ViewVC Help
View File | Revision Log | Show Annotations | Download File
Revision: 1.359
Committed: Tue Jan 11 10:56:01 2011 UTC (13 years, 6 months ago) by root
Branch: MAIN
CVS Tags: rel-4_03
Changes since 1.358: +1 -1 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 NAME
3 libev - a high performance full-featured event loop written in C
5 =head1 SYNOPSIS
7 #include <ev.h>
11 // a single header file is required
12 #include <ev.h>
14 #include <stdio.h> // for puts
16 // every watcher type has its own typedef'd struct
17 // with the name ev_TYPE
18 ev_io stdin_watcher;
19 ev_timer timeout_watcher;
21 // all watcher callbacks have a similar signature
22 // this callback is called when data is readable on stdin
23 static void
24 stdin_cb (EV_P_ ev_io *w, int revents)
25 {
26 puts ("stdin ready");
27 // for one-shot events, one must manually stop the watcher
28 // with its corresponding stop function.
29 ev_io_stop (EV_A_ w);
31 // this causes all nested ev_run's to stop iterating
32 ev_break (EV_A_ EVBREAK_ALL);
33 }
35 // another callback, this time for a time-out
36 static void
37 timeout_cb (EV_P_ ev_timer *w, int revents)
38 {
39 puts ("timeout");
40 // this causes the innermost ev_run to stop iterating
41 ev_break (EV_A_ EVBREAK_ONE);
42 }
44 int
45 main (void)
46 {
47 // use the default event loop unless you have special needs
48 struct ev_loop *loop = EV_DEFAULT;
50 // initialise an io watcher, then start it
51 // this one will watch for stdin to become readable
52 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
53 ev_io_start (loop, &stdin_watcher);
55 // initialise a timer watcher, then start it
56 // simple non-repeating 5.5 second timeout
57 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
58 ev_timer_start (loop, &timeout_watcher);
60 // now wait for events to arrive
61 ev_run (loop, 0);
63 // unloop was called, so exit
64 return 0;
65 }
69 This document documents the libev software package.
71 The newest version of this document is also available as an html-formatted
72 web page you might find easier to navigate when reading it for the first
73 time: L<>.
75 While this document tries to be as complete as possible in documenting
76 libev, its usage and the rationale behind its design, it is not a tutorial
77 on event-based programming, nor will it introduce event-based programming
78 with libev.
80 Familiarity with event based programming techniques in general is assumed
81 throughout this document.
85 This manual tries to be very detailed, but unfortunately, this also makes
86 it very long. If you just want to know the basics of libev, I suggest
87 reading L<ANATOMY OF A WATCHER>, then the L<EXAMPLE PROGRAM> above and
88 look up the missing functions in L<GLOBAL FUNCTIONS> and the C<ev_io> and
89 C<ev_timer> sections in L<WATCHER TYPES>.
91 =head1 ABOUT LIBEV
93 Libev is an event loop: you register interest in certain events (such as a
94 file descriptor being readable or a timeout occurring), and it will manage
95 these event sources and provide your program with events.
97 To do this, it must take more or less complete control over your process
98 (or thread) by executing the I<event loop> handler, and will then
99 communicate events via a callback mechanism.
101 You register interest in certain events by registering so-called I<event
102 watchers>, which are relatively small C structures you initialise with the
103 details of the event, and then hand it over to libev by I<starting> the
104 watcher.
106 =head2 FEATURES
108 Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
109 BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
110 for file descriptor events (C<ev_io>), the Linux C<inotify> interface
111 (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
112 inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
113 timers (C<ev_timer>), absolute timers with customised rescheduling
114 (C<ev_periodic>), synchronous signals (C<ev_signal>), process status
115 change events (C<ev_child>), and event watchers dealing with the event
116 loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
117 C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
118 limited support for fork events (C<ev_fork>).
120 It also is quite fast (see this
121 L<benchmark|> comparing it to libevent
122 for example).
124 =head2 CONVENTIONS
126 Libev is very configurable. In this manual the default (and most common)
127 configuration will be described, which supports multiple event loops. For
128 more info about various configuration options please have a look at
129 B<EMBED> section in this manual. If libev was configured without support
130 for multiple event loops, then all functions taking an initial argument of
131 name C<loop> (which is always of type C<struct ev_loop *>) will not have
132 this argument.
136 Libev represents time as a single floating point number, representing
137 the (fractional) number of seconds since the (POSIX) epoch (in practice
138 somewhere near the beginning of 1970, details are complicated, don't
139 ask). This type is called C<ev_tstamp>, which is what you should use
140 too. It usually aliases to the C<double> type in C. When you need to do
141 any calculations on it, you should treat it as some floating point value.
143 Unlike the name component C<stamp> might indicate, it is also used for
144 time differences (e.g. delays) throughout libev.
148 Libev knows three classes of errors: operating system errors, usage errors
149 and internal errors (bugs).
151 When libev catches an operating system error it cannot handle (for example
152 a system call indicating a condition libev cannot fix), it calls the callback
153 set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
154 abort. The default is to print a diagnostic message and to call C<abort
155 ()>.
157 When libev detects a usage error such as a negative timer interval, then
158 it will print a diagnostic message and abort (via the C<assert> mechanism,
159 so C<NDEBUG> will disable this checking): these are programming errors in
160 the libev caller and need to be fixed there.
162 Libev also has a few internal error-checking C<assert>ions, and also has
163 extensive consistency checking code. These do not trigger under normal
164 circumstances, as they indicate either a bug in libev or worse.
169 These functions can be called anytime, even before initialising the
170 library in any way.
172 =over 4
174 =item ev_tstamp ev_time ()
176 Returns the current time as libev would use it. Please note that the
177 C<ev_now> function is usually faster and also often returns the timestamp
178 you actually want to know. Also interesting is the combination of
179 C<ev_update_now> and C<ev_now>.
181 =item ev_sleep (ev_tstamp interval)
183 Sleep for the given interval: The current thread will be blocked until
184 either it is interrupted or the given time interval has passed. Basically
185 this is a sub-second-resolution C<sleep ()>.
187 =item int ev_version_major ()
189 =item int ev_version_minor ()
191 You can find out the major and minor ABI version numbers of the library
192 you linked against by calling the functions C<ev_version_major> and
193 C<ev_version_minor>. If you want, you can compare against the global
194 symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
195 version of the library your program was compiled against.
197 These version numbers refer to the ABI version of the library, not the
198 release version.
200 Usually, it's a good idea to terminate if the major versions mismatch,
201 as this indicates an incompatible change. Minor versions are usually
202 compatible to older versions, so a larger minor version alone is usually
203 not a problem.
205 Example: Make sure we haven't accidentally been linked against the wrong
206 version (note, however, that this will not detect other ABI mismatches,
207 such as LFS or reentrancy).
209 assert (("libev version mismatch",
210 ev_version_major () == EV_VERSION_MAJOR
211 && ev_version_minor () >= EV_VERSION_MINOR));
213 =item unsigned int ev_supported_backends ()
215 Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
216 value) compiled into this binary of libev (independent of their
217 availability on the system you are running on). See C<ev_default_loop> for
218 a description of the set values.
220 Example: make sure we have the epoll method, because yeah this is cool and
221 a must have and can we have a torrent of it please!!!11
223 assert (("sorry, no epoll, no sex",
224 ev_supported_backends () & EVBACKEND_EPOLL));
226 =item unsigned int ev_recommended_backends ()
228 Return the set of all backends compiled into this binary of libev and
229 also recommended for this platform, meaning it will work for most file
230 descriptor types. This set is often smaller than the one returned by
231 C<ev_supported_backends>, as for example kqueue is broken on most BSDs
232 and will not be auto-detected unless you explicitly request it (assuming
233 you know what you are doing). This is the set of backends that libev will
234 probe for if you specify no backends explicitly.
236 =item unsigned int ev_embeddable_backends ()
238 Returns the set of backends that are embeddable in other event loops. This
239 value is platform-specific but can include backends not available on the
240 current system. To find which embeddable backends might be supported on
241 the current system, you would need to look at C<ev_embeddable_backends ()
242 & ev_supported_backends ()>, likewise for recommended ones.
244 See the description of C<ev_embed> watchers for more info.
246 =item ev_set_allocator (void *(*cb)(void *ptr, long size))
248 Sets the allocation function to use (the prototype is similar - the
249 semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
250 used to allocate and free memory (no surprises here). If it returns zero
251 when memory needs to be allocated (C<size != 0>), the library might abort
252 or take some potentially destructive action.
254 Since some systems (at least OpenBSD and Darwin) fail to implement
255 correct C<realloc> semantics, libev will use a wrapper around the system
256 C<realloc> and C<free> functions by default.
258 You could override this function in high-availability programs to, say,
259 free some memory if it cannot allocate memory, to use a special allocator,
260 or even to sleep a while and retry until some memory is available.
262 Example: Replace the libev allocator with one that waits a bit and then
263 retries (example requires a standards-compliant C<realloc>).
265 static void *
266 persistent_realloc (void *ptr, size_t size)
267 {
268 for (;;)
269 {
270 void *newptr = realloc (ptr, size);
272 if (newptr)
273 return newptr;
275 sleep (60);
276 }
277 }
279 ...
280 ev_set_allocator (persistent_realloc);
282 =item ev_set_syserr_cb (void (*cb)(const char *msg))
284 Set the callback function to call on a retryable system call error (such
285 as failed select, poll, epoll_wait). The message is a printable string
286 indicating the system call or subsystem causing the problem. If this
287 callback is set, then libev will expect it to remedy the situation, no
288 matter what, when it returns. That is, libev will generally retry the
289 requested operation, or, if the condition doesn't go away, do bad stuff
290 (such as abort).
292 Example: This is basically the same thing that libev does internally, too.
294 static void
295 fatal_error (const char *msg)
296 {
297 perror (msg);
298 abort ();
299 }
301 ...
302 ev_set_syserr_cb (fatal_error);
304 =item ev_feed_signal (int signum)
306 This function can be used to "simulate" a signal receive. It is completely
307 safe to call this function at any time, from any context, including signal
308 handlers or random threads.
310 Its main use is to customise signal handling in your process, especially
311 in the presence of threads. For example, you could block signals
312 by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
313 creating any loops), and in one thread, use C<sigwait> or any other
314 mechanism to wait for signals, then "deliver" them to libev by calling
315 C<ev_feed_signal>.
317 =back
321 An event loop is described by a C<struct ev_loop *> (the C<struct> is
322 I<not> optional in this case unless libev 3 compatibility is disabled, as
323 libev 3 had an C<ev_loop> function colliding with the struct name).
325 The library knows two types of such loops, the I<default> loop, which
326 supports child process events, and dynamically created event loops which
327 do not.
329 =over 4
331 =item struct ev_loop *ev_default_loop (unsigned int flags)
333 This returns the "default" event loop object, which is what you should
334 normally use when you just need "the event loop". Event loop objects and
335 the C<flags> parameter are described in more detail in the entry for
336 C<ev_loop_new>.
338 If the default loop is already initialised then this function simply
339 returns it (and ignores the flags. If that is troubling you, check
340 C<ev_backend ()> afterwards). Otherwise it will create it with the given
341 flags, which should almost always be C<0>, unless the caller is also the
342 one calling C<ev_run> or otherwise qualifies as "the main program".
344 If you don't know what event loop to use, use the one returned from this
345 function (or via the C<EV_DEFAULT> macro).
347 Note that this function is I<not> thread-safe, so if you want to use it
348 from multiple threads, you have to employ some kind of mutex (note also
349 that this case is unlikely, as loops cannot be shared easily between
350 threads anyway).
352 The default loop is the only loop that can handle C<ev_child> watchers,
353 and to do this, it always registers a handler for C<SIGCHLD>. If this is
354 a problem for your application you can either create a dynamic loop with
355 C<ev_loop_new> which doesn't do that, or you can simply overwrite the
356 C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
358 Example: This is the most typical usage.
360 if (!ev_default_loop (0))
361 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
363 Example: Restrict libev to the select and poll backends, and do not allow
364 environment settings to be taken into account:
368 =item struct ev_loop *ev_loop_new (unsigned int flags)
370 This will create and initialise a new event loop object. If the loop
371 could not be initialised, returns false.
373 This function is thread-safe, and one common way to use libev with
374 threads is indeed to create one loop per thread, and using the default
375 loop in the "main" or "initial" thread.
377 The flags argument can be used to specify special behaviour or specific
378 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
380 The following flags are supported:
382 =over 4
384 =item C<EVFLAG_AUTO>
386 The default flags value. Use this if you have no clue (it's the right
387 thing, believe me).
389 =item C<EVFLAG_NOENV>
391 If this flag bit is or'ed into the flag value (or the program runs setuid
392 or setgid) then libev will I<not> look at the environment variable
393 C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
394 override the flags completely if it is found in the environment. This is
395 useful to try out specific backends to test their performance, or to work
396 around bugs.
400 Instead of calling C<ev_loop_fork> manually after a fork, you can also
401 make libev check for a fork in each iteration by enabling this flag.
403 This works by calling C<getpid ()> on every iteration of the loop,
404 and thus this might slow down your event loop if you do a lot of loop
405 iterations and little real work, but is usually not noticeable (on my
406 GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
407 without a system call and thus I<very> fast, but my GNU/Linux system also has
408 C<pthread_atfork> which is even faster).
410 The big advantage of this flag is that you can forget about fork (and
411 forget about forgetting to tell libev about forking) when you use this
412 flag.
414 This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
415 environment variable.
419 When this flag is specified, then libev will not attempt to use the
420 I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
421 testing, this flag can be useful to conserve inotify file descriptors, as
422 otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
426 When this flag is specified, then libev will attempt to use the
427 I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
428 delivers signals synchronously, which makes it both faster and might make
429 it possible to get the queued signal data. It can also simplify signal
430 handling with threads, as long as you properly block signals in your
431 threads that are not interested in handling them.
433 Signalfd will not be used by default as this changes your signal mask, and
434 there are a lot of shoddy libraries and programs (glib's threadpool for
435 example) that can't properly initialise their signal masks.
439 When this flag is specified, then libev will avoid to modify the signal
440 mask. Specifically, this means you ahve to make sure signals are unblocked
441 when you want to receive them.
443 This behaviour is useful when you want to do your own signal handling, or
444 want to handle signals only in specific threads and want to avoid libev
445 unblocking the signals.
447 This flag's behaviour will become the default in future versions of libev.
449 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
451 This is your standard select(2) backend. Not I<completely> standard, as
452 libev tries to roll its own fd_set with no limits on the number of fds,
453 but if that fails, expect a fairly low limit on the number of fds when
454 using this backend. It doesn't scale too well (O(highest_fd)), but its
455 usually the fastest backend for a low number of (low-numbered :) fds.
457 To get good performance out of this backend you need a high amount of
458 parallelism (most of the file descriptors should be busy). If you are
459 writing a server, you should C<accept ()> in a loop to accept as many
460 connections as possible during one iteration. You might also want to have
461 a look at C<ev_set_io_collect_interval ()> to increase the amount of
462 readiness notifications you get per iteration.
464 This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
465 C<writefds> set (and to work around Microsoft Windows bugs, also onto the
466 C<exceptfds> set on that platform).
468 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
470 And this is your standard poll(2) backend. It's more complicated
471 than select, but handles sparse fds better and has no artificial
472 limit on the number of fds you can use (except it will slow down
473 considerably with a lot of inactive fds). It scales similarly to select,
474 i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
475 performance tips.
477 This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
480 =item C<EVBACKEND_EPOLL> (value 4, Linux)
482 Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
483 kernels).
485 For few fds, this backend is a bit little slower than poll and select,
486 but it scales phenomenally better. While poll and select usually scale
487 like O(total_fds) where n is the total number of fds (or the highest fd),
488 epoll scales either O(1) or O(active_fds).
490 The epoll mechanism deserves honorable mention as the most misdesigned
491 of the more advanced event mechanisms: mere annoyances include silently
492 dropping file descriptors, requiring a system call per change per file
493 descriptor (and unnecessary guessing of parameters), problems with dup,
494 returning before the timeout value, resulting in additional iterations
495 (and only giving 5ms accuracy while select on the same platform gives
496 0.1ms) and so on. The biggest issue is fork races, however - if a program
497 forks then I<both> parent and child process have to recreate the epoll
498 set, which can take considerable time (one syscall per file descriptor)
499 and is of course hard to detect.
501 Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
502 of course I<doesn't>, and epoll just loves to report events for totally
503 I<different> file descriptors (even already closed ones, so one cannot
504 even remove them from the set) than registered in the set (especially
505 on SMP systems). Libev tries to counter these spurious notifications by
506 employing an additional generation counter and comparing that against the
507 events to filter out spurious ones, recreating the set when required. Last
508 not least, it also refuses to work with some file descriptors which work
509 perfectly fine with C<select> (files, many character devices...).
511 Epoll is truly the train wreck analog among event poll mechanisms,
512 a frankenpoll, cobbled together in a hurry, no thought to design or
513 interaction with others.
515 While stopping, setting and starting an I/O watcher in the same iteration
516 will result in some caching, there is still a system call per such
517 incident (because the same I<file descriptor> could point to a different
518 I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
519 file descriptors might not work very well if you register events for both
520 file descriptors.
522 Best performance from this backend is achieved by not unregistering all
523 watchers for a file descriptor until it has been closed, if possible,
524 i.e. keep at least one watcher active per fd at all times. Stopping and
525 starting a watcher (without re-setting it) also usually doesn't cause
526 extra overhead. A fork can both result in spurious notifications as well
527 as in libev having to destroy and recreate the epoll object, which can
528 take considerable time and thus should be avoided.
530 All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
531 faster than epoll for maybe up to a hundred file descriptors, depending on
532 the usage. So sad.
534 While nominally embeddable in other event loops, this feature is broken in
535 all kernel versions tested so far.
537 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
540 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
542 Kqueue deserves special mention, as at the time of this writing, it
543 was broken on all BSDs except NetBSD (usually it doesn't work reliably
544 with anything but sockets and pipes, except on Darwin, where of course
545 it's completely useless). Unlike epoll, however, whose brokenness
546 is by design, these kqueue bugs can (and eventually will) be fixed
547 without API changes to existing programs. For this reason it's not being
548 "auto-detected" unless you explicitly specify it in the flags (i.e. using
549 C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
550 system like NetBSD.
552 You still can embed kqueue into a normal poll or select backend and use it
553 only for sockets (after having made sure that sockets work with kqueue on
554 the target platform). See C<ev_embed> watchers for more info.
556 It scales in the same way as the epoll backend, but the interface to the
557 kernel is more efficient (which says nothing about its actual speed, of
558 course). While stopping, setting and starting an I/O watcher does never
559 cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
560 two event changes per incident. Support for C<fork ()> is very bad (but
561 sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
562 cases
564 This backend usually performs well under most conditions.
566 While nominally embeddable in other event loops, this doesn't work
567 everywhere, so you might need to test for this. And since it is broken
568 almost everywhere, you should only use it when you have a lot of sockets
569 (for which it usually works), by embedding it into another event loop
570 (e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
571 also broken on OS X)) and, did I mention it, using it only for sockets.
