - enable automake dependency tracking.
  183. - ev_loop_new no longer leaks memory when loop creation failed.
  184. - new ev_cleanup watcher type.
  185. 3.9 Thu Dec 31 07:59:59 CET 2009
  186. - signalfd is no longer used by default and has to be requested
  187. explicitly - this means that easy to catch bugs become hard to
  188. catch race conditions, but the users have spoken.
  189. - point out the unspecified signal mask in the documentation, and
  190. that this is a race condition regardless of EV_SIGNALFD.
  191. - backport inotify code to C89.
  192. - inotify file descriptors could leak into child processes.
  193. - ev_stat watchers could keep an erroneous extra ref on the loop,
  194. preventing exit when unregistering all watchers (testcases
  195. provided by ry@tinyclouds.org).
  196. - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
  197. symbols to make it easier for apps to do their own fd management.
  198. - support EV_IDLE_ENABLE being disabled in ev++.h
  199. (patch by Didier Spezia).
  200. - take advantage of inotify_init1, if available, to set cloexec/nonblock
  201. on fd creation, to avoid races.
  202. - the signal handling pipe wasn't always initialised under windows
  203. (analysed by lekma).
  204. - changed minimum glibc requirement from glibc 2.9 to 2.7, for
  205. signalfd.
  206. - add missing string.h include (Denis F. Latypoff).
  207. - only replace ev_stat.prev when we detect an actual difference,
  208. so prev is (almost) always different to attr. this might
  209. have caused the problems with 04_stat.t.
  210. - add ev::timer->remaining () method to C++ API.
  211. 3.8 Sun Aug 9 14:30:45 CEST 2009
  212. - incompatible change: do not necessarily reset signal handler
  213. to SIG_DFL when a sighandler is stopped.
  214. - ev_default_destroy did not properly free or zero some members,
  215. potentially causing crashes and memory corruption on repeated
  216. ev_default_destroy/ev_default_loop calls.
  217. - take advantage of signalfd on GNU/Linux systems.
  218. - document that the signal mask might be in an unspecified
  219. state when using libev's signal handling.
  220. - take advantage of some GNU/Linux calls to set cloexec/nonblock
  221. on fd creation, to avoid race conditions.
  222. 3.7 Fri Jul 17 16:36:32 CEST 2009
  223. - ev_unloop and ev_loop wrongly used a global variable to exit loops,
  224. instead of using a per-loop variable (bug caught by accident...).
  225. - the ev_set_io_collect_interval interpretation has changed.
  226. - add new functionality: ev_set_userdata, ev_userdata,
  227. ev_set_invoke_pending_cb, ev_set_loop_release_cb,
  228. ev_invoke_pending, ev_pending_count, together with a long example
  229. about thread locking.
  230. - add ev_timer_remaining (as requested by Denis F. Latypoff).
  231. - add ev_loop_depth.
  232. - calling ev_unloop in fork/prepare watchers will no longer poll
  233. for new events.
  234. - Denis F. Latypoff corrected many typos in example code snippets.
  235. - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
  236. check that the syscall number is available before trying to
  237. use it (reported by ry@tinyclouds).
  238. - use GetSystemTimeAsFileTime instead of _timeb on windows, for
  239. slightly higher accuracy.
  240. - properly declare ev_loop_verify and ev_now_update even when
  241. !EV_MULTIPLICITY.
  242. - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
  243. - support EV_MINIMAL==2 for a reduced API.
  244. - actually 0-initialise struct sigaction when installing signals.
  245. - add section on hibernate and stopped processes to ev_timer docs.
  246. 3.6 Tue Apr 28 02:49:30 CEST 2009
  247. - multiple timers becoming ready within an event loop iteration
  248. will be invoked in the "correct" order now.
  249. - do not leave the event loop early just because we have no active
  250. watchers, fixing a problem when embedding a kqueue loop
  251. that has active kernel events but no registered watchers
  252. (reported by blacksand blacksand).
  253. - correctly zero the idx values for arrays, so destroying and
  254. reinitialising the default loop actually works (patch by
  255. Malek Hadj-Ali).
  256. - implement ev_suspend and ev_resume.
  257. - new EV_CUSTOM revents flag for use by applications.
  258. - add documentation section about priorities.
  259. - add a glossary to the documentation.
  260. - extend the ev_fork description slightly.
  261. - optimize a jump out of call_pending.
  262. 3.53 Sun Feb 15 02:38:20 CET 2009
  263. - fix a bug in event pipe creation on win32 that would cause a
  264. failed assertion on event loop creation (patch by Malek Hadj-Ali).
  265. - probe for CLOCK_REALTIME support at runtime as well and fall
  266. back to gettimeofday if there is an error, to support older
  267. operating systems with newer header files/libraries.
  268. - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
  269. (default most everywhere), otherwise not.
  270. 3.52 Wed Jan 7 21:43:02 CET 2009
  271. - fix compilation of select backend in fd_set mode when NFDBITS is
  272. missing (to get it to compile on QNX, reported by Rodrigo Campos).
  273. - better select-nfds handling when select backend is in fd_set mode.
  274. - diagnose fd_set overruns when select backend is in fd_set mode.