573 This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
574 C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
575 C<NOTE_EOF>.
577 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
579 This is not implemented yet (and might never be, unless you send me an
580 implementation). According to reports, C</dev/poll> only supports sockets
581 and is not embeddable, which would limit the usefulness of this backend
582 immensely.
584 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
586 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
587 it's really slow, but it still scales very well (O(active_fds)).
589 While this backend scales well, it requires one system call per active
590 file descriptor per loop iteration. For small and medium numbers of file
591 descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
592 might perform better.
594 On the positive side, this backend actually performed fully to
595 specification in all tests and is fully embeddable, which is a rare feat
596 among the OS-specific backends (I vastly prefer correctness over speed
597 hacks).
599 On the negative side, the interface is I<bizarre> - so bizarre that
600 even sun itself gets it wrong in their code examples: The event polling
601 function sometimes returning events to the caller even though an error
602 occurred, but with no indication whether it has done so or not (yes, it's
603 even documented that way) - deadly for edge-triggered interfaces where
604 you absolutely have to know whether an event occurred or not because you
605 have to re-arm the watcher.
607 Fortunately libev seems to be able to work around these idiocies.
609 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
612 =item C<EVBACKEND_ALL>
614 Try all backends (even potentially broken ones that wouldn't be tried
615 with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
618 It is definitely not recommended to use this flag, use whatever
619 C<ev_recommended_backends ()> returns, or simply do not specify a backend
620 at all.
624 Not a backend at all, but a mask to select all backend bits from a
625 C<flags> value, in case you want to mask out any backends from a flags
626 value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
628 =back
630 If one or more of the backend flags are or'ed into the flags value,
631 then only these backends will be tried (in the reverse order as listed
632 here). If none are specified, all backends in C<ev_recommended_backends
633 ()> will be tried.
635 Example: Try to create a event loop that uses epoll and nothing else.
637 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
638 if (!epoller)
639 fatal ("no epoll found here, maybe it hides under your chair");
641 Example: Use whatever libev has to offer, but make sure that kqueue is
642 used if available.
644 struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
646 =item ev_loop_destroy (loop)
648 Destroys an event loop object (frees all memory and kernel state
649 etc.). None of the active event watchers will be stopped in the normal
650 sense, so e.g. C<ev_is_active> might still return true. It is your
651 responsibility to either stop all watchers cleanly yourself I<before>
652 calling this function, or cope with the fact afterwards (which is usually
653 the easiest thing, you can just ignore the watchers and/or C<free ()> them
654 for example).
656 Note that certain global state, such as signal state (and installed signal
657 handlers), will not be freed by this function, and related watchers (such
658 as signal and child watchers) would need to be stopped manually.
660 This function is normally used on loop objects allocated by
661 C<ev_loop_new>, but it can also be used on the default loop returned by
662 C<ev_default_loop>, in which case it is not thread-safe.
664 Note that it is not advisable to call this function on the default loop
665 except in the rare occasion where you really need to free its resources.
666 If you need dynamically allocated loops it is better to use C<ev_loop_new>
667 and C<ev_loop_destroy>.
669 =item ev_loop_fork (loop)
671 This function sets a flag that causes subsequent C<ev_run> iterations to
672 reinitialise the kernel state for backends that have one. Despite the
673 name, you can call it anytime, but it makes most sense after forking, in
674 the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
675 child before resuming or calling C<ev_run>.
677 Again, you I<have> to call it on I<any> loop that you want to re-use after
678 a fork, I<even if you do not plan to use the loop in the parent>. This is
679 because some kernel interfaces *cough* I<kqueue> *cough* do funny things
680 during fork.
682 On the other hand, you only need to call this function in the child
683 process if and only if you want to use the event loop in the child. If
684 you just fork+exec or create a new loop in the child, you don't have to
685 call it at all (in fact, C<epoll> is so badly broken that it makes a
686 difference, but libev will usually detect this case on its own and do a
687 costly reset of the backend).
689 The function itself is quite fast and it's usually not a problem to call
690 it just in case after a fork.
692 Example: Automate calling C<ev_loop_fork> on the default loop when
693 using pthreads.
695 static void
696 post_fork_child (void)
697 {
698 ev_loop_fork (EV_DEFAULT);
699 }
701 ...
702 pthread_atfork (0, 0, post_fork_child);
704 =item int ev_is_default_loop (loop)
706 Returns true when the given loop is, in fact, the default loop, and false
707 otherwise.
709 =item unsigned int ev_iteration (loop)
711 Returns the current iteration count for the event loop, which is identical
712 to the number of times libev did poll for new events. It starts at C<0>
713 and happily wraps around with enough iterations.
715 This value can sometimes be useful as a generation counter of sorts (it
716 "ticks" the number of loop iterations), as it roughly corresponds with
717 C<ev_prepare> and C<ev_check> calls - and is incremented between the
718 prepare and check phases.
720 =item unsigned int ev_depth (loop)
722 Returns the number of times C<ev_run> was entered minus the number of
723 times C<ev_run> was exited normally, in other words, the recursion depth.
725 Outside C<ev_run>, this number is zero. In a callback, this number is
726 C<1>, unless C<ev_run> was invoked recursively (or from another thread),
727 in which case it is higher.
729 Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
730 throwing an exception etc.), doesn't count as "exit" - consider this
731 as a hint to avoid such ungentleman-like behaviour unless it's really
732 convenient, in which case it is fully supported.
734 =item unsigned int ev_backend (loop)
736 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
737 use.
739 =item ev_tstamp ev_now (loop)
741 Returns the current "event loop time", which is the time the event loop
742 received events and started processing them. This timestamp does not
743 change as long as callbacks are being processed, and this is also the base
744 time used for relative timers. You can treat it as the timestamp of the
745 event occurring (or more correctly, libev finding out about it).
747 =item ev_now_update (loop)
749 Establishes the current time by querying the kernel, updating the time
750 returned by C<ev_now ()> in the progress. This is a costly operation and
751 is usually done automatically within C<ev_run ()>.
753 This function is rarely useful, but when some event callback runs for a
754 very long time without entering the event loop, updating libev's idea of
755 the current time is a good idea.
757 See also L<The special problem of time updates> in the C<ev_timer> section.
759 =item ev_suspend (loop)
761 =item ev_resume (loop)
763 These two functions suspend and resume an event loop, for use when the
764 loop is not used for a while and timeouts should not be processed.
766 A typical use case would be an interactive program such as a game: When
767 the user presses C<^Z> to suspend the game and resumes it an hour later it
768 would be best to handle timeouts as if no time had actually passed while
769 the program was suspended. This can be achieved by calling C<ev_suspend>
770 in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
771 C<ev_resume> directly afterwards to resume timer processing.
773 Effectively, all C<ev_timer> watchers will be delayed by the time spend
774 between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
775 will be rescheduled (that is, they will lose any events that would have
776 occurred while suspended).
778 After calling C<ev_suspend> you B<must not> call I<any> function on the
779 given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
780 without a previous call to C<ev_suspend>.
782 Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
783 event loop time (see C<ev_now_update>).
785 =item ev_run (loop, int flags)
787 Finally, this is it, the event handler. This function usually is called
788 after you have initialised all your watchers and you want to start
789 handling events. It will ask the operating system for any new events, call
790 the watcher callbacks, an then repeat the whole process indefinitely: This
791 is why event loops are called I<loops>.
793 If the flags argument is specified as C<0>, it will keep handling events
794 until either no event watchers are active anymore or C<ev_break> was
795 called.
797 Please note that an explicit C<ev_break> is usually better than
798 relying on all watchers to be stopped when deciding when a program has
799 finished (especially in interactive programs), but having a program
800 that automatically loops as long as it has to and no longer by virtue
801 of relying on its watchers stopping correctly, that is truly a thing of
802 beauty.
804 This function is also I<mostly> exception-safe - you can break out of
805 a C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
806 exception and so on. This does not decrement the C<ev_depth> value, nor
807 will it clear any outstanding C<EVBREAK_ONE> breaks.
809 A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
810 those events and any already outstanding ones, but will not wait and
811 block your process in case there are no events and will return after one
812 iteration of the loop. This is sometimes useful to poll and handle new
813 events while doing lengthy calculations, to keep the program responsive.
815 A flags value of C<EVRUN_ONCE> will look for new events (waiting if
816 necessary) and will handle those and any already outstanding ones. It
817 will block your process until at least one new event arrives (which could
818 be an event internal to libev itself, so there is no guarantee that a
819 user-registered callback will be called), and will return after one
820 iteration of the loop.
822 This is useful if you are waiting for some external event in conjunction
823 with something not expressible using other libev watchers (i.e. "roll your
824 own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
825 usually a better approach for this kind of thing.
827 Here are the gory details of what C<ev_run> does:
829 - Increment loop depth.
830 - Reset the ev_break status.
831 - Before the first iteration, call any pending watchers.
832 LOOP:
833 - If EVFLAG_FORKCHECK was used, check for a fork.
834 - If a fork was detected (by any means), queue and call all fork watchers.
835 - Queue and call all prepare watchers.
836 - If ev_break was called, goto FINISH.
837 - If we have been forked, detach and recreate the kernel state
838 as to not disturb the other process.
839 - Update the kernel state with all outstanding changes.
840 - Update the "event loop time" (ev_now ()).
841 - Calculate for how long to sleep or block, if at all
842 (active idle watchers, EVRUN_NOWAIT or not having
843 any active watchers at all will result in not sleeping).
844 - Sleep if the I/O and timer collect interval say so.
845 - Increment loop iteration counter.
846 - Block the process, waiting for any events.
847 - Queue all outstanding I/O (fd) events.
848 - Update the "event loop time" (ev_now ()), and do time jump adjustments.
849 - Queue all expired timers.
850 - Queue all expired periodics.
851 - Queue all idle watchers with priority higher than that of pending events.
852 - Queue all check watchers.
853 - Call all queued watchers in reverse order (i.e. check watchers first).
854 Signals and child watchers are implemented as I/O watchers, and will
855 be handled here by queueing them when their watcher gets executed.
856 - If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
857 were used, or there are no active watchers, goto FINISH, otherwise
858 continue with step LOOP.
860 - Reset the ev_break status iff it was EVBREAK_ONE.
861 - Decrement the loop depth.
862 - Return.
864 Example: Queue some jobs and then loop until no events are outstanding
865 anymore.
867 ... queue jobs here, make sure they register event watchers as long
868 ... as they still have work to do (even an idle watcher will do..)
869 ev_run (my_loop, 0);
870 ... jobs done or somebody called unloop. yeah!
872 =item ev_break (loop, how)
874 Can be used to make a call to C<ev_run> return early (but only after it
875 has processed all outstanding events). The C<how> argument must be either
876 C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
877 C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
879 This "break state" will be cleared on the next call to C<ev_run>.
881 It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
882 which case it will have no effect.
884 =item ev_ref (loop)
886 =item ev_unref (loop)
888 Ref/unref can be used to add or remove a reference count on the event
889 loop: Every watcher keeps one reference, and as long as the reference
890 count is nonzero, C<ev_run> will not return on its own.
892 This is useful when you have a watcher that you never intend to
893 unregister, but that nevertheless should not keep C<ev_run> from
894 returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
895 before stopping it.
897 As an example, libev itself uses this for its internal signal pipe: It
898 is not visible to the libev user and should not keep C<ev_run> from
899 exiting if no event watchers registered by it are active. It is also an
900 excellent way to do this for generic recurring timers or from within
901 third-party libraries. Just remember to I<unref after start> and I<ref
902 before stop> (but only if the watcher wasn't active before, or was active
903 before, respectively. Note also that libev might stop watchers itself
904 (e.g. non-repeating timers) in which case you have to C<ev_ref>
905 in the callback).
907 Example: Create a signal watcher, but keep it from keeping C<ev_run>
908 running when nothing else is active.
910 ev_signal exitsig;
911 ev_signal_init (&exitsig, sig_cb, SIGINT);
912 ev_signal_start (loop, &exitsig);
913 ev_unref (loop);
915 Example: For some weird reason, unregister the above signal handler again.
917 ev_ref (loop);
918 ev_signal_stop (loop, &exitsig);
920 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
922 =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
924 These advanced functions influence the time that libev will spend waiting
925 for events. Both time intervals are by default C<0>, meaning that libev
926 will try to invoke timer/periodic callbacks and I/O callbacks with minimum
927 latency.
929 Setting these to a higher value (the C<interval> I<must> be >= C<0>)
930 allows libev to delay invocation of I/O and timer/periodic callbacks
931 to increase efficiency of loop iterations (or to increase power-saving
932 opportunities).
934 The idea is that sometimes your program runs just fast enough to handle
935 one (or very few) event(s) per loop iteration. While this makes the
936 program responsive, it also wastes a lot of CPU time to poll for new
937 events, especially with backends like C<select ()> which have a high
938 overhead for the actual polling but can deliver many events at once.
940 By setting a higher I<io collect interval> you allow libev to spend more
941 time collecting I/O events, so you can handle more events per iteration,
942 at the cost of increasing latency. Timeouts (both C<ev_periodic> and
943 C<ev_timer>) will be not affected. Setting this to a non-null value will
944 introduce an additional C<ev_sleep ()> call into most loop iterations. The
945 sleep time ensures that libev will not poll for I/O events more often then
946 once per this interval, on average.
948 Likewise, by setting a higher I<timeout collect interval> you allow libev
949 to spend more time collecting timeouts, at the expense of increased
950 latency/jitter/inexactness (the watcher callback will be called
951 later). C<ev_io> watchers will not be affected. Setting this to a non-null
952 value will not introduce any overhead in libev.
954 Many (busy) programs can usually benefit by setting the I/O collect
955 interval to a value near C<0.1> or so, which is often enough for
956 interactive servers (of course not for games), likewise for timeouts. It
957 usually doesn't make much sense to set it to a lower value than C<0.01>,
958 as this approaches the timing granularity of most systems. Note that if
959 you do transactions with the outside world and you can't increase the
960 parallelity, then this setting will limit your transaction rate (if you
961 need to poll once per transaction and the I/O collect interval is 0.01,
962 then you can't do more than 100 transactions per second).
964 Setting the I<timeout collect interval> can improve the opportunity for
965 saving power, as the program will "bundle" timer callback invocations that
966 are "near" in time together, by delaying some, thus reducing the number of
967 times the process sleeps and wakes up again. Another useful technique to
968 reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
969 they fire on, say, one-second boundaries only.
971 Example: we only need 0.1s timeout granularity, and we wish not to poll
972 more often than 100 times per second:
974 ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
975 ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
977 =item ev_invoke_pending (loop)
979 This call will simply invoke all pending watchers while resetting their
980 pending state. Normally, C<ev_run> does this automatically when required,
981 but when overriding the invoke callback this call comes handy. This
982 function can be invoked from a watcher - this can be useful for example
983 when you want to do some lengthy calculation and want to pass further
984 event handling to another thread (you still have to make sure only one
985 thread executes within C<ev_invoke_pending> or C<ev_run> of course).
987 =item int ev_pending_count (loop)
989 Returns the number of pending watchers - zero indicates that no watchers
990 are pending.
992 =item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
994 This overrides the invoke pending functionality of the loop: Instead of
995 invoking all pending watchers when there are any, C<ev_run> will call
996 this callback instead. This is useful, for example, when you want to
997 invoke the actual watchers inside another context (another thread etc.).
999 If you want to reset the callback, use C<ev_invoke_pending> as new
1000 callback.
1002 =item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
1004 Sometimes you want to share the same loop between multiple threads. This
1005 can be done relatively simply by putting mutex_lock/unlock calls around
1006 each call to a libev function.
1008 However, C<ev_run> can run an indefinite time, so it is not feasible
1009 to wait for it to return. One way around this is to wake up the event
1010 loop via C<ev_break> and C<av_async_send>, another way is to set these
1011 I<release> and I<acquire> callbacks on the loop.
1013 When set, then C<release> will be called just before the thread is
1014 suspended waiting for new events, and C<acquire> is called just
1015 afterwards.
1017 Ideally, C<release> will just call your mutex_unlock function, and
1018 C<acquire> will just call the mutex_lock function again.
1020 While event loop modifications are allowed between invocations of
1021 C<release> and C<acquire> (that's their only purpose after all), no
1022 modifications done will affect the event loop, i.e. adding watchers will
1023 have no effect on the set of file descriptors being watched, or the time
1024 waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
1025 to take note of any changes you made.
1027 In theory, threads executing C<ev_run> will be async-cancel safe between
1028 invocations of C<release> and C<acquire>.
1030 See also the locking example in the C<THREADS> section later in this
1031 document.
1033 =item ev_set_userdata (loop, void *data)
1035 =item void *ev_userdata (loop)
1037 Set and retrieve a single C<void *> associated with a loop. When
1038 C<ev_set_userdata> has never been called, then C<ev_userdata> returns
1039 C<0>.
1041 These two functions can be used to associate arbitrary data with a loop,
1042 and are intended solely for the C<invoke_pending_cb>, C<release> and
1043 C<acquire> callbacks described above, but of course can be (ab-)used for
1044 any other purpose as well.
1046 =item ev_verify (loop)
1048 This function only does something when C<EV_VERIFY> support has been
1049 compiled in, which is the default for non-minimal builds. It tries to go
1050 through all internal structures and checks them for validity. If anything
1051 is found to be inconsistent, it will print an error message to standard
1052 error and call C<abort ()>.
1054 This can be used to catch bugs inside libev itself: under normal
1055 circumstances, this function will never abort as of course libev keeps its
1056 data structures consistent.
1058 =back
1063 In the following description, uppercase C<TYPE> in names stands for the
1064 watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
1065 watchers and C<ev_io_start> for I/O watchers.
1067 A watcher is an opaque structure that you allocate and register to record
1068 your interest in some event. To make a concrete example, imagine you want
1069 to wait for STDIN to become readable, you would create an C<ev_io> watcher
1070 for that:
1072 static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
1073 {
1074 ev_io_stop (w);
1075 ev_break (loop, EVBREAK_ALL);
1076 }
1078 struct ev_loop *loop = ev_default_loop (0);
1080 ev_io stdin_watcher;
1082 ev_init (&stdin_watcher, my_cb);
1083 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
1084 ev_io_start (loop, &stdin_watcher);
1086 ev_run (loop, 0);
1088 As you can see, you are responsible for allocating the memory for your
1089 watcher structures (and it is I<usually> a bad idea to do this on the
1090 stack).
1092 Each watcher has an associated watcher structure (called C<struct ev_TYPE>
1093 or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
1095 Each watcher structure must be initialised by a call to C<ev_init (watcher
1096 *, callback)>, which expects a callback to be provided. This callback is
1097 invoked each time the event occurs (or, in the case of I/O watchers, each
1098 time the event loop detects that the file descriptor given is readable
1099 and/or writable).
1101 Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
1102 macro to configure it, with arguments specific to the watcher type. There
1103 is also a macro to combine initialisation and setting in one call: C<<
1104 ev_TYPE_init (watcher *, callback, ...) >>.
1106 To make the watcher actually watch out for events, you have to start it
1107 with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
1108 *) >>), and you can stop watching for events at any time by calling the
1109 corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
1111 As long as your watcher is active (has been started but not stopped) you
1112 must not touch the values stored in it. Most specifically you must never
1113 reinitialise it or call its C<ev_TYPE_set> macro.
1115 Each and every callback receives the event loop pointer as first, the
1116 registered watcher structure as second, and a bitset of received events as
1117 third argument.
1119 The received events usually include a single bit per event type received
1120 (you can receive multiple events at the same time). The possible bit masks
1121 are:
1123 =over 4
1125 =item C<EV_READ>
1127 =item C<EV_WRITE>
1129 The file descriptor in the C<ev_io> watcher has become readable and/or
1130 writable.
1132 =item C<EV_TIMER>
1134 The C<ev_timer> watcher has timed out.
1136 =item C<EV_PERIODIC>
1138 The C<ev_periodic> watcher has timed out.
1140 =item C<EV_SIGNAL>
1142 The signal specified in the C<ev_signal> watcher has been received by a thread.
1144 =item C<EV_CHILD>
1146 The pid specified in the C<ev_child> watcher has received a status change.
1148 =item C<EV_STAT>
1150 The path specified in the C<ev_stat> watcher changed its attributes somehow.
1152 =item C<EV_IDLE>
1154 The C<ev_idle> watcher has determined that you have nothing better to do.
1156 =item C<EV_PREPARE>
1158 =item C<EV_CHECK>
1160 All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
1161 to gather new events, and all C<ev_check> watchers are invoked just after
1162 C<ev_run> has gathered them, but before it invokes any callbacks for any
1163 received events. Callbacks of both watcher types can start and stop as
1164 many watchers as they want, and all of them will be taken into account
1165 (for example, a C<ev_prepare> watcher might start an idle watcher to keep
1166 C<ev_run> from blocking).
1168 =item C<EV_EMBED>
1170 The embedded event loop specified in the C<ev_embed> watcher needs attention.
1172 =item C<EV_FORK>
1174 The event loop has been resumed in the child process after fork (see
1175 C<ev_fork>).
1177 =item C<EV_CLEANUP>
1179 The event loop is about to be destroyed (see C<ev_cleanup>).
1181 =item C<EV_ASYNC>
1183 The given async watcher has been asynchronously notified (see C<ev_async>).
1185 =item C<EV_CUSTOM>
1187 Not ever sent (or otherwise used) by libev itself, but can be freely used
1188 by libev users to signal watchers (e.g. via C<ev_feed_event>).
1190 =item C<EV_ERROR>
1192 An unspecified error has occurred, the watcher has been stopped. This might
1193 happen because the watcher could not be properly started because libev
1194 ran out of memory, a file descriptor was found to be closed or any other
1195 problem. Libev considers these application bugs.
1197 You best act on it by reporting the problem and somehow coping with the
1198 watcher being stopped. Note that well-written programs should not receive
1199 an error ever, so when your watcher receives it, this usually indicates a
1200 bug in your program.
1202 Libev will usually signal a few "dummy" events together with an error, for
1203 example it might indicate that a fd is readable or writable, and if your
1204 callbacks is well-written it can just attempt the operation and cope with
1205 the error from read() or write(). This will not work in multi-threaded
1206 programs, though, as the fd could already be closed and reused for another
1207 thing, so beware.
1209 =back
1213 =over 4
1215 =item C<ev_init> (ev_TYPE *watcher, callback)
1217 This macro initialises the generic portion of a watcher. The contents
1218 of the watcher object can be arbitrary (so C<malloc> will do). Only
1219 the generic parts of the watcher are initialised, you I<need> to call
1220 the type-specific C<ev_TYPE_set> macro afterwards to initialise the
1221 type-specific parts. For each type there is also a C<ev_TYPE_init> macro
1222 which rolls both calls into one.
1224 You can reinitialise a watcher at any time as long as it has been stopped
1225 (or never started) and there are no pending events outstanding.
1227 The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
1228 int revents)>.
1230 Example: Initialise an C<ev_io> watcher in two steps.
1232 ev_io w;
1233 ev_init (&w, my_cb);
1234 ev_io_set (&w, STDIN_FILENO, EV_READ);
1236 =item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
1238 This macro initialises the type-specific parts of a watcher. You need to
1239 call C<ev_init> at least once before you call this macro, but you can
1240 call C<ev_TYPE_set> any number of times. You must not, however, call this
1241 macro on a watcher that is active (it can be pending, however, which is a
1242 difference to the C<ev_init> macro).
1244 Although some watcher types do not have type-specific arguments
1245 (e.g. C<ev_prepare>) you still need to call its C<set> macro.
1247 See C<ev_init>, above, for an example.
1249 =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
1251 This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
1252 calls into a single call. This is the most convenient method to initialise
1253 a watcher. The same limitations apply, of course.
1255 Example: Initialise and set an C<ev_io> watcher in one step.
1257 ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
1259 =item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
1261 Starts (activates) the given watcher. Only active watchers will receive
1262 events. If the watcher is already active nothing will happen.
1264 Example: Start the C<ev_io> watcher that is being abused as example in this
1265 whole section.
1267 ev_io_start (EV_DEFAULT_UC, &w);
1269 =item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
1271 Stops the given watcher if active, and clears the pending status (whether
1272 the watcher was active or not).
1274 It is possible that stopped watchers are pending - for example,
1275 non-repeating timers are being stopped when they become pending - but
1276 calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
1277 pending. If you want to free or reuse the memory used by the watcher it is
1278 therefore a good idea to always call its C<ev_TYPE_stop> function.
1280 =item bool ev_is_active (ev_TYPE *watcher)
1282 Returns a true value iff the watcher is active (i.e. it has been started
1283 and not yet been stopped). As long as a watcher is active you must not modify
1284 it.
1286 =item bool ev_is_pending (ev_TYPE *watcher)
1288 Returns a true value iff the watcher is pending, (i.e. it has outstanding
1289 events but its callback has not yet been invoked). As long as a watcher
1290 is pending (but not active) you must not call an init function on it (but
1291 C<ev_TYPE_set> is safe), you must not change its priority, and you must
1292 make sure the watcher is available to libev (e.g. you cannot C<free ()>
1293 it).
1295 =item callback ev_cb (ev_TYPE *watcher)
1297 Returns the callback currently set on the watcher.
1299 =item ev_cb_set (ev_TYPE *watcher, callback)
1301 Change the callback. You can change the callback at virtually any time
1302 (modulo threads).
1304 =item ev_set_priority (ev_TYPE *watcher, int priority)
1306 =item int ev_priority (ev_TYPE *watcher)
1308 Set and query the priority of the watcher. The priority is a small
1309 integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
1310 (default: C<-2>). Pending watchers with higher priority will be invoked
1311 before watchers with lower priority, but priority will not keep watchers
1312 from being executed (except for C<ev_idle> watchers).
1314 If you need to suppress invocation when higher priority events are pending
1315 you need to look at C<ev_idle> watchers, which provide this functionality.
1317 You I<must not> change the priority of a watcher as long as it is active or
1318 pending.
1320 Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
1321 fine, as long as you do not mind that the priority value you query might
1322 or might not have been clamped to the valid range.
1324 The default priority used by watchers when no priority has been set is
1325 always C<0>, which is supposed to not be too high and not be too low :).
1327 See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1328 priorities.
1330 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
1332 Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1333 C<loop> nor C<revents> need to be valid as long as the watcher callback
1334 can deal with that fact, as both are simply passed through to the
1335 callback.
1337 =item int ev_clear_pending (loop, ev_TYPE *watcher)
1339 If the watcher is pending, this function clears its pending status and
1340 returns its C<revents> bitset (as if its callback was invoked). If the
1341 watcher isn't pending it does nothing and returns C<0>.
1343 Sometimes it can be useful to "poll" a watcher instead of waiting for its
1344 callback to be invoked, which can be accomplished with this function.
1346 =item ev_feed_event (loop, ev_TYPE *watcher, int revents)
1348 Feeds the given event set into the event loop, as if the specified event
1349 had happened for the specified watcher (which must be a pointer to an
1350 initialised but not necessarily started event watcher). Obviously you must
1351 not free the watcher as long as it has pending events.
1353 Stopping the watcher, letting libev invoke it, or calling
1354 C<ev_clear_pending> will clear the pending event, even if the watcher was
1355 not started in the first place.
1357 See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1358 functions that do not need a watcher.
1360 =back
1365 =head2 WATCHER STATES
1367 There are various watcher states mentioned throughout this manual -
1368 active, pending and so on. In this section these states and the rules to
1369 transition between them will be described in more detail - and while these
1370 rules might look complicated, they usually do "the right thing".
1372 =over 4
1374 =item initialiased
1376 Before a watcher can be registered with the event looop it has to be
1377 initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1378 C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1380 In this state it is simply some block of memory that is suitable for use
1381 in an event loop. It can be moved around, freed, reused etc. at will.
1383 =item started/running/active
1385 Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1386 property of the event loop, and is actively waiting for events. While in
1387 this state it cannot be accessed (except in a few documented ways), moved,
1388 freed or anything else - the only legal thing is to keep a pointer to it,
1389 and call libev functions on it that are documented to work on active watchers.
1391 =item pending
1393 If a watcher is active and libev determines that an event it is interested
1394 in has occurred (such as a timer expiring), it will become pending. It will
1395 stay in this pending state until either it is stopped or its callback is
1396 about to be invoked, so it is not normally pending inside the watcher
1397 callback.
1399 The watcher might or might not be active while it is pending (for example,
1400 an expired non-repeating timer can be pending but no longer active). If it
1401 is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
1402 but it is still property of the event loop at this time, so cannot be
1403 moved, freed or reused. And if it is active the rules described in the
1404 previous item still apply.
1406 It is also possible to feed an event on a watcher that is not active (e.g.
1407 via C<ev_feed_event>), in which case it becomes pending without being
1408 active.
1410 =item stopped
1412 A watcher can be stopped implicitly by libev (in which case it might still
1413 be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
1414 latter will clear any pending state the watcher might be in, regardless
1415 of whether it was active or not, so stopping a watcher explicitly before
1416 freeing it is often a good idea.
1418 While stopped (and not pending) the watcher is essentially in the
1419 initialised state, that is it can be reused, moved, modified in any way
1420 you wish.
1422 =back
1426 Many event loops support I<watcher priorities>, which are usually small
1427 integers that influence the ordering of event callback invocation
1428 between watchers in some way, all else being equal.
1430 In libev, Watcher priorities can be set using C<ev_set_priority>. See its
1431 description for the more technical details such as the actual priority
1432 range.
1434 There are two common ways how these these priorities are being interpreted
1435 by event loops:
1437 In the more common lock-out model, higher priorities "lock out" invocation
1438 of lower priority watchers, which means as long as higher priority
1439 watchers receive events, lower priority watchers are not being invoked.
1441 The less common only-for-ordering model uses priorities solely to order
1442 callback invocation within a single event loop iteration: Higher priority
1443 watchers are invoked before lower priority ones, but they all get invoked
1444 before polling for new events.
1446 Libev uses the second (only-for-ordering) model for all its watchers
1447 except for idle watchers (which use the lock-out model).
1449 The rationale behind this is that implementing the lock-out model for
1450 watchers is not well supported by most kernel interfaces, and most event
1451 libraries will just poll for the same events again and again as long as
1452 their callbacks have not been executed, which is very inefficient in the
1453 common case of one high-priority watcher locking out a mass of lower
1454 priority ones.
1456 Static (ordering) priorities are most useful when you have two or more
1457 watchers handling the same resource: a typical usage example is having an
1458 C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
1459 timeouts. Under load, data might be received while the program handles
1460 other jobs, but since timers normally get invoked first, the timeout
1461 handler will be executed before checking for data. In that case, giving
1462 the timer a lower priority than the I/O watcher ensures that I/O will be
1463 handled first even under adverse conditions (which is usually, but not
1464 always, what you want).
1466 Since idle watchers use the "lock-out" model, meaning that idle watchers
1467 will only be executed when no same or higher priority watchers have
1468 received events, they can be used to implement the "lock-out" model when
1469 required.
1471 For example, to emulate how many other event libraries handle priorities,
1472 you can associate an C<ev_idle> watcher to each such watcher, and in
1473 the normal watcher callback, you just start the idle watcher. The real
1474 processing is done in the idle watcher callback. This causes libev to
1475 continuously poll and process kernel event data for the watcher, but when
1476 the lock-out case is known to be rare (which in turn is rare :), this is
1477 workable.
1479 Usually, however, the lock-out model implemented that way will perform
1480 miserably under the type of load it was designed to handle. In that case,
1481 it might be preferable to stop the real watcher before starting the
1482 idle watcher, so the kernel will not have to process the event in case
1483 the actual processing will be delayed for considerable time.
1485 Here is an example of an I/O watcher that should run at a strictly lower
1486 priority than the default, and which should only process data when no
1487 other events are pending:
1489 ev_idle idle; // actual processing watcher
1490 ev_io io; // actual event watcher
1492 static void
1493 io_cb (EV_P_ ev_io *w, int revents)
1494 {
1495 // stop the I/O watcher, we received the event, but
1496 // are not yet ready to handle it.
1497 ev_io_stop (EV_A_ w);
1499 // start the idle watcher to handle the actual event.
1500 // it will not be executed as long as other watchers
1501 // with the default priority are receiving events.
1502 ev_idle_start (EV_A_ &idle);
1503 }
1505 static void
1506 idle_cb (EV_P_ ev_idle *w, int revents)
1507 {
1508 // actual processing
1509 read (STDIN_FILENO, ...);
1511 // have to start the I/O watcher again, as
1512 // we have handled the event
1513 ev_io_start (EV_P_ &io);
1514 }
1516 // initialisation
1517 ev_idle_init (&idle, idle_cb);
1518 ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
1519 ev_io_start (EV_DEFAULT_ &io);
1521 In the "real" world, it might also be beneficial to start a timer, so that
1522 low-priority connections can not be locked out forever under load. This
1523 enables your program to keep a lower latency for important connections
1524 during short periods of high load, while not completely locking out less
1525 important ones.
1528 =head1 WATCHER TYPES
1530 This section describes each watcher in detail, but will not repeat
1531 information given in the last section. Any initialisation/set macros,
1532 functions and members specific to the watcher type are explained.
1534 Members are additionally marked with either I<[read-only]>, meaning that,
1535 while the watcher is active, you can look at the member and expect some
1536 sensible content, but you must not modify it (you can modify it while the
1537 watcher is stopped to your hearts content), or I<[read-write]>, which
1538 means you can expect it to have some sensible content while the watcher
1539 is active, but you can also modify it. Modifying it may not do something
1540 sensible or take immediate effect (or do anything at all), but libev will
1541 not crash or malfunction in any way.
1544 =head2 C<ev_io> - is this file descriptor readable or writable?
1546 I/O watchers check whether a file descriptor is readable or writable
1547 in each iteration of the event loop, or, more precisely, when reading
1548 would not block the process and writing would at least be able to write
1549 some data. This behaviour is called level-triggering because you keep
1550 receiving events as long as the condition persists. Remember you can stop
1551 the watcher if you don't want to act on the event and neither want to
1552 receive future events.
1554 In general you can register as many read and/or write event watchers per
1555 fd as you want (as long as you don't confuse yourself). Setting all file
1556 descriptors to non-blocking mode is also usually a good idea (but not
1557 required if you know what you are doing).
1559 Another thing you have to watch out for is that it is quite easy to
1560 receive "spurious" readiness notifications, that is, your callback might
1561 be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1562 because there is no data. It is very easy to get into this situation even
1563 with a relatively standard program structure. Thus it is best to always
1564 use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1565 preferable to a program hanging until some data arrives.
1567 If you cannot run the fd in non-blocking mode (for example you should
1568 not play around with an Xlib connection), then you have to separately
1569 re-test whether a file descriptor is really ready with a known-to-be good
1570 interface such as poll (fortunately in the case of Xlib, it already does
1571 this on its own, so its quite safe to use). Some people additionally
1572 use C<SIGALRM> and an interval timer, just to be sure you won't block
1573 indefinitely.
1575 But really, best use non-blocking mode.
1577 =head3 The special problem of disappearing file descriptors
1579 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1580 descriptor (either due to calling C<close> explicitly or any other means,
1581 such as C<dup2>). The reason is that you register interest in some file
1582 descriptor, but when it goes away, the operating system will silently drop
1583 this interest. If another file descriptor with the same number then is
1584 registered with libev, there is no efficient way to see that this is, in
1585 fact, a different file descriptor.
1587 To avoid having to explicitly tell libev about such cases, libev follows
1588 the following policy: Each time C<ev_io_set> is being called, libev
1589 will assume that this is potentially a new file descriptor, otherwise
1590 it is assumed that the file descriptor stays the same. That means that
1591 you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1592 descriptor even if the file descriptor number itself did not change.
1594 This is how one would do it normally anyway, the important point is that
1595 the libev application should not optimise around libev but should leave
1596 optimisations to libev.
1598 =head3 The special problem of dup'ed file descriptors
1600 Some backends (e.g. epoll), cannot register events for file descriptors,
1601 but only events for the underlying file descriptions. That means when you
1602 have C<dup ()>'ed file descriptors or weirder constellations, and register
1603 events for them, only one file descriptor might actually receive events.
1605 There is no workaround possible except not registering events
1606 for potentially C<dup ()>'ed file descriptors, or to resort to
1609 =head3 The special problem of files
1611 Many people try to use C<select> (or libev) on file descriptors
1612 representing files, and expect it to become ready when their program
1613 doesn't block on disk accesses (which can take a long time on their own).
1615 However, this cannot ever work in the "expected" way - you get a readiness
1616 notification as soon as the kernel knows whether and how much data is
1617 there, and in the case of open files, that's always the case, so you
1618 always get a readiness notification instantly, and your read (or possibly
1619 write) will still block on the disk I/O.
1621 Another way to view it is that in the case of sockets, pipes, character
1622 devices and so on, there is another party (the sender) that delivers data
1623 on its own, but in the case of files, there is no such thing: the disk
1624 will not send data on its own, simply because it doesn't know what you
1625 wish to read - you would first have to request some data.
1627 Since files are typically not-so-well supported by advanced notification
1628 mechanism, libev tries hard to emulate POSIX behaviour with respect
1629 to files, even though you should not use it. The reason for this is
1630 convenience: sometimes you want to watch STDIN or STDOUT, which is
1631 usually a tty, often a pipe, but also sometimes files or special devices
1632 (for example, C<epoll> on Linux works with F</dev/random> but not with
1633 F</dev/urandom>), and even though the file might better be served with
1634 asynchronous I/O instead of with non-blocking I/O, it is still useful when
1635 it "just works" instead of freezing.
1637 So avoid file descriptors pointing to files when you know it (e.g. use
1638 libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
1639 when you rarely read from a file instead of from a socket, and want to
1640 reuse the same code path.
1642 =head3 The special problem of fork
1644 Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1645 useless behaviour. Libev fully supports fork, but needs to be told about
1646 it in the child if you want to continue to use it in the child.
1648 To support fork in your child processes, you have to call C<ev_loop_fork
1649 ()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1652 =head3 The special problem of SIGPIPE
1654 While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1655 when writing to a pipe whose other end has been closed, your program gets
1656 sent a SIGPIPE, which, by default, aborts your program. For most programs
1657 this is sensible behaviour, for daemons, this is usually undesirable.
1659 So when you encounter spurious, unexplained daemon exits, make sure you
1660 ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1661 somewhere, as that would have given you a big clue).