  275. - due to a thinko, instead of disabling everything but
  276. select on the borked OS X platform, everything but select was
  277. allowed (reported by Emanuele Giaquinta).
  278. - actually verify that local and remote port are matching in
  279. libev's socketpair emulation, which makes denial-of-service
  280. attacks harder (but not impossible - it's windows). Make sure
  281. it even works under vista, which thinks that getpeer/sockname
  282. should return fantasy port numbers.
  283. - include "libev" in all assertion messages for potentially
  284. clearer diagnostics.
  285. - event_get_version (libevent compatibility) returned
  286. a useless string instead of the expected version string
  287. (patch by W.C.A. Wijngaards).
  288. 3.51 Wed Dec 24 23:00:11 CET 2008
  289. - fix a bug where an inotify watcher was added twice, causing
  290. freezes on hash collisions (reported and analysed by Graham Leggett).
  291. - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
  292. a direct syscall - slower, but no dependency on librt et al.
  293. - assume negative return values != -1 signals success of port_getn
  294. (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
  295. (no known failure reports, but it doesn't hurt).
  296. - fork detection in ev_embed now stops and restarts the watcher
  297. automatically.
  298. - EXPERIMENTAL: default the method to operator () in ev++.h,
  299. to make it nicer to use functors (requested by Benedek László).
  300. - fixed const object callbacks in ev++.h.
  301. - replaced loop_ref argument of watcher.set (loop) by a direct
  302. ev_loop * in ev++.h, to avoid clashes with functor patch.
  303. - do not try to watch the empty string via inotify.
  304. - inotify watchers could be leaked under certain circumstances.
  305. - OS X 10.5 is actually even more broken than earlier versions,
  306. so fall back to select on that piece of garbage.
  307. - fixed some weirdness in the ev_embed documentation.
  308. 3.49 Wed Nov 19 11:26:53 CET 2008
  309. - ev_stat watchers will now use inotify as a mere hint on
  310. kernels <2.6.25, or if the filesystem is not in the
  311. "known to be good" list.
  312. - better mingw32 compatibility (it's not as borked as native win32)
  313. (analysed by Roger Pack).
  314. - include stdio.h in the example program, as too many people are
  315. confused by the weird C language otherwise. I guess the next thing
  316. I get told is that the "..." ellipses in the examples don't compile
  317. with their C compiler.
  318. 3.48 Thu Oct 30 09:02:37 CET 2008
  319. - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
  320. backend by assuming the kernel event mask hasn't changed if
  321. ADD fails with EEXIST.
  322. - work around spurious event notification bugs in epoll by using
  323. a 32-bit generation counter. recreate kernel state if we receive
  324. spurious notifications or unwanted events. this is very costly,
  325. but I didn't come up with this horrible design.
  326. - use memset to initialise most arrays now and do away with the
  327. init functions.
  328. - expand time-out strategies into a "Be smart about timeouts" section.
  329. - drop the "struct" from all ev_watcher declarations in the
  330. documentation and did other clarifications (yeah, it was a mistake
  331. to have a struct AND a function called ev_loop).
  332. - fix a bug where ev_default would not initialise the default
  333. loop again after it was destroyed with ev_default_destroy.
  334. - rename syserr to ev_syserr to avoid name clashes when embedding,
  335. do similar changes for event.c.
  336. 3.45 Tue Oct 21 21:59:26 CEST 2008
  337. - disable inotify usage on linux <2.6.25, as it is broken
  338. (reported by Yoann Vandoorselaere).
  339. - ev_stat erroneously would try to add inotify watchers
  340. even when inotify wasn't available (this should only
  341. have a performance impact).
  342. - ev_once now passes both timeout and io to the callback if both
  343. occur concurrently, instead of giving timeouts precedence.
  344. - disable EV_USE_INOTIFY when sys/inotify.h is too old.
  345. 3.44 Mon Sep 29 05:18:39 CEST 2008
  346. - embed watchers now automatically invoke ev_loop_fork on the
  347. embedded loop when the parent loop forks.
  348. - new function: ev_now_update (loop).
  349. - verify_watcher was not marked static.
  350. - improve the "associating..." manpage section.
  351. - documentation tweaks here and there.
  352. 3.43 Sun Jul 6 05:34:41 CEST 2008
  353. - include more include files on windows to get struct _stati64
  354. (reported by Chris Hulbert, but doesn't quite fix his issue).
  355. - add missing #include <io.h> in ev.c on windows (reported by
  356. Matt Tolton).
  357. 3.42 Tue Jun 17 12:12:07 CEST 2008
  358. - work around yet another windows bug: FD_SET actually adds fd's
  359. multiple times to the fd_*SET*, despite official MSN docs claiming
  360. otherwise. Reported and well-analysed by Matt Tolton.
  361. - define NFDBITS to 0 when EV_SELECT_IS_WINSOCKET to make it compile
  362. (reported any analysed by Chris Hulbert).
  363. - fix a bug in ev_ebadf (this function is only used to catch
  364. programming errors in the libev user). reported by Matt Tolton.