1663 =head3 The special problem of accept()ing when you can't
1665 Many implementations of the POSIX C<accept> function (for example,
1666 found in post-2004 Linux) have the peculiar behaviour of not removing a
1667 connection from the pending queue in all error cases.
1669 For example, larger servers often run out of file descriptors (because
1670 of resource limits), causing C<accept> to fail with C<ENFILE> but not
1671 rejecting the connection, leading to libev signalling readiness on
1672 the next iteration again (the connection still exists after all), and
1673 typically causing the program to loop at 100% CPU usage.
1675 Unfortunately, the set of errors that cause this issue differs between
1676 operating systems, there is usually little the app can do to remedy the
1677 situation, and no known thread-safe method of removing the connection to
1678 cope with overload is known (to me).
1680 One of the easiest ways to handle this situation is to just ignore it
1681 - when the program encounters an overload, it will just loop until the
1682 situation is over. While this is a form of busy waiting, no OS offers an
1683 event-based way to handle this situation, so it's the best one can do.
1685 A better way to handle the situation is to log any errors other than
1686 C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
1687 messages, and continue as usual, which at least gives the user an idea of
1688 what could be wrong ("raise the ulimit!"). For extra points one could stop
1689 the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
1690 usage.
1692 If your program is single-threaded, then you could also keep a dummy file
1693 descriptor for overload situations (e.g. by opening F</dev/null>), and
1694 when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
1695 close that fd, and create a new dummy fd. This will gracefully refuse
1696 clients under typical overload conditions.
1698 The last way to handle it is to simply log the error and C<exit>, as
1699 is often done with C<malloc> failures, but this results in an easy
1700 opportunity for a DoS attack.
1702 =head3 Watcher-Specific Functions
1704 =over 4
1706 =item ev_io_init (ev_io *, callback, int fd, int events)
1708 =item ev_io_set (ev_io *, int fd, int events)
1710 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1711 receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
1712 C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
1714 =item int fd [read-only]
1716 The file descriptor being watched.
1718 =item int events [read-only]
1720 The events being watched.
1722 =back
1724 =head3 Examples
1726 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1727 readable, but only once. Since it is likely line-buffered, you could
1728 attempt to read a whole line in the callback.
1730 static void
1731 stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
1732 {
1733 ev_io_stop (loop, w);
1734 .. read from stdin here (or from w->fd) and handle any I/O errors
1735 }
1737 ...
1738 struct ev_loop *loop = ev_default_init (0);
1739 ev_io stdin_readable;
1740 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1741 ev_io_start (loop, &stdin_readable);
1742 ev_run (loop, 0);
1745 =head2 C<ev_timer> - relative and optionally repeating timeouts
1747 Timer watchers are simple relative timers that generate an event after a
1748 given time, and optionally repeating in regular intervals after that.
1750 The timers are based on real time, that is, if you register an event that
1751 times out after an hour and you reset your system clock to January last
1752 year, it will still time out after (roughly) one hour. "Roughly" because
1753 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1754 monotonic clock option helps a lot here).
1756 The callback is guaranteed to be invoked only I<after> its timeout has
1757 passed (not I<at>, so on systems with very low-resolution clocks this
1758 might introduce a small delay). If multiple timers become ready during the
1759 same loop iteration then the ones with earlier time-out values are invoked
1760 before ones of the same priority with later time-out values (but this is
1761 no longer true when a callback calls C<ev_run> recursively).
1763 =head3 Be smart about timeouts
1765 Many real-world problems involve some kind of timeout, usually for error
1766 recovery. A typical example is an HTTP request - if the other side hangs,
1767 you want to raise some error after a while.
1769 What follows are some ways to handle this problem, from obvious and
1770 inefficient to smart and efficient.
1772 In the following, a 60 second activity timeout is assumed - a timeout that
1773 gets reset to 60 seconds each time there is activity (e.g. each time some
1774 data or other life sign was received).
1776 =over 4
1778 =item 1. Use a timer and stop, reinitialise and start it on activity.
1780 This is the most obvious, but not the most simple way: In the beginning,
1781 start the watcher:
1783 ev_timer_init (timer, callback, 60., 0.);
1784 ev_timer_start (loop, timer);
1786 Then, each time there is some activity, C<ev_timer_stop> it, initialise it
1787 and start it again:
1789 ev_timer_stop (loop, timer);
1790 ev_timer_set (timer, 60., 0.);
1791 ev_timer_start (loop, timer);
1793 This is relatively simple to implement, but means that each time there is
1794 some activity, libev will first have to remove the timer from its internal
1795 data structure and then add it again. Libev tries to be fast, but it's
1796 still not a constant-time operation.
1798 =item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
1800 This is the easiest way, and involves using C<ev_timer_again> instead of
1801 C<ev_timer_start>.
1803 To implement this, configure an C<ev_timer> with a C<repeat> value
1804 of C<60> and then call C<ev_timer_again> at start and each time you
1805 successfully read or write some data. If you go into an idle state where
1806 you do not expect data to travel on the socket, you can C<ev_timer_stop>
1807 the timer, and C<ev_timer_again> will automatically restart it if need be.
1809 That means you can ignore both the C<ev_timer_start> function and the
1810 C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
1811 member and C<ev_timer_again>.
1813 At start:
1815 ev_init (timer, callback);
1816 timer->repeat = 60.;
1817 ev_timer_again (loop, timer);
1819 Each time there is some activity:
1821 ev_timer_again (loop, timer);
1823 It is even possible to change the time-out on the fly, regardless of
1824 whether the watcher is active or not:
1826 timer->repeat = 30.;
1827 ev_timer_again (loop, timer);
1829 This is slightly more efficient then stopping/starting the timer each time
1830 you want to modify its timeout value, as libev does not have to completely
1831 remove and re-insert the timer from/into its internal data structure.
1833 It is, however, even simpler than the "obvious" way to do it.
1835 =item 3. Let the timer time out, but then re-arm it as required.
1837 This method is more tricky, but usually most efficient: Most timeouts are
1838 relatively long compared to the intervals between other activity - in
1839 our example, within 60 seconds, there are usually many I/O events with
1840 associated activity resets.
1842 In this case, it would be more efficient to leave the C<ev_timer> alone,
1843 but remember the time of last activity, and check for a real timeout only
1844 within the callback:
1846 ev_tstamp last_activity; // time of last activity
1848 static void
1849 callback (EV_P_ ev_timer *w, int revents)
1850 {
1851 ev_tstamp now = ev_now (EV_A);
1852 ev_tstamp timeout = last_activity + 60.;
1854 // if last_activity + 60. is older than now, we did time out
1855 if (timeout < now)
1856 {
1857 // timeout occurred, take action
1858 }
1859 else
1860 {
1861 // callback was invoked, but there was some activity, re-arm
1862 // the watcher to fire in last_activity + 60, which is
1863 // guaranteed to be in the future, so "again" is positive:
1864 w->repeat = timeout - now;
1865 ev_timer_again (EV_A_ w);
1866 }
1867 }
1869 To summarise the callback: first calculate the real timeout (defined
1870 as "60 seconds after the last activity"), then check if that time has
1871 been reached, which means something I<did>, in fact, time out. Otherwise
1872 the callback was invoked too early (C<timeout> is in the future), so
1873 re-schedule the timer to fire at that future time, to see if maybe we have
1874 a timeout then.
1876 Note how C<ev_timer_again> is used, taking advantage of the
1877 C<ev_timer_again> optimisation when the timer is already running.
1879 This scheme causes more callback invocations (about one every 60 seconds
1880 minus half the average time between activity), but virtually no calls to
1881 libev to change the timeout.
1883 To start the timer, simply initialise the watcher and set C<last_activity>
1884 to the current time (meaning we just have some activity :), then call the
1885 callback, which will "do the right thing" and start the timer:
1887 ev_init (timer, callback);
1888 last_activity = ev_now (loop);
1889 callback (loop, timer, EV_TIMER);
1891 And when there is some activity, simply store the current time in
1892 C<last_activity>, no libev calls at all:
1894 last_activity = ev_now (loop);
1896 This technique is slightly more complex, but in most cases where the
1897 time-out is unlikely to be triggered, much more efficient.
1899 Changing the timeout is trivial as well (if it isn't hard-coded in the
1900 callback :) - just change the timeout and invoke the callback, which will
1901 fix things for you.
1903 =item 4. Wee, just use a double-linked list for your timeouts.
1905 If there is not one request, but many thousands (millions...), all
1906 employing some kind of timeout with the same timeout value, then one can
1907 do even better:
1909 When starting the timeout, calculate the timeout value and put the timeout
1910 at the I<end> of the list.
1912 Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
1913 the list is expected to fire (for example, using the technique #3).
1915 When there is some activity, remove the timer from the list, recalculate
1916 the timeout, append it to the end of the list again, and make sure to
1917 update the C<ev_timer> if it was taken from the beginning of the list.
1919 This way, one can manage an unlimited number of timeouts in O(1) time for
1920 starting, stopping and updating the timers, at the expense of a major
1921 complication, and having to use a constant timeout. The constant timeout
1922 ensures that the list stays sorted.
1924 =back
1926 So which method the best?
1928 Method #2 is a simple no-brain-required solution that is adequate in most
1929 situations. Method #3 requires a bit more thinking, but handles many cases
1930 better, and isn't very complicated either. In most case, choosing either
1931 one is fine, with #3 being better in typical situations.
1933 Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1934 rather complicated, but extremely efficient, something that really pays
1935 off after the first million or so of active timers, i.e. it's usually
1936 overkill :)
1938 =head3 The special problem of time updates
1940 Establishing the current time is a costly operation (it usually takes at
1941 least two system calls): EV therefore updates its idea of the current
1942 time only before and after C<ev_run> collects new events, which causes a
1943 growing difference between C<ev_now ()> and C<ev_time ()> when handling
1944 lots of events in one iteration.
1946 The relative timeouts are calculated relative to the C<ev_now ()>
1947 time. This is usually the right thing as this timestamp refers to the time
1948 of the event triggering whatever timeout you are modifying/starting. If
1949 you suspect event processing to be delayed and you I<need> to base the
1950 timeout on the current time, use something like this to adjust for this:
1952 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1954 If the event loop is suspended for a long time, you can also force an
1955 update of the time returned by C<ev_now ()> by calling C<ev_now_update
1956 ()>.
1958 =head3 The special problems of suspended animation
1960 When you leave the server world it is quite customary to hit machines that
1961 can suspend/hibernate - what happens to the clocks during such a suspend?
1963 Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
1964 all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
1965 to run until the system is suspended, but they will not advance while the
1966 system is suspended. That means, on resume, it will be as if the program
1967 was frozen for a few seconds, but the suspend time will not be counted
1968 towards C<ev_timer> when a monotonic clock source is used. The real time
1969 clock advanced as expected, but if it is used as sole clocksource, then a
1970 long suspend would be detected as a time jump by libev, and timers would
1971 be adjusted accordingly.
1973 I would not be surprised to see different behaviour in different between
1974 operating systems, OS versions or even different hardware.
1976 The other form of suspend (job control, or sending a SIGSTOP) will see a
1977 time jump in the monotonic clocks and the realtime clock. If the program
1978 is suspended for a very long time, and monotonic clock sources are in use,
1979 then you can expect C<ev_timer>s to expire as the full suspension time
1980 will be counted towards the timers. When no monotonic clock source is in
1981 use, then libev will again assume a timejump and adjust accordingly.
1983 It might be beneficial for this latter case to call C<ev_suspend>
1984 and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
1985 deterministic behaviour in this case (you can do nothing against
1986 C<SIGSTOP>).
1988 =head3 Watcher-Specific Functions and Data Members
1990 =over 4
1992 =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1994 =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1996 Configure the timer to trigger after C<after> seconds. If C<repeat>
1997 is C<0.>, then it will automatically be stopped once the timeout is
1998 reached. If it is positive, then the timer will automatically be
1999 configured to trigger again C<repeat> seconds later, again, and again,
2000 until stopped manually.
2002 The timer itself will do a best-effort at avoiding drift, that is, if
2003 you configure a timer to trigger every 10 seconds, then it will normally
2004 trigger at exactly 10 second intervals. If, however, your program cannot
2005 keep up with the timer (because it takes longer than those 10 seconds to
2006 do stuff) the timer will not fire more than once per event loop iteration.
2008 =item ev_timer_again (loop, ev_timer *)
2010 This will act as if the timer timed out and restart it again if it is
2011 repeating. The exact semantics are:
2013 If the timer is pending, its pending status is cleared.
2015 If the timer is started but non-repeating, stop it (as if it timed out).
2017 If the timer is repeating, either start it if necessary (with the
2018 C<repeat> value), or reset the running timer to the C<repeat> value.
2020 This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
2021 usage example.
2023 =item ev_tstamp ev_timer_remaining (loop, ev_timer *)
2025 Returns the remaining time until a timer fires. If the timer is active,
2026 then this time is relative to the current event loop time, otherwise it's
2027 the timeout value currently configured.
2029 That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
2030 C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
2031 will return C<4>. When the timer expires and is restarted, it will return
2032 roughly C<7> (likely slightly less as callback invocation takes some time,
2033 too), and so on.
2035 =item ev_tstamp repeat [read-write]
2037 The current C<repeat> value. Will be used each time the watcher times out
2038 or C<ev_timer_again> is called, and determines the next timeout (if any),
2039 which is also when any modifications are taken into account.
2041 =back
2043 =head3 Examples
2045 Example: Create a timer that fires after 60 seconds.
2047 static void
2048 one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
2049 {
2050 .. one minute over, w is actually stopped right here
2051 }
2053 ev_timer mytimer;
2054 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
2055 ev_timer_start (loop, &mytimer);
2057 Example: Create a timeout timer that times out after 10 seconds of
2058 inactivity.
2060 static void
2061 timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
2062 {
2063 .. ten seconds without any activity
2064 }
2066 ev_timer mytimer;
2067 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
2068 ev_timer_again (&mytimer); /* start timer */
2069 ev_run (loop, 0);
2071 // and in some piece of code that gets executed on any "activity":
2072 // reset the timeout to start ticking again at 10 seconds
2073 ev_timer_again (&mytimer);
2076 =head2 C<ev_periodic> - to cron or not to cron?
2078 Periodic watchers are also timers of a kind, but they are very versatile
2079 (and unfortunately a bit complex).
2081 Unlike C<ev_timer>, periodic watchers are not based on real time (or
2082 relative time, the physical time that passes) but on wall clock time
2083 (absolute time, the thing you can read on your calender or clock). The
2084 difference is that wall clock time can run faster or slower than real
2085 time, and time jumps are not uncommon (e.g. when you adjust your
2086 wrist-watch).
2088 You can tell a periodic watcher to trigger after some specific point
2089 in time: for example, if you tell a periodic watcher to trigger "in 10
2090 seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
2091 not a delay) and then reset your system clock to January of the previous
2092 year, then it will take a year or more to trigger the event (unlike an
2093 C<ev_timer>, which would still trigger roughly 10 seconds after starting
2094 it, as it uses a relative timeout).
2096 C<ev_periodic> watchers can also be used to implement vastly more complex
2097 timers, such as triggering an event on each "midnight, local time", or
2098 other complicated rules. This cannot be done with C<ev_timer> watchers, as
2099 those cannot react to time jumps.
2101 As with timers, the callback is guaranteed to be invoked only when the
2102 point in time where it is supposed to trigger has passed. If multiple
2103 timers become ready during the same loop iteration then the ones with
2104 earlier time-out values are invoked before ones with later time-out values
2105 (but this is no longer true when a callback calls C<ev_run> recursively).
2107 =head3 Watcher-Specific Functions and Data Members
2109 =over 4
2111 =item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
2113 =item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
2115 Lots of arguments, let's sort it out... There are basically three modes of
2116 operation, and we will explain them from simplest to most complex:
2118 =over 4
2120 =item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
2122 In this configuration the watcher triggers an event after the wall clock
2123 time C<offset> has passed. It will not repeat and will not adjust when a
2124 time jump occurs, that is, if it is to be run at January 1st 2011 then it
2125 will be stopped and invoked when the system clock reaches or surpasses
2126 this point in time.
2128 =item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
2130 In this mode the watcher will always be scheduled to time out at the next
2131 C<offset + N * interval> time (for some integer N, which can also be
2132 negative) and then repeat, regardless of any time jumps. The C<offset>
2133 argument is merely an offset into the C<interval> periods.
2135 This can be used to create timers that do not drift with respect to the
2136 system clock, for example, here is an C<ev_periodic> that triggers each
2137 hour, on the hour (with respect to UTC):
2139 ev_periodic_set (&periodic, 0., 3600., 0);
2141 This doesn't mean there will always be 3600 seconds in between triggers,
2142 but only that the callback will be called when the system time shows a
2143 full hour (UTC), or more correctly, when the system time is evenly divisible
2144 by 3600.
2146 Another way to think about it (for the mathematically inclined) is that
2147 C<ev_periodic> will try to run the callback in this mode at the next possible
2148 time where C<time = offset (mod interval)>, regardless of any time jumps.
2150 For numerical stability it is preferable that the C<offset> value is near
2151 C<ev_now ()> (the current time), but there is no range requirement for
2152 this value, and in fact is often specified as zero.
2154 Note also that there is an upper limit to how often a timer can fire (CPU
2155 speed for example), so if C<interval> is very small then timing stability
2156 will of course deteriorate. Libev itself tries to be exact to be about one
2157 millisecond (if the OS supports it and the machine is fast enough).
2159 =item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
2161 In this mode the values for C<interval> and C<offset> are both being
2162 ignored. Instead, each time the periodic watcher gets scheduled, the
2163 reschedule callback will be called with the watcher as first, and the
2164 current time as second argument.
2166 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
2167 or make ANY other event loop modifications whatsoever, unless explicitly
2168 allowed by documentation here>.
2170 If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
2171 it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
2172 only event loop modification you are allowed to do).
2174 The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
2175 *w, ev_tstamp now)>, e.g.:
2177 static ev_tstamp
2178 my_rescheduler (ev_periodic *w, ev_tstamp now)
2179 {
2180 return now + 60.;
2181 }
2183 It must return the next time to trigger, based on the passed time value
2184 (that is, the lowest time value larger than to the second argument). It
2185 will usually be called just before the callback will be triggered, but
2186 might be called at other times, too.
2188 NOTE: I<< This callback must always return a time that is higher than or
2189 equal to the passed C<now> value >>.
2191 This can be used to create very complex timers, such as a timer that
2192 triggers on "next midnight, local time". To do this, you would calculate the
2193 next midnight after C<now> and return the timestamp value for this. How
2194 you do this is, again, up to you (but it is not trivial, which is the main
2195 reason I omitted it as an example).
2197 =back
2199 =item ev_periodic_again (loop, ev_periodic *)
2201 Simply stops and restarts the periodic watcher again. This is only useful
2202 when you changed some parameters or the reschedule callback would return
2203 a different time than the last time it was called (e.g. in a crond like
2204 program when the crontabs have changed).
2206 =item ev_tstamp ev_periodic_at (ev_periodic *)
2208 When active, returns the absolute time that the watcher is supposed
2209 to trigger next. This is not the same as the C<offset> argument to
2210 C<ev_periodic_set>, but indeed works even in interval and manual
2211 rescheduling modes.
2213 =item ev_tstamp offset [read-write]
2215 When repeating, this contains the offset value, otherwise this is the
2216 absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
2217 although libev might modify this value for better numerical stability).
2219 Can be modified any time, but changes only take effect when the periodic
2220 timer fires or C<ev_periodic_again> is being called.
2222 =item ev_tstamp interval [read-write]
2224 The current interval value. Can be modified any time, but changes only
2225 take effect when the periodic timer fires or C<ev_periodic_again> is being
2226 called.
2228 =item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
2230 The current reschedule callback, or C<0>, if this functionality is
2231 switched off. Can be changed any time, but changes only take effect when
2232 the periodic timer fires or C<ev_periodic_again> is being called.
2234 =back
2236 =head3 Examples
2238 Example: Call a callback every hour, or, more precisely, whenever the
2239 system time is divisible by 3600. The callback invocation times have
2240 potentially a lot of jitter, but good long-term stability.
2242 static void
2243 clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
2244 {
2245 ... its now a full hour (UTC, or TAI or whatever your clock follows)
2246 }
2248 ev_periodic hourly_tick;
2249 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
2250 ev_periodic_start (loop, &hourly_tick);
2252 Example: The same as above, but use a reschedule callback to do it:
2254 #include <math.h>
2256 static ev_tstamp
2257 my_scheduler_cb (ev_periodic *w, ev_tstamp now)
2258 {
2259 return now + (3600. - fmod (now, 3600.));
2260 }
2262 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
2264 Example: Call a callback every hour, starting now:
2266 ev_periodic hourly_tick;
2267 ev_periodic_init (&hourly_tick, clock_cb,
2268 fmod (ev_now (loop), 3600.), 3600., 0);
2269 ev_periodic_start (loop, &hourly_tick);
2272 =head2 C<ev_signal> - signal me when a signal gets signalled!
2274 Signal watchers will trigger an event when the process receives a specific
2275 signal one or more times. Even though signals are very asynchronous, libev
2276 will try its best to deliver signals synchronously, i.e. as part of the
2277 normal event processing, like any other event.
2279 If you want signals to be delivered truly asynchronously, just use
2280 C<sigaction> as you would do without libev and forget about sharing
2281 the signal. You can even use C<ev_async> from a signal handler to
2282 synchronously wake up an event loop.
2284 You can configure as many watchers as you like for the same signal, but
2285 only within the same loop, i.e. you can watch for C<SIGINT> in your
2286 default loop and for C<SIGIO> in another loop, but you cannot watch for
2287 C<SIGINT> in both the default loop and another loop at the same time. At
2288 the moment, C<SIGCHLD> is permanently tied to the default loop.
2290 When the first watcher gets started will libev actually register something
2291 with the kernel (thus it coexists with your own signal handlers as long as
2292 you don't register any with libev for the same signal).
2294 If possible and supported, libev will install its handlers with
2295 C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
2296 not be unduly interrupted. If you have a problem with system calls getting
2297 interrupted by signals you can block all signals in an C<ev_check> watcher
2298 and unblock them in an C<ev_prepare> watcher.
2300 =head3 The special problem of inheritance over fork/execve/pthread_create
2302 Both the signal mask (C<sigprocmask>) and the signal disposition
2303 (C<sigaction>) are unspecified after starting a signal watcher (and after
2304 stopping it again), that is, libev might or might not block the signal,
2305 and might or might not set or restore the installed signal handler.
2307 While this does not matter for the signal disposition (libev never
2308 sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2309 C<execve>), this matters for the signal mask: many programs do not expect
2310 certain signals to be blocked.
2312 This means that before calling C<exec> (from the child) you should reset
2313 the signal mask to whatever "default" you expect (all clear is a good
2314 choice usually).
2316 The simplest way to ensure that the signal mask is reset in the child is
2317 to install a fork handler with C<pthread_atfork> that resets it. That will
2318 catch fork calls done by libraries (such as the libc) as well.
2320 In current versions of libev, the signal will not be blocked indefinitely
2321 unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
2322 the window of opportunity for problems, it will not go away, as libev
2323 I<has> to modify the signal mask, at least temporarily.
2325 So I can't stress this enough: I<If you do not reset your signal mask when
2326 you expect it to be empty, you have a race condition in your code>. This
2327 is not a libev-specific thing, this is true for most event libraries.
2329 =head3 The special problem of threads signal handling
2331 POSIX threads has problematic signal handling semantics, specifically,
2332 a lot of functionality (sigfd, sigwait etc.) only really works if all
2333 threads in a process block signals, which is hard to achieve.
2335 When you want to use sigwait (or mix libev signal handling with your own
2336 for the same signals), you can tackle this problem by globally blocking
2337 all signals before creating any threads (or creating them with a fully set
2338 sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
2339 loops. Then designate one thread as "signal receiver thread" which handles
2340 these signals. You can pass on any signals that libev might be interested
2341 in by calling C<ev_feed_signal>.
2343 =head3 Watcher-Specific Functions and Data Members
2345 =over 4
2347 =item ev_signal_init (ev_signal *, callback, int signum)
2349 =item ev_signal_set (ev_signal *, int signum)
2351 Configures the watcher to trigger on the given signal number (usually one
2352 of the C<SIGxxx> constants).
2354 =item int signum [read-only]
2356 The signal the watcher watches out for.
2358 =back
2360 =head3 Examples
2362 Example: Try to exit cleanly on SIGINT.
2364 static void
2365 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
2366 {
2367 ev_break (loop, EVBREAK_ALL);
2368 }
2370 ev_signal signal_watcher;
2371 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
2372 ev_signal_start (loop, &signal_watcher);
2375 =head2 C<ev_child> - watch out for process status changes
2377 Child watchers trigger when your process receives a SIGCHLD in response to
2378 some child status changes (most typically when a child of yours dies or
2379 exits). It is permissible to install a child watcher I<after> the child
2380 has been forked (which implies it might have already exited), as long
2381 as the event loop isn't entered (or is continued from a watcher), i.e.,
2382 forking and then immediately registering a watcher for the child is fine,
2383 but forking and registering a watcher a few event loop iterations later or
2384 in the next callback invocation is not.
2386 Only the default event loop is capable of handling signals, and therefore
2387 you can only register child watchers in the default event loop.
2389 Due to some design glitches inside libev, child watchers will always be
2390 handled at maximum priority (their priority is set to C<EV_MAXPRI> by
2391 libev)
2393 =head3 Process Interaction
2395 Libev grabs C<SIGCHLD> as soon as the default event loop is
2396 initialised. This is necessary to guarantee proper behaviour even if the
2397 first child watcher is started after the child exits. The occurrence
2398 of C<SIGCHLD> is recorded asynchronously, but child reaping is done
2399 synchronously as part of the event loop processing. Libev always reaps all
2400 children, even ones not watched.
2402 =head3 Overriding the Built-In Processing
2404 Libev offers no special support for overriding the built-in child
2405 processing, but if your application collides with libev's default child
2406 handler, you can override it easily by installing your own handler for
2407 C<SIGCHLD> after initialising the default loop, and making sure the
2408 default loop never gets destroyed. You are encouraged, however, to use an
2409 event-based approach to child reaping and thus use libev's support for
2410 that, so other libev users can use C<ev_child> watchers freely.
2412 =head3 Stopping the Child Watcher
2414 Currently, the child watcher never gets stopped, even when the
2415 child terminates, so normally one needs to stop the watcher in the
2416 callback. Future versions of libev might stop the watcher automatically
2417 when a child exit is detected (calling C<ev_child_stop> twice is not a
2418 problem).
2420 =head3 Watcher-Specific Functions and Data Members
2422 =over 4
2424 =item ev_child_init (ev_child *, callback, int pid, int trace)
2426 =item ev_child_set (ev_child *, int pid, int trace)
2428 Configures the watcher to wait for status changes of process C<pid> (or
2429 I<any> process if C<pid> is specified as C<0>). The callback can look
2430 at the C<rstatus> member of the C<ev_child> watcher structure to see
2431 the status word (use the macros from C<sys/wait.h> and see your systems
2432 C<waitpid> documentation). The C<rpid> member contains the pid of the
2433 process causing the status change. C<trace> must be either C<0> (only
2434 activate the watcher when the process terminates) or C<1> (additionally
2435 activate the watcher when the process is stopped or continued).
2437 =item int pid [read-only]
2439 The process id this watcher watches out for, or C<0>, meaning any process id.
2441 =item int rpid [read-write]
2443 The process id that detected a status change.
2445 =item int rstatus [read-write]
2447 The process exit/trace status caused by C<rpid> (see your systems
2448 C<waitpid> and C<sys/wait.h> documentation for details).
2450 =back
2452 =head3 Examples
2454 Example: C<fork()> a new process and install a child handler to wait for
2455 its completion.
2457 ev_child cw;
2459 static void
2460 child_cb (EV_P_ ev_child *w, int revents)
2461 {
2462 ev_child_stop (EV_A_ w);
2463 printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
2464 }
2466 pid_t pid = fork ();
2468 if (pid < 0)
2469 // error
2470 else if (pid == 0)
2471 {
2472 // the forked child executes here
2473 exit (1);
2474 }
2475 else
2476 {
2477 ev_child_init (&cw, child_cb, pid, 0);
2478 ev_child_start (EV_DEFAULT_ &cw);
2479 }
2482 =head2 C<ev_stat> - did the file attributes just change?
2484 This watches a file system path for attribute changes. That is, it calls
2485 C<stat> on that path in regular intervals (or when the OS says it changed)
2486 and sees if it changed compared to the last time, invoking the callback if
2487 it did.
2489 The path does not need to exist: changing from "path exists" to "path does
2490 not exist" is a status change like any other. The condition "path does not
2491 exist" (or more correctly "path cannot be stat'ed") is signified by the
2492 C<st_nlink> field being zero (which is otherwise always forced to be at
2493 least one) and all the other fields of the stat buffer having unspecified
2494 contents.
2496 The path I<must not> end in a slash or contain special components such as
2497 C<.> or C<..>. The path I<should> be absolute: If it is relative and
2498 your working directory changes, then the behaviour is undefined.
2500 Since there is no portable change notification interface available, the
2501 portable implementation simply calls C<stat(2)> regularly on the path
2502 to see if it changed somehow. You can specify a recommended polling
2503 interval for this case. If you specify a polling interval of C<0> (highly
2504 recommended!) then a I<suitable, unspecified default> value will be used
2505 (which you can expect to be around five seconds, although this might
2506 change dynamically). Libev will also impose a minimum interval which is
2507 currently around C<0.1>, but that's usually overkill.
2509 This watcher type is not meant for massive numbers of stat watchers,
2510 as even with OS-supported change notifications, this can be
2511 resource-intensive.
2513 At the time of this writing, the only OS-specific interface implemented
2514 is the Linux inotify interface (implementing kqueue support is left as an
2515 exercise for the reader. Note, however, that the author sees no way of
2516 implementing C<ev_stat> semantics with kqueue, except as a hint).
2518 =head3 ABI Issues (Largefile Support)
2520 Libev by default (unless the user overrides this) uses the default
2521 compilation environment, which means that on systems with large file
2522 support disabled by default, you get the 32 bit version of the stat
2523 structure. When using the library from programs that change the ABI to
2524 use 64 bit file offsets the programs will fail. In that case you have to
2525 compile libev with the same flags to get binary compatibility. This is
2526 obviously the case with any flags that change the ABI, but the problem is
2527 most noticeably displayed with ev_stat and large file support.
2529 The solution for this is to lobby your distribution maker to make large
2530 file interfaces available by default (as e.g. FreeBSD does) and not
2531 optional. Libev cannot simply switch on large file support because it has
2532 to exchange stat structures with application programs compiled using the
2533 default compilation environment.
2535 =head3 Inotify and Kqueue
2537 When C<inotify (7)> support has been compiled into libev and present at
2538 runtime, it will be used to speed up change detection where possible. The
2539 inotify descriptor will be created lazily when the first C<ev_stat>
2540 watcher is being started.
2542 Inotify presence does not change the semantics of C<ev_stat> watchers
2543 except that changes might be detected earlier, and in some cases, to avoid
2544 making regular C<stat> calls. Even in the presence of inotify support
2545 there are many cases where libev has to resort to regular C<stat> polling,
2546 but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
2547 many bugs), the path exists (i.e. stat succeeds), and the path resides on
2548 a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
2549 xfs are fully working) libev usually gets away without polling.
2551 There is no support for kqueue, as apparently it cannot be used to
2552 implement this functionality, due to the requirement of having a file
2553 descriptor open on the object at all times, and detecting renames, unlinks
2554 etc. is difficult.
2556 =head3 C<stat ()> is a synchronous operation
2558 Libev doesn't normally do any kind of I/O itself, and so is not blocking
2559 the process. The exception are C<ev_stat> watchers - those call C<stat
2560 ()>, which is a synchronous operation.
2562 For local paths, this usually doesn't matter: unless the system is very
2563 busy or the intervals between stat's are large, a stat call will be fast,
2564 as the path data is usually in memory already (except when starting the
2565 watcher).
2567 For networked file systems, calling C<stat ()> can block an indefinite
2568 time due to network issues, and even under good conditions, a stat call
2569 often takes multiple milliseconds.
2571 Therefore, it is best to avoid using C<ev_stat> watchers on networked
2572 paths, although this is fully supported by libev.
2574 =head3 The special problem of stat time resolution
2576 The C<stat ()> system call only supports full-second resolution portably,
2577 and even on systems where the resolution is higher, most file systems
2578 still only support whole seconds.
2580 That means that, if the time is the only thing that changes, you can
2581 easily miss updates: on the first update, C<ev_stat> detects a change and
2582 calls your callback, which does something. When there is another update
2583 within the same second, C<ev_stat> will be unable to detect unless the
2584 stat data does change in other ways (e.g. file size).
2586 The solution to this is to delay acting on a change for slightly more
2587 than a second (or till slightly after the next full second boundary), using
2588 a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
2589 ev_timer_again (loop, w)>).
2591 The C<.02> offset is added to work around small timing inconsistencies
2592 of some operating systems (where the second counter of the current time
2593 might be be delayed. One such system is the Linux kernel, where a call to
2594 C<gettimeofday> might return a timestamp with a full second later than
2595 a subsequent C<time> call - if the equivalent of C<time ()> is used to
2596 update file times then there will be a small window where the kernel uses
2597 the previous second to update file times but libev might already execute
2598 the timer callback).
2600 =head3 Watcher-Specific Functions and Data Members
2602 =over 4
2604 =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
2606 =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
2608 Configures the watcher to wait for status changes of the given
2609 C<path>. The C<interval> is a hint on how quickly a change is expected to
2610 be detected and should normally be specified as C<0> to let libev choose
2611 a suitable value. The memory pointed to by C<path> must point to the same
2612 path for as long as the watcher is active.
2614 The callback will receive an C<EV_STAT> event when a change was detected,
2615 relative to the attributes at the time the watcher was started (or the
2616 last change was detected).
2618 =item ev_stat_stat (loop, ev_stat *)
2620 Updates the stat buffer immediately with new values. If you change the
2621 watched path in your callback, you could call this function to avoid
2622 detecting this change (while introducing a race condition if you are not
2623 the only one changing the path). Can also be useful simply to find out the
2624 new values.
2626 =item ev_statdata attr [read-only]
2628 The most-recently detected attributes of the file. Although the type is
2629 C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
2630 suitable for your system, but you can only rely on the POSIX-standardised
2631 members to be present. If the C<st_nlink> member is C<0>, then there was
2632 some error while C<stat>ing the file.
2634 =item ev_statdata prev [read-only]
2636 The previous attributes of the file. The callback gets invoked whenever
2637 C<prev> != C<attr>, or, more precisely, one or more of these members
2638 differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
2639 C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
2641 =item ev_tstamp interval [read-only]
2643 The specified interval.
2645 =item const char *path [read-only]
2647 The file system path that is being watched.
2649 =back
2651 =head3 Examples
2653 Example: Watch C</etc/passwd> for attribute changes.
2655 static void
2656 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
2657 {
2658 /* /etc/passwd changed in some way */
2659 if (w->attr.st_nlink)
2660 {
2661 printf ("passwd current size %ld\n", (long)w->attr.st_size);
2662 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
2663 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
2664 }
2665 else
2666 /* you shalt not abuse printf for puts */
2667 puts ("wow, /etc/passwd is not there, expect problems. "
2668 "if this is windows, they already arrived\n");
2669 }
2671 ...
2672 ev_stat passwd;
2674 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
2675 ev_stat_start (loop, &passwd);
2677 Example: Like above, but additionally use a one-second delay so we do not
2678 miss updates (however, frequent updates will delay processing, too, so
2679 one might do the work both on C<ev_stat> callback invocation I<and> on
2680 C<ev_timer> callback invocation).
2682 static ev_stat passwd;
2683 static ev_timer timer;
2685 static void
2686 timer_cb (EV_P_ ev_timer *w, int revents)
2687 {
2688 ev_timer_stop (EV_A_ w);
2690 /* now it's one second after the most recent passwd change */
2691 }
2693 static void
2694 stat_cb (EV_P_ ev_stat *w, int revents)
2695 {
2696 /* reset the one-second timer */
2697 ev_timer_again (EV_A_ &timer);
2698 }
2700 ...
2701 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2702 ev_stat_start (loop, &passwd);
2703 ev_timer_init (&timer, timer_cb, 0., 1.02);
2706 =head2 C<ev_idle> - when you've got nothing better to do...
2708 Idle watchers trigger events when no other events of the same or higher
2709 priority are pending (prepare, check and other idle watchers do not count
2710 as receiving "events").
2712 That is, as long as your process is busy handling sockets or timeouts
2713 (or even signals, imagine) of the same or higher priority it will not be
2714 triggered. But when your process is idle (or only lower-priority watchers
2715 are pending), the idle watchers are being called once per event loop
2716 iteration - until stopped, that is, or your process receives more events
2717 and becomes busy again with higher priority stuff.
2719 The most noteworthy effect is that as long as any idle watchers are
2720 active, the process will not block when waiting for new events.
2722 Apart from keeping your process non-blocking (which is a useful
2723 effect on its own sometimes), idle watchers are a good place to do
2724 "pseudo-background processing", or delay processing stuff to after the
2725 event loop has handled all outstanding events.
2727 =head3 Watcher-Specific Functions and Data Members
2729 =over 4
2731 =item ev_idle_init (ev_idle *, callback)
2733 Initialises and configures the idle watcher - it has no parameters of any
2734 kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
2735 believe me.
2737 =back
2739 =head3 Examples
2741 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
2742 callback, free it. Also, use no error checking, as usual.
2744 static void
2745 idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
2746 {
2747 free (w);
2748 // now do something you wanted to do when the program has
2749 // no longer anything immediate to do.
2750 }
2752 ev_idle *idle_watcher = malloc (sizeof (ev_idle));
2753 ev_idle_init (idle_watcher, idle_cb);
2754 ev_idle_start (loop, idle_watcher);
2757 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2759 Prepare and check watchers are usually (but not always) used in pairs:
2760 prepare watchers get invoked before the process blocks and check watchers
2761 afterwards.