  365. - fix a bug in fd_intern on win32 (could lead to compile errors
  366. under some circumstances, but would work correctly if it compiles).
  367. reported by Matt Tolton.
  368. - (try to) work around missing lstat on windows.
  369. - pass in the write fd set as except fd set under windows. windows
  370. is so uncontrollably lame that it requires this. this means that
  371. switching off oobinline is not supported (but tcp/ip doesn't
  372. have oob, so that would be stupid anyways.
  373. - use posix module symbol to auto-detect monotonic clock presence
  374. and some other default values.
  375. 3.41 Fri May 23 18:42:54 CEST 2008
  376. - work around an obscure bug in winsocket select: if you
  377. provide only empty fd sets then select returns WSAEINVAL. how sucky.
  378. - improve timer scheduling stability and reduce use of time_epsilon.
  379. - use 1-based 2-heap for EV_MINIMAL, simplifies code, reduces
  380. codesize and makes for better cache-efficiency.
  381. - use 3-based 4-heap for !EV_MINIMAL. this makes better use
  382. of cpu cache lines and gives better growth behaviour than
  383. 2-based heaps.
  384. - cache timestamp within heap for !EV_MINIMAL, to avoid random
  385. memory accesses.
  386. - document/add EV_USE_4HEAP and EV_HEAP_CACHE_AT.
  387. - fix a potential aliasing issue in ev_timer_again.
  388. - add/document ev_periodic_at, retract direct access to ->at.
  389. - improve ev_stat docs.
  390. - add portability requirements section.
  391. - fix manpage headers etc.
  392. - normalise WSA error codes to lower range on windows.
  393. - add consistency check code that can be called automatically
  394. or on demand to check for internal structures (ev_loop_verify).
  395. 3.31 Wed Apr 16 20:45:04 CEST 2008
  396. - added last minute fix for ev_poll.c by Brandon Black.
  397. 3.3 Wed Apr 16 19:04:10 CEST 2008
  398. - event_base_loopexit should return 0 on success
  399. (W.C.A. Wijngaards).
  400. - added linux eventfd support.
  401. - try to autodetect epoll and inotify support
  402. by libc header version if not using autoconf.
  403. - new symbols: EV_DEFAULT_UC and EV_DEFAULT_UC_.
  404. - declare functions defined in ev.h as inline if
  405. C99 or gcc are available.
  406. - enable inlining with gcc versions 2 and 3.
  407. - work around broken poll implementations potentially
  408. not clearing revents field in ev_poll (Brandon Black)
  409. (no such systems are known at this time).
  410. - work around a bug in realloc on openbsd and darwin,
  411. also makes the erroneous valgrind complaints
  412. go away (noted by various people).
  413. - fix ev_async_pending, add c++ wrapper for ev_async
  414. (based on patch sent by Johannes Deisenhofer).
  415. - add sensible set method to ev::embed.
  416. - made integer constants type int in ev.h.
  417. 3.2 Wed Apr 2 17:11:19 CEST 2008
  418. - fix a 64 bit overflow issue in the select backend,
  419. by using fd_mask instead of int for the mask.
  420. - rename internal sighandler to avoid clash with very old perls.
  421. - entering ev_loop will not clear the ONESHOT or NONBLOCKING
  422. flags of any outer loops anymore.
  423. - add ev_async_pending.
  424. 3.1 Thu Mar 13 13:45:22 CET 2008
  425. - implement ev_async watchers.
  426. - only initialise signal pipe on demand.
  427. - make use of sig_atomic_t configurable.
  428. - improved documentation.
  429. 3.0 Mon Jan 28 13:14:47 CET 2008
  430. - API/ABI bump to version 3.0.
  431. - ev++.h includes "ev.h" by default now, not <ev.h>.
  432. - slightly improved documentation.
  433. - speed up signal detection after a fork.
  434. - only optionally return trace status changed in ev_child
  435. watchers.
  436. - experimental (and undocumented) loop wrappers for ev++.h.
  437. 2.01 Tue Dec 25 08:04:41 CET 2007
  438. - separate Changes file.
  439. - fix ev_path_set => ev_stat_set typo.
  440. - remove event_compat.h from the libev tarball.
  441. - change how include files are found.
  442. - doc updates.
  443. - update licenses, explicitly allow for GPL relicensing.
  444. 2.0 Sat Dec 22 17:47:03 CET 2007
  445. - new ev_sleep, ev_set_(io|timeout)_collect_interval.
  446. - removed epoll from embeddable fd set.
  447. - fix embed watchers.
  448. - renamed ev_embed.loop to other.
  449. - added exported Symbol tables.
  450. - undefine member wrapper macros at the end of ev.c.
  451. - respect EV_H in ev++.h.
  452. 1.86 Tue Dec 18 02:36:57 CET 2007
  453. - fix memleak on loop destroy (not relevant for perl).
  454. 1.85 Fri Dec 14 20:32:40 CET 2007
  455. - fix some aliasing issues w.r.t. timers and periodics
  456. (not relevant for perl).
  457. (for historic versions refer to EV/Changes, found in the Perl interface)
  458. 0.1 Wed Oct 31 21:31:48 CET 2007
  459. - original version; hacked together in <24h.