2763 You I<must not> call C<ev_run> or similar functions that enter
2764 the current event loop from either C<ev_prepare> or C<ev_check>
2765 watchers. Other loops than the current one are fine, however. The
2766 rationale behind this is that you do not need to check for recursion in
2767 those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
2768 C<ev_check> so if you have one watcher of each kind they will always be
2769 called in pairs bracketing the blocking call.
2771 Their main purpose is to integrate other event mechanisms into libev and
2772 their use is somewhat advanced. They could be used, for example, to track
2773 variable changes, implement your own watchers, integrate net-snmp or a
2774 coroutine library and lots more. They are also occasionally useful if
2775 you cache some data and want to flush it before blocking (for example,
2776 in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
2777 watcher).
2779 This is done by examining in each prepare call which file descriptors
2780 need to be watched by the other library, registering C<ev_io> watchers
2781 for them and starting an C<ev_timer> watcher for any timeouts (many
2782 libraries provide exactly this functionality). Then, in the check watcher,
2783 you check for any events that occurred (by checking the pending status
2784 of all watchers and stopping them) and call back into the library. The
2785 I/O and timer callbacks will never actually be called (but must be valid
2786 nevertheless, because you never know, you know?).
2788 As another example, the Perl Coro module uses these hooks to integrate
2789 coroutines into libev programs, by yielding to other active coroutines
2790 during each prepare and only letting the process block if no coroutines
2791 are ready to run (it's actually more complicated: it only runs coroutines
2792 with priority higher than or equal to the event loop and one coroutine
2793 of lower priority, but only once, using idle watchers to keep the event
2794 loop from blocking if lower-priority coroutines are active, thus mapping
2795 low-priority coroutines to idle/background tasks).
2797 It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
2798 priority, to ensure that they are being run before any other watchers
2799 after the poll (this doesn't matter for C<ev_prepare> watchers).
2801 Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
2802 activate ("feed") events into libev. While libev fully supports this, they
2803 might get executed before other C<ev_check> watchers did their job. As
2804 C<ev_check> watchers are often used to embed other (non-libev) event
2805 loops those other event loops might be in an unusable state until their
2806 C<ev_check> watcher ran (always remind yourself to coexist peacefully with
2807 others).
2809 =head3 Watcher-Specific Functions and Data Members
2811 =over 4
2813 =item ev_prepare_init (ev_prepare *, callback)
2815 =item ev_check_init (ev_check *, callback)
2817 Initialises and configures the prepare or check watcher - they have no
2818 parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
2819 macros, but using them is utterly, utterly, utterly and completely
2820 pointless.
2822 =back
2824 =head3 Examples
2826 There are a number of principal ways to embed other event loops or modules
2827 into libev. Here are some ideas on how to include libadns into libev
2828 (there is a Perl module named C<EV::ADNS> that does this, which you could
2829 use as a working example. Another Perl module named C<EV::Glib> embeds a
2830 Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
2831 Glib event loop).
2833 Method 1: Add IO watchers and a timeout watcher in a prepare handler,
2834 and in a check watcher, destroy them and call into libadns. What follows
2835 is pseudo-code only of course. This requires you to either use a low
2836 priority for the check watcher or use C<ev_clear_pending> explicitly, as
2837 the callbacks for the IO/timeout watchers might not have been called yet.
2839 static ev_io iow [nfd];
2840 static ev_timer tw;
2842 static void
2843 io_cb (struct ev_loop *loop, ev_io *w, int revents)
2844 {
2845 }
2847 // create io watchers for each fd and a timer before blocking
2848 static void
2849 adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
2850 {
2851 int timeout = 3600000;
2852 struct pollfd fds [nfd];
2853 // actual code will need to loop here and realloc etc.
2854 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2856 /* the callback is illegal, but won't be called as we stop during check */
2857 ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
2858 ev_timer_start (loop, &tw);
2860 // create one ev_io per pollfd
2861 for (int i = 0; i < nfd; ++i)
2862 {
2863 ev_io_init (iow + i, io_cb, fds [i].fd,
2864 ((fds [i].events & POLLIN ? EV_READ : 0)
2865 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
2867 fds [i].revents = 0;
2868 ev_io_start (loop, iow + i);
2869 }
2870 }
2872 // stop all watchers after blocking
2873 static void
2874 adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
2875 {
2876 ev_timer_stop (loop, &tw);
2878 for (int i = 0; i < nfd; ++i)
2879 {
2880 // set the relevant poll flags
2881 // could also call adns_processreadable etc. here
2882 struct pollfd *fd = fds + i;
2883 int revents = ev_clear_pending (iow + i);
2884 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
2885 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
2887 // now stop the watcher
2888 ev_io_stop (loop, iow + i);
2889 }
2891 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
2892 }
2894 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
2895 in the prepare watcher and would dispose of the check watcher.
2897 Method 3: If the module to be embedded supports explicit event
2898 notification (libadns does), you can also make use of the actual watcher
2899 callbacks, and only destroy/create the watchers in the prepare watcher.
2901 static void
2902 timer_cb (EV_P_ ev_timer *w, int revents)
2903 {
2904 adns_state ads = (adns_state)w->data;
2905 update_now (EV_A);
2907 adns_processtimeouts (ads, &tv_now);
2908 }
2910 static void
2911 io_cb (EV_P_ ev_io *w, int revents)
2912 {
2913 adns_state ads = (adns_state)w->data;
2914 update_now (EV_A);
2916 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
2917 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
2918 }
2920 // do not ever call adns_afterpoll
2922 Method 4: Do not use a prepare or check watcher because the module you
2923 want to embed is not flexible enough to support it. Instead, you can
2924 override their poll function. The drawback with this solution is that the
2925 main loop is now no longer controllable by EV. The C<Glib::EV> module uses
2926 this approach, effectively embedding EV as a client into the horrible
2927 libglib event loop.
2929 static gint
2930 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2931 {
2932 int got_events = 0;
2934 for (n = 0; n < nfds; ++n)
2935 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
2937 if (timeout >= 0)
2938 // create/start timer
2940 // poll
2941 ev_run (EV_A_ 0);
2943 // stop timer again
2944 if (timeout >= 0)
2945 ev_timer_stop (EV_A_ &to);
2947 // stop io watchers again - their callbacks should have set
2948 for (n = 0; n < nfds; ++n)
2949 ev_io_stop (EV_A_ iow [n]);
2951 return got_events;
2952 }
2955 =head2 C<ev_embed> - when one backend isn't enough...
2957 This is a rather advanced watcher type that lets you embed one event loop
2958 into another (currently only C<ev_io> events are supported in the embedded
2959 loop, other types of watchers might be handled in a delayed or incorrect
2960 fashion and must not be used).
2962 There are primarily two reasons you would want that: work around bugs and
2963 prioritise I/O.
2965 As an example for a bug workaround, the kqueue backend might only support
2966 sockets on some platform, so it is unusable as generic backend, but you
2967 still want to make use of it because you have many sockets and it scales
2968 so nicely. In this case, you would create a kqueue-based loop and embed
2969 it into your default loop (which might use e.g. poll). Overall operation
2970 will be a bit slower because first libev has to call C<poll> and then
2971 C<kevent>, but at least you can use both mechanisms for what they are
2972 best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
2974 As for prioritising I/O: under rare circumstances you have the case where
2975 some fds have to be watched and handled very quickly (with low latency),
2976 and even priorities and idle watchers might have too much overhead. In
2977 this case you would put all the high priority stuff in one loop and all
2978 the rest in a second one, and embed the second one in the first.
2980 As long as the watcher is active, the callback will be invoked every
2981 time there might be events pending in the embedded loop. The callback
2982 must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
2983 sweep and invoke their callbacks (the callback doesn't need to invoke the
2984 C<ev_embed_sweep> function directly, it could also start an idle watcher
2985 to give the embedded loop strictly lower priority for example).
2987 You can also set the callback to C<0>, in which case the embed watcher
2988 will automatically execute the embedded loop sweep whenever necessary.
2990 Fork detection will be handled transparently while the C<ev_embed> watcher
2991 is active, i.e., the embedded loop will automatically be forked when the
2992 embedding loop forks. In other cases, the user is responsible for calling
2993 C<ev_loop_fork> on the embedded loop.
2995 Unfortunately, not all backends are embeddable: only the ones returned by
2996 C<ev_embeddable_backends> are, which, unfortunately, does not include any
2997 portable one.
2999 So when you want to use this feature you will always have to be prepared
3000 that you cannot get an embeddable loop. The recommended way to get around
3001 this is to have a separate variables for your embeddable loop, try to
3002 create it, and if that fails, use the normal loop for everything.
3004 =head3 C<ev_embed> and fork
3006 While the C<ev_embed> watcher is running, forks in the embedding loop will
3007 automatically be applied to the embedded loop as well, so no special
3008 fork handling is required in that case. When the watcher is not running,
3009 however, it is still the task of the libev user to call C<ev_loop_fork ()>
3010 as applicable.
3012 =head3 Watcher-Specific Functions and Data Members
3014 =over 4
3016 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
3018 =item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
3020 Configures the watcher to embed the given loop, which must be
3021 embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
3022 invoked automatically, otherwise it is the responsibility of the callback
3023 to invoke it (it will continue to be called until the sweep has been done,
3024 if you do not want that, you need to temporarily stop the embed watcher).
3026 =item ev_embed_sweep (loop, ev_embed *)
3028 Make a single, non-blocking sweep over the embedded loop. This works
3029 similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
3030 appropriate way for embedded loops.
3032 =item struct ev_loop *other [read-only]
3034 The embedded event loop.
3036 =back
3038 =head3 Examples
3040 Example: Try to get an embeddable event loop and embed it into the default
3041 event loop. If that is not possible, use the default loop. The default
3042 loop is stored in C<loop_hi>, while the embeddable loop is stored in
3043 C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
3044 used).
3046 struct ev_loop *loop_hi = ev_default_init (0);
3047 struct ev_loop *loop_lo = 0;
3048 ev_embed embed;
3050 // see if there is a chance of getting one that works
3051 // (remember that a flags value of 0 means autodetection)
3052 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
3053 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
3054 : 0;
3056 // if we got one, then embed it, otherwise default to loop_hi
3057 if (loop_lo)
3058 {
3059 ev_embed_init (&embed, 0, loop_lo);
3060 ev_embed_start (loop_hi, &embed);
3061 }
3062 else
3063 loop_lo = loop_hi;
3065 Example: Check if kqueue is available but not recommended and create
3066 a kqueue backend for use with sockets (which usually work with any
3067 kqueue implementation). Store the kqueue/socket-only event loop in
3068 C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
3070 struct ev_loop *loop = ev_default_init (0);
3071 struct ev_loop *loop_socket = 0;
3072 ev_embed embed;
3074 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
3075 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
3076 {
3077 ev_embed_init (&embed, 0, loop_socket);
3078 ev_embed_start (loop, &embed);
3079 }
3081 if (!loop_socket)
3082 loop_socket = loop;
3084 // now use loop_socket for all sockets, and loop for everything else
3087 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
3089 Fork watchers are called when a C<fork ()> was detected (usually because
3090 whoever is a good citizen cared to tell libev about it by calling
3091 C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
3092 event loop blocks next and before C<ev_check> watchers are being called,
3093 and only in the child after the fork. If whoever good citizen calling
3094 C<ev_default_fork> cheats and calls it in the wrong process, the fork
3095 handlers will be invoked, too, of course.
3097 =head3 The special problem of life after fork - how is it possible?
3099 Most uses of C<fork()> consist of forking, then some simple calls to set
3100 up/change the process environment, followed by a call to C<exec()>. This
3101 sequence should be handled by libev without any problems.
3103 This changes when the application actually wants to do event handling
3104 in the child, or both parent in child, in effect "continuing" after the
3105 fork.
3107 The default mode of operation (for libev, with application help to detect
3108 forks) is to duplicate all the state in the child, as would be expected
3109 when I<either> the parent I<or> the child process continues.
3111 When both processes want to continue using libev, then this is usually the
3112 wrong result. In that case, usually one process (typically the parent) is
3113 supposed to continue with all watchers in place as before, while the other
3114 process typically wants to start fresh, i.e. without any active watchers.
3116 The cleanest and most efficient way to achieve that with libev is to
3117 simply create a new event loop, which of course will be "empty", and
3118 use that for new watchers. This has the advantage of not touching more
3119 memory than necessary, and thus avoiding the copy-on-write, and the
3120 disadvantage of having to use multiple event loops (which do not support
3121 signal watchers).
3123 When this is not possible, or you want to use the default loop for
3124 other reasons, then in the process that wants to start "fresh", call
3125 C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
3126 Destroying the default loop will "orphan" (not stop) all registered
3127 watchers, so you have to be careful not to execute code that modifies
3128 those watchers. Note also that in that case, you have to re-register any
3129 signal watchers.
3131 =head3 Watcher-Specific Functions and Data Members
3133 =over 4
3135 =item ev_fork_init (ev_fork *, callback)
3137 Initialises and configures the fork watcher - it has no parameters of any
3138 kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
3139 really.
3141 =back
3144 =head2 C<ev_cleanup> - even the best things end
3146 Cleanup watchers are called just before the event loop is being destroyed
3147 by a call to C<ev_loop_destroy>.
3149 While there is no guarantee that the event loop gets destroyed, cleanup
3150 watchers provide a convenient method to install cleanup hooks for your
3151 program, worker threads and so on - you just to make sure to destroy the
3152 loop when you want them to be invoked.
3154 Cleanup watchers are invoked in the same way as any other watcher. Unlike
3155 all other watchers, they do not keep a reference to the event loop (which
3156 makes a lot of sense if you think about it). Like all other watchers, you
3157 can call libev functions in the callback, except C<ev_cleanup_start>.
3159 =head3 Watcher-Specific Functions and Data Members
3161 =over 4
3163 =item ev_cleanup_init (ev_cleanup *, callback)
3165 Initialises and configures the cleanup watcher - it has no parameters of
3166 any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
3167 pointless, I assure you.
3169 =back
3171 Example: Register an atexit handler to destroy the default loop, so any
3172 cleanup functions are called.
3174 static void
3175 program_exits (void)
3176 {
3177 ev_loop_destroy (EV_DEFAULT_UC);
3178 }
3180 ...
3181 atexit (program_exits);
3184 =head2 C<ev_async> - how to wake up an event loop
3186 In general, you cannot use an C<ev_run> from multiple threads or other
3187 asynchronous sources such as signal handlers (as opposed to multiple event
3188 loops - those are of course safe to use in different threads).
3190 Sometimes, however, you need to wake up an event loop you do not control,
3191 for example because it belongs to another thread. This is what C<ev_async>
3192 watchers do: as long as the C<ev_async> watcher is active, you can signal
3193 it by calling C<ev_async_send>, which is thread- and signal safe.
3195 This functionality is very similar to C<ev_signal> watchers, as signals,
3196 too, are asynchronous in nature, and signals, too, will be compressed
3197 (i.e. the number of callback invocations may be less than the number of
3198 C<ev_async_sent> calls). In fact, you could use signal watchers as a kind
3199 of "global async watchers" by using a watcher on an otherwise unused
3200 signal, and C<ev_feed_signal> to signal this watcher from another thread,
3201 even without knowing which loop owns the signal.
3203 Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
3204 just the default loop.
3206 =head3 Queueing
3208 C<ev_async> does not support queueing of data in any way. The reason
3209 is that the author does not know of a simple (or any) algorithm for a
3210 multiple-writer-single-reader queue that works in all cases and doesn't
3211 need elaborate support such as pthreads or unportable memory access
3212 semantics.
3214 That means that if you want to queue data, you have to provide your own
3215 queue. But at least I can tell you how to implement locking around your
3216 queue:
3218 =over 4
3220 =item queueing from a signal handler context
3222 To implement race-free queueing, you simply add to the queue in the signal
3223 handler but you block the signal handler in the watcher callback. Here is
3224 an example that does that for some fictitious SIGUSR1 handler:
3226 static ev_async mysig;
3228 static void
3229 sigusr1_handler (void)
3230 {
3231 sometype data;
3233 // no locking etc.
3234 queue_put (data);
3235 ev_async_send (EV_DEFAULT_ &mysig);
3236 }
3238 static void
3239 mysig_cb (EV_P_ ev_async *w, int revents)
3240 {
3241 sometype data;
3242 sigset_t block, prev;
3244 sigemptyset (&block);
3245 sigaddset (&block, SIGUSR1);
3246 sigprocmask (SIG_BLOCK, &block, &prev);
3248 while (queue_get (&data))
3249 process (data);
3251 if (sigismember (&prev, SIGUSR1)
3252 sigprocmask (SIG_UNBLOCK, &block, 0);
3253 }
3255 (Note: pthreads in theory requires you to use C<pthread_setmask>
3256 instead of C<sigprocmask> when you use threads, but libev doesn't do it
3257 either...).
3259 =item queueing from a thread context
3261 The strategy for threads is different, as you cannot (easily) block
3262 threads but you can easily preempt them, so to queue safely you need to
3263 employ a traditional mutex lock, such as in this pthread example:
3265 static ev_async mysig;
3266 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
3268 static void
3269 otherthread (void)
3270 {
3271 // only need to lock the actual queueing operation
3272 pthread_mutex_lock (&mymutex);
3273 queue_put (data);
3274 pthread_mutex_unlock (&mymutex);
3276 ev_async_send (EV_DEFAULT_ &mysig);
3277 }
3279 static void
3280 mysig_cb (EV_P_ ev_async *w, int revents)
3281 {
3282 pthread_mutex_lock (&mymutex);
3284 while (queue_get (&data))
3285 process (data);
3287 pthread_mutex_unlock (&mymutex);
3288 }
3290 =back
3293 =head3 Watcher-Specific Functions and Data Members
3295 =over 4
3297 =item ev_async_init (ev_async *, callback)
3299 Initialises and configures the async watcher - it has no parameters of any
3300 kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
3301 trust me.
3303 =item ev_async_send (loop, ev_async *)
3305 Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3306 an C<EV_ASYNC> event on the watcher into the event loop. Unlike
3307 C<ev_feed_event>, this call is safe to do from other threads, signal or
3308 similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
3309 section below on what exactly this means).
3311 Note that, as with other watchers in libev, multiple events might get
3312 compressed into a single callback invocation (another way to look at this
3313 is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
3314 reset when the event loop detects that).
3316 This call incurs the overhead of a system call only once per event loop
3317 iteration, so while the overhead might be noticeable, it doesn't apply to
3318 repeated calls to C<ev_async_send> for the same event loop.
3320 =item bool = ev_async_pending (ev_async *)
3322 Returns a non-zero value when C<ev_async_send> has been called on the
3323 watcher but the event has not yet been processed (or even noted) by the
3324 event loop.
3326 C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
3327 the loop iterates next and checks for the watcher to have become active,
3328 it will reset the flag again. C<ev_async_pending> can be used to very
3329 quickly check whether invoking the loop might be a good idea.
3331 Not that this does I<not> check whether the watcher itself is pending,
3332 only whether it has been requested to make this watcher pending: there
3333 is a time window between the event loop checking and resetting the async
3334 notification, and the callback being invoked.
3336 =back
3341 There are some other functions of possible interest. Described. Here. Now.
3343 =over 4
3345 =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
3347 This function combines a simple timer and an I/O watcher, calls your
3348 callback on whichever event happens first and automatically stops both
3349 watchers. This is useful if you want to wait for a single event on an fd
3350 or timeout without having to allocate/configure/start/stop/free one or
3351 more watchers yourself.
3353 If C<fd> is less than 0, then no I/O watcher will be started and the
3354 C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
3355 the given C<fd> and C<events> set will be created and started.
3357 If C<timeout> is less than 0, then no timeout watcher will be
3358 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
3359 repeat = 0) will be started. C<0> is a valid timeout.
3361 The callback has the type C<void (*cb)(int revents, void *arg)> and is
3362 passed an C<revents> set like normal event callbacks (a combination of
3363 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
3364 value passed to C<ev_once>. Note that it is possible to receive I<both>
3365 a timeout and an io event at the same time - you probably should give io
3366 events precedence.
3368 Example: wait up to ten seconds for data to appear on STDIN_FILENO.
3370 static void stdin_ready (int revents, void *arg)
3371 {
3372 if (revents & EV_READ)
3373 /* stdin might have data for us, joy! */;
3374 else if (revents & EV_TIMER)
3375 /* doh, nothing entered */;
3376 }
3378 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3380 =item ev_feed_fd_event (loop, int fd, int revents)
3382 Feed an event on the given fd, as if a file descriptor backend detected
3383 the given events it.
3385 =item ev_feed_signal_event (loop, int signum)
3387 Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3388 which is async-safe.
3390 =back
3395 This section explains some common idioms that are not immediately
3396 obvious. Note that examples are sprinkled over the whole manual, and this
3397 section only contains stuff that wouldn't fit anywhere else.
3401 Each watcher has, by default, a C<void *data> member that you can read
3402 or modify at any time: libev will completely ignore it. This can be used
3403 to associate arbitrary data with your watcher. If you need more data and
3404 don't want to allocate memory separately and store a pointer to it in that
3405 data member, you can also "subclass" the watcher type and provide your own
3406 data:
3408 struct my_io
3409 {
3410 ev_io io;
3411 int otherfd;
3412 void *somedata;
3413 struct whatever *mostinteresting;
3414 };
3416 ...
3417 struct my_io w;
3418 ev_io_init (&, my_cb, fd, EV_READ);
3420 And since your callback will be called with a pointer to the watcher, you
3421 can cast it back to your own type:
3423 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
3424 {
3425 struct my_io *w = (struct my_io *)w_;
3426 ...
3427 }
3429 More interesting and less C-conformant ways of casting your callback
3430 function type instead have been omitted.
3434 Another common scenario is to use some data structure with multiple
3435 embedded watchers, in effect creating your own watcher that combines
3436 multiple libev event sources into one "super-watcher":
3438 struct my_biggy
3439 {
3440 int some_data;
3441 ev_timer t1;
3442 ev_timer t2;
3443 }
3445 In this case getting the pointer to C<my_biggy> is a bit more
3446 complicated: Either you store the address of your C<my_biggy> struct in
3447 the C<data> member of the watcher (for woozies or C++ coders), or you need
3448 to use some pointer arithmetic using C<offsetof> inside your watchers (for
3449 real programmers):
3451 #include <stddef.h>
3453 static void
3454 t1_cb (EV_P_ ev_timer *w, int revents)
3455 {
3456 struct my_biggy big = (struct my_biggy *)
3457 (((char *)w) - offsetof (struct my_biggy, t1));
3458 }
3460 static void
3461 t2_cb (EV_P_ ev_timer *w, int revents)
3462 {
3463 struct my_biggy big = (struct my_biggy *)
3464 (((char *)w) - offsetof (struct my_biggy, t2));
3465 }
3469 Often (especially in GUI toolkits) there are places where you have
3470 I<modal> interaction, which is most easily implemented by recursively
3471 invoking C<ev_run>.
3473 This brings the problem of exiting - a callback might want to finish the
3474 main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
3475 a modal "Are you sure?" dialog is still waiting), or just the nested one
3476 and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
3477 other combination: In these cases, C<ev_break> will not work alone.
3479 The solution is to maintain "break this loop" variable for each C<ev_run>
3480 invocation, and use a loop around C<ev_run> until the condition is
3481 triggered, using C<EVRUN_ONCE>:
3483 // main loop
3484 int exit_main_loop = 0;
3486 while (!exit_main_loop)
3487 ev_run (EV_DEFAULT_ EVRUN_ONCE);
3489 // in a model watcher
3490 int exit_nested_loop = 0;
3492 while (!exit_nested_loop)
3493 ev_run (EV_A_ EVRUN_ONCE);
3495 To exit from any of these loops, just set the corresponding exit variable:
3497 // exit modal loop
3498 exit_nested_loop = 1;
3500 // exit main program, after modal loop is finished
3501 exit_main_loop = 1;
3503 // exit both
3504 exit_main_loop = exit_nested_loop = 1;
3508 Here is a fictitious example of how to run an event loop in a different
3509 thread from where callbacks are being invoked and watchers are
3510 created/added/removed.
3512 For a real-world example, see the C<EV::Loop::Async> perl module,
3513 which uses exactly this technique (which is suited for many high-level
3514 languages).
3516 The example uses a pthread mutex to protect the loop data, a condition
3517 variable to wait for callback invocations, an async watcher to notify the
3518 event loop thread and an unspecified mechanism to wake up the main thread.
3520 First, you need to associate some data with the event loop:
3522 typedef struct {
3523 mutex_t lock; /* global loop lock */
3524 ev_async async_w;
3525 thread_t tid;
3526 cond_t invoke_cv;
3527 } userdata;
3529 void prepare_loop (EV_P)
3530 {
3531 // for simplicity, we use a static userdata struct.
3532 static userdata u;
3534 ev_async_init (&u->async_w, async_cb);
3535 ev_async_start (EV_A_ &u->async_w);
3537 pthread_mutex_init (&u->lock, 0);
3538 pthread_cond_init (&u->invoke_cv, 0);
3540 // now associate this with the loop
3541 ev_set_userdata (EV_A_ u);
3542 ev_set_invoke_pending_cb (EV_A_ l_invoke);
3543 ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
3545 // then create the thread running ev_loop
3546 pthread_create (&u->tid, 0, l_run, EV_A);
3547 }
3549 The callback for the C<ev_async> watcher does nothing: the watcher is used
3550 solely to wake up the event loop so it takes notice of any new watchers
3551 that might have been added:
3553 static void
3554 async_cb (EV_P_ ev_async *w, int revents)
3555 {
3556 // just used for the side effects
3557 }
3559 The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
3560 protecting the loop data, respectively.
3562 static void
3563 l_release (EV_P)
3564 {
3565 userdata *u = ev_userdata (EV_A);
3566 pthread_mutex_unlock (&u->lock);
3567 }
3569 static void
3570 l_acquire (EV_P)
3571 {
3572 userdata *u = ev_userdata (EV_A);
3573 pthread_mutex_lock (&u->lock);
3574 }
3576 The event loop thread first acquires the mutex, and then jumps straight
3577 into C<ev_run>:
3579 void *
3580 l_run (void *thr_arg)
3581 {
3582 struct ev_loop *loop = (struct ev_loop *)thr_arg;
3584 l_acquire (EV_A);
3585 pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
3586 ev_run (EV_A_ 0);
3587 l_release (EV_A);
3589 return 0;
3590 }
3592 Instead of invoking all pending watchers, the C<l_invoke> callback will
3593 signal the main thread via some unspecified mechanism (signals? pipe
3594 writes? C<Async::Interrupt>?) and then waits until all pending watchers
3595 have been called (in a while loop because a) spurious wakeups are possible
3596 and b) skipping inter-thread-communication when there are no pending
3597 watchers is very beneficial):
3599 static void
3600 l_invoke (EV_P)
3601 {
3602 userdata *u = ev_userdata (EV_A);
3604 while (ev_pending_count (EV_A))
3605 {
3606 wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
3607 pthread_cond_wait (&u->invoke_cv, &u->lock);
3608 }
3609 }
3611 Now, whenever the main thread gets told to invoke pending watchers, it
3612 will grab the lock, call C<ev_invoke_pending> and then signal the loop
3613 thread to continue:
3615 static void
3616 real_invoke_pending (EV_P)
3617 {
3618 userdata *u = ev_userdata (EV_A);
3620 pthread_mutex_lock (&u->lock);
3621 ev_invoke_pending (EV_A);
3622 pthread_cond_signal (&u->invoke_cv);
3623 pthread_mutex_unlock (&u->lock);
3624 }
3626 Whenever you want to start/stop a watcher or do other modifications to an
3627 event loop, you will now have to lock:
3629 ev_timer timeout_watcher;
3630 userdata *u = ev_userdata (EV_A);
3632 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
3634 pthread_mutex_lock (&u->lock);
3635 ev_timer_start (EV_A_ &timeout_watcher);
3636 ev_async_send (EV_A_ &u->async_w);
3637 pthread_mutex_unlock (&u->lock);
3639 Note that sending the C<ev_async> watcher is required because otherwise
3640 an event loop currently blocking in the kernel will have no knowledge
3641 about the newly added timer. By waking up the loop it will pick up any new
3642 watchers in the next event loop iteration.
3646 While the overhead of a callback that e.g. schedules a thread is small, it
3647 is still an overhead. If you embed libev, and your main usage is with some
3648 kind of threads or coroutines, you might want to customise libev so that
3649 doesn't need callbacks anymore.
3651 Imagine you have coroutines that you can switch to using a function
3652 C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
3653 and that due to some magic, the currently active coroutine is stored in a
3654 global called C<current_coro>. Then you can build your own "wait for libev
3655 event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
3656 the differing C<;> conventions):
3658 #define EV_CB_DECLARE(type) struct my_coro *cb;
3659 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
3661 That means instead of having a C callback function, you store the
3662 coroutine to switch to in each watcher, and instead of having libev call
3663 your callback, you instead have it switch to that coroutine.
3665 A coroutine might now wait for an event with a function called
3666 C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
3667 matter when, or whether the watcher is active or not when this function is
3668 called):
3670 void
3671 wait_for_event (ev_watcher *w)
3672 {
3673 ev_cb_set (w) = current_coro;
3674 switch_to (libev_coro);
3675 }
3677 That basically suspends the coroutine inside C<wait_for_event> and
3678 continues the libev coroutine, which, when appropriate, switches back to
3679 this or any other coroutine. I am sure if you sue this your own :)
3681 You can do similar tricks if you have, say, threads with an event queue -
3682 instead of storing a coroutine, you store the queue object and instead of
3683 switching to a coroutine, you push the watcher onto the queue and notify
3684 any waiters.
3686 To embed libev, see L<EMBEDDING>, but in short, it's easiest to create two
3687 files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
3689 // my_ev.h
3690 #define EV_CB_DECLARE(type) struct my_coro *cb;
3691 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
3692 #include "../libev/ev.h"
3694 // my_ev.c
3695 #define EV_H "my_ev.h"
3696 #include "../libev/ev.c"
3698 And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
3699 F<my_ev.c> into your project. When properly specifying include paths, you
3700 can even use F<ev.h> as header file name directly.
3705 Libev offers a compatibility emulation layer for libevent. It cannot
3706 emulate the internals of libevent, so here are some usage hints:
3708 =over 4
3710 =item * Only the libevent-1.4.1-beta API is being emulated.
3712 This was the newest libevent version available when libev was implemented,
3713 and is still mostly unchanged in 2010.
3715 =item * Use it by including <event.h>, as usual.
3717 =item * The following members are fully supported: ev_base, ev_callback,
3718 ev_arg, ev_fd, ev_res, ev_events.
3720 =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
3721 maintained by libev, it does not work exactly the same way as in libevent (consider
3722 it a private API).
3724 =item * Priorities are not currently supported. Initialising priorities
3725 will fail and all watchers will have the same priority, even though there
3726 is an ev_pri field.
3728 =item * In libevent, the last base created gets the signals, in libev, the
3729 base that registered the signal gets the signals.
3731 =item * Other members are not supported.
3733 =item * The libev emulation is I<not> ABI compatible to libevent, you need
3734 to use the libev header file and library.
3736 =back
3738 =head1 C++ SUPPORT
3740 Libev comes with some simplistic wrapper classes for C++ that mainly allow
3741 you to use some convenience methods to start/stop watchers and also change
3742 the callback model to a model using method callbacks on objects.
3744 To use it,
3746 #include <ev++.h>
3748 This automatically includes F<ev.h> and puts all of its definitions (many
3749 of them macros) into the global namespace. All C++ specific things are
3750 put into the C<ev> namespace. It should support all the same embedding
3751 options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
3753 Care has been taken to keep the overhead low. The only data member the C++
3754 classes add (compared to plain C-style watchers) is the event loop pointer
3755 that the watcher is associated with (or no additional members at all if
3756 you disable C<EV_MULTIPLICITY> when embedding libev).
3758 Currently, functions, static and non-static member functions and classes
3759 with C<operator ()> can be used as callbacks. Other types should be easy
3760 to add as long as they only need one additional pointer for context. If
3761 you need support for other types of functors please contact the author
3762 (preferably after implementing it).
3764 Here is a list of things available in the C<ev> namespace:
3766 =over 4
3768 =item C<ev::READ>, C<ev::WRITE> etc.
3770 These are just enum values with the same values as the C<EV_READ> etc.
3771 macros from F<ev.h>.
3773 =item C<ev::tstamp>, C<ev::now>
3775 Aliases to the same types/functions as with the C<ev_> prefix.
3777 =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
3779 For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
3780 the same name in the C<ev> namespace, with the exception of C<ev_signal>
3781 which is called C<ev::sig> to avoid clashes with the C<signal> macro
3782 defines by many implementations.
3784 All of those classes have these methods:
3786 =over 4
3788 =item ev::TYPE::TYPE ()
3790 =item ev::TYPE::TYPE (loop)
3792 =item ev::TYPE::~TYPE
3794 The constructor (optionally) takes an event loop to associate the watcher
3795 with. If it is omitted, it will use C<EV_DEFAULT>.
3797 The constructor calls C<ev_init> for you, which means you have to call the
3798 C<set> method before starting it.
3800 It will not set a callback, however: You have to call the templated C<set>
3801 method to set a callback before you can start the watcher.
3803 (The reason why you have to use a method is a limitation in C++ which does
3804 not allow explicit template arguments for constructors).
3806 The destructor automatically stops the watcher if it is active.
3808 =item w->set<class, &class::method> (object *)
3810 This method sets the callback method to call. The method has to have a
3811 signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
3812 first argument and the C<revents> as second. The object must be given as
3813 parameter and is stored in the C<data> member of the watcher.
3815 This method synthesizes efficient thunking code to call your method from
3816 the C callback that libev requires. If your compiler can inline your
3817 callback (i.e. it is visible to it at the place of the C<set> call and
3818 your compiler is good :), then the method will be fully inlined into the
3819 thunking function, making it as fast as a direct C callback.
3821 Example: simple class declaration and watcher initialisation
3823 struct myclass
3824 {
3825 void io_cb (ev::io &w, int revents) { }
3826 }
3828 myclass obj;
3829 ev::io iow;
3830 iow.set <myclass, &myclass::io_cb> (&obj);
3832 =item w->set (object *)
3834 This is a variation of a method callback - leaving out the method to call
3835 will default the method to C<operator ()>, which makes it possible to use
3836 functor objects without having to manually specify the C<operator ()> all
3837 the time. Incidentally, you can then also leave out the template argument
3838 list.
3840 The C<operator ()> method prototype must be C<void operator ()(watcher &w,
3841 int revents)>.
3843 See the method-C<set> above for more details.
3845 Example: use a functor object as callback.
3847 struct myfunctor
3848 {
3849 void operator() (ev::io &w, int revents)
3850 {
3851 ...
3852 }
3853 }
3855 myfunctor f;
3857 ev::io w;
3858 w.set (&f);
3860 =item w->set<function> (void *data = 0)
3862 Also sets a callback, but uses a static method or plain function as
3863 callback. The optional C<data> argument will be stored in the watcher's
3864 C<data> member and is free for you to use.
3866 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
3868 See the method-C<set> above for more details.
3870 Example: Use a plain function as callback.
3872 static void io_cb (ev::io &w, int revents) { }
3873 iow.set <io_cb> ();
3875 =item w->set (loop)
3877 Associates a different C<struct ev_loop> with this watcher. You can only
3878 do this when the watcher is inactive (and not pending either).
3880 =item w->set ([arguments])
3882 Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
3883 method or a suitable start method must be called at least once. Unlike the
3884 C counterpart, an active watcher gets automatically stopped and restarted
3885 when reconfiguring it with this method.
3887 =item w->start ()
3889 Starts the watcher. Note that there is no C<loop> argument, as the
3890 constructor already stores the event loop.
3892 =item w->start ([arguments])
3894 Instead of calling C<set> and C<start> methods separately, it is often
3895 convenient to wrap them in one call. Uses the same type of arguments as
3896 the configure C<set> method of the watcher.
3898 =item w->stop ()
3900 Stops the watcher if it is active. Again, no C<loop> argument.
3902 =item w->again () (C<ev::timer>, C<ev::periodic> only)
3904 For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
3905 C<ev_TYPE_again> function.
3907 =item w->sweep () (C<ev::embed> only)
3909 Invokes C<ev_embed_sweep>.
3911 =item w->update () (C<ev::stat> only)
3913 Invokes C<ev_stat_stat>.
3915 =back
3917 =back
3919 Example: Define a class with two I/O and idle watchers, start the I/O
3920 watchers in the constructor.
3922 class myclass
3923 {
3924 ev::io io ; void io_cb (ev::io &w, int revents);
3925 ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
3926 ev::idle idle; void idle_cb (ev::idle &w, int revents);
3928 myclass (int fd)
3929 {
3930 io .set <myclass, &myclass::io_cb > (this);
3931 io2 .set <myclass, &myclass::io2_cb > (this);
3932 idle.set <myclass, &myclass::idle_cb> (this);
3934 io.set (fd, ev::WRITE); // configure the watcher
3935 io.start (); // start it whenever convenient
3937 io2.start (fd, ev::READ); // set + start in one call
3938 }
3939 };
3944 Libev does not offer other language bindings itself, but bindings for a
3945 number of languages exist in the form of third-party packages. If you know
3946 any interesting language binding in addition to the ones listed here, drop
3947 me a note.
3949 =over 4
3951 =item Perl
3953 The EV module implements the full libev API and is actually used to test
3954 libev. EV is developed together with libev. Apart from the EV core module,
3955 there are additional modules that implement libev-compatible interfaces
3956 to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
3957 C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
3958 and C<EV::Glib>).
3960 It can be found and installed via CPAN, its homepage is at
3961 L<>.
3963 =item Python
3965 Python bindings can be found at L<>. It
3966 seems to be quite complete and well-documented.
3968 =item Ruby
3970 Tony Arcieri has written a ruby extension that offers access to a subset
3971 of the libev API and adds file handle abstractions, asynchronous DNS and
3972 more on top of it. It can be found via gem servers. Its homepage is at
3973 L<>.
3975 Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
3976 makes rev work even on mingw.
3978 =item Haskell
3980 A haskell binding to libev is available at
3981 L<>.
3983 =item D
3985 Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3986 be found at L<>.
3988 =item Ocaml
3990 Erkki Seppala has written Ocaml bindings for libev, to be found at
3991 L<>.
3993 =item Lua
3995 Brian Maher has written a partial interface to libev for lua (at the
3996 time of this writing, only C<ev_io> and C<ev_timer>), to be found at
3997 L<>.
3999 =back
4002 =head1 MACRO MAGIC
4004 Libev can be compiled with a variety of options, the most fundamental
4005 of which is C<EV_MULTIPLICITY>. This option determines whether (most)
4006 functions and callbacks have an initial C<struct ev_loop *> argument.
4008 To make it easier to write programs that cope with either variant, the
4009 following macros are defined:
4011 =over 4
4013 =item C<EV_A>, C<EV_A_>
4015 This provides the loop I<argument> for functions, if one is required ("ev
4016 loop argument"). The C<EV_A> form is used when this is the sole argument,
4017 C<EV_A_> is used when other arguments are following. Example:
4019 ev_unref (EV_A);
4020 ev_timer_add (EV_A_ watcher);
4021 ev_run (EV_A_ 0);
4023 It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
4024 which is often provided by the following macro.
4026 =item C<EV_P>, C<EV_P_>
4028 This provides the loop I<parameter> for functions, if one is required ("ev
4029 loop parameter"). The C<EV_P> form is used when this is the sole parameter,
4030 C<EV_P_> is used when other parameters are following. Example:
4032 // this is how ev_unref is being declared
4033 static void ev_unref (EV_P);
4035 // this is how you can declare your typical callback
4036 static void cb (EV_P_ ev_timer *w, int revents)
4038 It declares a parameter C<loop> of type C<struct ev_loop *>, quite
4039 suitable for use with C<EV_A>.
4041 =item C<EV_DEFAULT>, C<EV_DEFAULT_>
4043 Similar to the other two macros, this gives you the value of the default
4044 loop, if multiple loops are supported ("ev loop default").
4048 Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
4049 default loop has been initialised (C<UC> == unchecked). Their behaviour
4050 is undefined when the default loop has not been initialised by a previous
4051 execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
4053 It is often prudent to use C<EV_DEFAULT> when initialising the first
4054 watcher in a function but use C<EV_DEFAULT_UC> afterwards.
4056 =back
4058 Example: Declare and initialise a check watcher, utilising the above
4059 macros so it will work regardless of whether multiple loops are supported
4060 or not.
4062 static void
4063 check_cb (EV_P_ ev_timer *w, int revents)
4064 {
4065 ev_check_stop (EV_A_ w);
4066 }
4068 ev_check check;
4069 ev_check_init (&check, check_cb);
4070 ev_check_start (EV_DEFAULT_ &check);
4071 ev_run (EV_DEFAULT_ 0);
4073 =head1 EMBEDDING
4075 Libev can (and often is) directly embedded into host
4076 applications. Examples of applications that embed it include the Deliantra
4077 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
4078 and rxvt-unicode.
4080 The goal is to enable you to just copy the necessary files into your
4081 source directory without having to change even a single line in them, so
4082 you can easily upgrade by simply copying (or having a checked-out copy of
4083 libev somewhere in your source tree).
4085 =head2 FILESETS
4087 Depending on what features you need you need to include one or more sets of files
4088 in your application.
4090 =head3 CORE EVENT LOOP
4092 To include only the libev core (all the C<ev_*> functions), with manual
4093 configuration (no autoconf):
4095 #define EV_STANDALONE 1
4096 #include "ev.c"
4098 This will automatically include F<ev.h>, too, and should be done in a
4099 single C source file only to provide the function implementations. To use
4100 it, do the same for F<ev.h> in all files wishing to use this API (best
4101 done by writing a wrapper around F<ev.h> that you can include instead and
4102 where you can put other configuration options):
4104 #define EV_STANDALONE 1
4105 #include "ev.h"
4107 Both header files and implementation files can be compiled with a C++
4108 compiler (at least, that's a stated goal, and breakage will be treated
4109 as a bug).
4111 You need the following files in your source tree, or in a directory
4112 in your include path (e.g. in libev/ when using -Ilibev):
4114 ev.h
4115 ev.c
4116 ev_vars.h
4117 ev_wrap.h
4119 ev_win32.c required on win32 platforms only
4121 ev_select.c only when select backend is enabled (which is enabled by default)
4122 ev_poll.c only when poll backend is enabled (disabled by default)
4123 ev_epoll.c only when the epoll backend is enabled (disabled by default)
4124 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
4125 ev_port.c only when the solaris port backend is enabled (disabled by default)
4127 F<ev.c> includes the backend files directly when enabled, so you only need
4128 to compile this single file.
4132 To include the libevent compatibility API, also include:
4134 #include "event.c"
4136 in the file including F<ev.c>, and:
4138 #include "event.h"
4140 in the files that want to use the libevent API. This also includes F<ev.h>.
4142 You need the following additional files for this:
4144 event.h
4145 event.c
4149 Instead of using C<EV_STANDALONE=1> and providing your configuration in
4150 whatever way you want, you can also C<m4_include([libev.m4])> in your
4151 F<> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
4152 include F<config.h> and configure itself accordingly.
4154 For this of course you need the m4 file:
4156 libev.m4
4160 Libev can be configured via a variety of preprocessor symbols you have to
4161 define before including (or compiling) any of its files. The default in
4162 the absence of autoconf is documented for every option.
4164 Symbols marked with "(h)" do not change the ABI, and can have different
4165 values when compiling libev vs. including F<ev.h>, so it is permissible
4166 to redefine them before including F<ev.h> without breaking compatibility
4167 to a compiled library. All other symbols change the ABI, which means all
4168 users of libev and the libev code itself must be compiled with compatible
4169 settings.
4171 =over 4
4173 =item EV_COMPAT3 (h)
4175 Backwards compatibility is a major concern for libev. This is why this
4176 release of libev comes with wrappers for the functions and symbols that
4177 have been renamed between libev version 3 and 4.
4179 You can disable these wrappers (to test compatibility with future
4180 versions) by defining C<EV_COMPAT3> to C<0> when compiling your
4181 sources. This has the additional advantage that you can drop the C<struct>
4182 from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
4183 typedef in that case.
4185 In some future version, the default for C<EV_COMPAT3> will become C<0>,
4186 and in some even more future version the compatibility code will be
4187 removed completely.
4189 =item EV_STANDALONE (h)
4191 Must always be C<1> if you do not use autoconf configuration, which
4192 keeps libev from including F<config.h>, and it also defines dummy
4193 implementations for some libevent functions (such as logging, which is not
4194 supported). It will also not define any of the structs usually found in
4195 F<event.h> that are not directly supported by the libev core alone.
4197 In standalone mode, libev will still try to automatically deduce the
4198 configuration, but has to be more conservative.
4202 If defined to be C<1>, libev will try to detect the availability of the
4203 monotonic clock option at both compile time and runtime. Otherwise no
4204 use of the monotonic clock option will be attempted. If you enable this,
4205 you usually have to link against librt or something similar. Enabling it
4206 when the functionality isn't available is safe, though, although you have
4207 to make sure you link against any libraries where the C<clock_gettime>
4208 function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
4210 =item EV_USE_REALTIME
4212 If defined to be C<1>, libev will try to detect the availability of the
4213 real-time clock option at compile time (and assume its availability
4214 at runtime if successful). Otherwise no use of the real-time clock
4215 option will be attempted. This effectively replaces C<gettimeofday>
4216 by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
4217 correctness. See the note about libraries in the description of
4218 C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
4223 If defined to be C<1>, libev will try to use a direct syscall instead
4224 of calling the system-provided C<clock_gettime> function. This option
4225 exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
4226 unconditionally pulls in C<libpthread>, slowing down single-threaded
4227 programs needlessly. Using a direct syscall is slightly slower (in
4228 theory), because no optimised vdso implementation can be used, but avoids
4229 the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
4230 higher, as it simplifies linking (no need for C<-lrt>).
4234 If defined to be C<1>, libev will assume that C<nanosleep ()> is available
4235 and will use it for delays. Otherwise it will use C<select ()>.
4237 =item EV_USE_EVENTFD
4239 If defined to be C<1>, then libev will assume that C<eventfd ()> is
4240 available and will probe for kernel support at runtime. This will improve
4241 C<ev_signal> and C<ev_async> performance and reduce resource consumption.
4242 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
4243 2.7 or newer, otherwise disabled.
4245 =item EV_USE_SELECT
4247 If undefined or defined to be C<1>, libev will compile in support for the
4248 C<select>(2) backend. No attempt at auto-detection will be done: if no
4249 other method takes over, select will be it. Otherwise the select backend
4250 will not be compiled in.
4254 If defined to C<1>, then the select backend will use the system C<fd_set>
4255 structure. This is useful if libev doesn't compile due to a missing
4256 C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
4257 on exotic systems. This usually limits the range of file descriptors to
4258 some low limit such as 1024 or might have other limitations (winsocket
4259 only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
4260 configures the maximum size of the C<fd_set>.
4264 When defined to C<1>, the select backend will assume that
4265 select/socket/connect etc. don't understand file descriptors but
4266 wants osf handles on win32 (this is the case when the select to
4267 be used is the winsock select). This means that it will call
4268 C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
4269 it is assumed that all these functions actually work on fds, even
4270 on win32. Should not be defined on non-win32 platforms.
4272 =item EV_FD_TO_WIN32_HANDLE(fd)
4274 If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
4275 file descriptors to socket handles. When not defining this symbol (the
4276 default), then libev will call C<_get_osfhandle>, which is usually
4277 correct. In some cases, programs use their own file descriptor management,
4278 in which case they can provide this function to map fds to socket handles.
4280 =item EV_WIN32_HANDLE_TO_FD(handle)
4282 If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
4283 using the standard C<_open_osfhandle> function. For programs implementing
4284 their own fd to handle mapping, overwriting this function makes it easier
4285 to do so. This can be done by defining this macro to an appropriate value.
4287 =item EV_WIN32_CLOSE_FD(fd)
4289 If programs implement their own fd to handle mapping on win32, then this
4290 macro can be used to override the C<close> function, useful to unregister
4291 file descriptors again. Note that the replacement function has to close
4292 the underlying OS handle.
4294 =item EV_USE_POLL
4296 If defined to be C<1>, libev will compile in support for the C<poll>(2)
4297 backend. Otherwise it will be enabled on non-win32 platforms. It
4298 takes precedence over select.
4300 =item EV_USE_EPOLL
4302 If defined to be C<1>, libev will compile in support for the Linux
4303 C<epoll>(7) backend. Its availability will be detected at runtime,
4304 otherwise another method will be used as fallback. This is the preferred
4305 backend for GNU/Linux systems. If undefined, it will be enabled if the
4306 headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4308 =item EV_USE_KQUEUE
4310 If defined to be C<1>, libev will compile in support for the BSD style
4311 C<kqueue>(2) backend. Its actual availability will be detected at runtime,
4312 otherwise another method will be used as fallback. This is the preferred
4313 backend for BSD and BSD-like systems, although on most BSDs kqueue only
4314 supports some types of fds correctly (the only platform we found that
4315 supports ptys for example was NetBSD), so kqueue might be compiled in, but
4316 not be used unless explicitly requested. The best way to use it is to find
4317 out whether kqueue supports your type of fd properly and use an embedded
4318 kqueue loop.
4320 =item EV_USE_PORT
4322 If defined to be C<1>, libev will compile in support for the Solaris
4323 10 port style backend. Its availability will be detected at runtime,
4324 otherwise another method will be used as fallback. This is the preferred
4325 backend for Solaris 10 systems.
4327 =item EV_USE_DEVPOLL
4329 Reserved for future expansion, works like the USE symbols above.
4331 =item EV_USE_INOTIFY
4333 If defined to be C<1>, libev will compile in support for the Linux inotify
4334 interface to speed up C<ev_stat> watchers. Its actual availability will
4335 be detected at runtime. If undefined, it will be enabled if the headers
4336 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4338 =item EV_ATOMIC_T
4340 Libev requires an integer type (suitable for storing C<0> or C<1>) whose
4341 access is atomic with respect to other threads or signal contexts. No such
4342 type is easily found in the C language, so you can provide your own type
4343 that you know is safe for your purposes. It is used both for signal handler "locking"
4344 as well as for signal and thread safety in C<ev_async> watchers.
4346 In the absence of this define, libev will use C<sig_atomic_t volatile>
4347 (from F<signal.h>), which is usually good enough on most platforms.
4349 =item EV_H (h)
4351 The name of the F<ev.h> header file used to include it. The default if
4352 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
4353 used to virtually rename the F<ev.h> header file in case of conflicts.
4355 =item EV_CONFIG_H (h)
4357 If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
4358 F<ev.c>'s idea of where to find the F<config.h> file, similarly to
4359 C<EV_H>, above.
4361 =item EV_EVENT_H (h)
4363 Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
4364 of how the F<event.h> header can be found, the default is C<"event.h">.
4366 =item EV_PROTOTYPES (h)
4368 If defined to be C<0>, then F<ev.h> will not define any function
4369 prototypes, but still define all the structs and other symbols. This is
4370 occasionally useful if you want to provide your own wrapper functions
4371 around libev functions.
4375 If undefined or defined to C<1>, then all event-loop-specific functions
4376 will have the C<struct ev_loop *> as first argument, and you can create
4377 additional independent event loops. Otherwise there will be no support
4378 for multiple event loops and there is no first event loop pointer
4379 argument. Instead, all functions act on the single default loop.
4381 =item EV_MINPRI
4383 =item EV_MAXPRI
4385 The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
4386 C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
4387 provide for more priorities by overriding those symbols (usually defined
4388 to be C<-2> and C<2>, respectively).
4390 When doing priority-based operations, libev usually has to linearly search
4391 all the priorities, so having many of them (hundreds) uses a lot of space
4392 and time, so using the defaults of five priorities (-2 .. +2) is usually
4393 fine.
4395 If your embedding application does not need any priorities, defining these
4396 both to C<0> will save some memory and CPU.
4402 If undefined or defined to be C<1> (and the platform supports it), then
4403 the respective watcher type is supported. If defined to be C<0>, then it
4404 is not. Disabling watcher types mainly saves code size.
4406 =item EV_FEATURES
4408 If you need to shave off some kilobytes of code at the expense of some
4409 speed (but with the full API), you can define this symbol to request
4410 certain subsets of functionality. The default is to enable all features
4411 that can be enabled on the platform.
4413 A typical way to use this symbol is to define it to C<0> (or to a bitset
4414 with some broad features you want) and then selectively re-enable
4415 additional parts you want, for example if you want everything minimal,
4416 but multiple event loop support, async and child watchers and the poll
4417 backend, use this:
4419 #define EV_FEATURES 0
4420 #define EV_MULTIPLICITY 1
4421 #define EV_USE_POLL 1
4422 #define EV_CHILD_ENABLE 1
4423 #define EV_ASYNC_ENABLE 1
4425 The actual value is a bitset, it can be a combination of the following
4426 values:
4428 =over 4
4430 =item C<1> - faster/larger code
4432 Use larger code to speed up some operations.
4434 Currently this is used to override some inlining decisions (enlarging the
4435 code size by roughly 30% on amd64).
4437 When optimising for size, use of compiler flags such as C<-Os> with
4438 gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
4439 assertions.
4441 =item C<2> - faster/larger data structures
4443 Replaces the small 2-heap for timer management by a faster 4-heap, larger
4444 hash table sizes and so on. This will usually further increase code size
4445 and can additionally have an effect on the size of data structures at
4446 runtime.
4448 =item C<4> - full API configuration
4450 This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
4451 enables multiplicity (C<EV_MULTIPLICITY>=1).
4453 =item C<8> - full API
4455 This enables a lot of the "lesser used" API functions. See C<ev.h> for
4456 details on which parts of the API are still available without this
4457 feature, and do not complain if this subset changes over time.
4459 =item C<16> - enable all optional watcher types
4461 Enables all optional watcher types. If you want to selectively enable
4462 only some watcher types other than I/O and timers (e.g. prepare,
4463 embed, async, child...) you can enable them manually by defining
4464 C<EV_watchertype_ENABLE> to C<1> instead.
4466 =item C<32> - enable all backends
4468 This enables all backends - without this feature, you need to enable at
4469 least one backend manually (C<EV_USE_SELECT> is a good choice).
4471 =item C<64> - enable OS-specific "helper" APIs
4473 Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
4474 default.
4476 =back
4478 Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
4479 reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
4480 code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
4481 watchers, timers and monotonic clock support.
4483 With an intelligent-enough linker (gcc+binutils are intelligent enough
4484 when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4485 your program might be left out as well - a binary starting a timer and an
4486 I/O watcher then might come out at only 5Kb.
4488 =item EV_AVOID_STDIO
4490 If this is set to C<1> at compiletime, then libev will avoid using stdio
4491 functions (printf, scanf, perror etc.). This will increase the code size
4492 somewhat, but if your program doesn't otherwise depend on stdio and your
4493 libc allows it, this avoids linking in the stdio library which is quite
4494 big.
4496 Note that error messages might become less precise when this option is
4497 enabled.
4499 =item EV_NSIG
4501 The highest supported signal number, +1 (or, the number of
4502 signals): Normally, libev tries to deduce the maximum number of signals
4503 automatically, but sometimes this fails, in which case it can be
4504 specified. Also, using a lower number than detected (C<32> should be
4505 good for about any system in existence) can save some memory, as libev
4506 statically allocates some 12-24 bytes per signal number.
4508 =item EV_PID_HASHSIZE
4510 C<ev_child> watchers use a small hash table to distribute workload by
4511 pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
4512 usually more than enough. If you need to manage thousands of children you
4513 might want to increase this value (I<must> be a power of two).
4517 C<ev_stat> watchers use a small hash table to distribute workload by
4518 inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
4519 disabled), usually more than enough. If you need to manage thousands of
4520 C<ev_stat> watchers you might want to increase this value (I<must> be a
4521 power of two).
4523 =item EV_USE_4HEAP
4525 Heaps are not very cache-efficient. To improve the cache-efficiency of the
4526 timer and periodics heaps, libev uses a 4-heap when this symbol is defined
4527 to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
4528 faster performance with many (thousands) of watchers.
4530 The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
4531 will be C<0>.
4533 =item EV_HEAP_CACHE_AT
4535 Heaps are not very cache-efficient. To improve the cache-efficiency of the
4536 timer and periodics heaps, libev can cache the timestamp (I<at>) within
4537 the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
4538 which uses 8-12 bytes more per watcher and a few hundred bytes more code,
4539 but avoids random read accesses on heap changes. This improves performance
4540 noticeably with many (hundreds) of watchers.
4542 The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
4543 will be C<0>.
4545 =item EV_VERIFY
4547 Controls how much internal verification (see C<ev_verify ()>) will
4548 be done: If set to C<0>, no internal verification code will be compiled
4549 in. If set to C<1>, then verification code will be compiled in, but not
4550 called. If set to C<2>, then the internal verification code will be
4551 called once per loop, which can slow down libev. If set to C<3>, then the
4552 verification code will be called very frequently, which will slow down
4553 libev considerably.
4555 The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
4556 will be C<0>.
4558 =item EV_COMMON
4560 By default, all watchers have a C<void *data> member. By redefining
4561 this macro to something else you can include more and other types of
4562 members. You have to define it each time you include one of the files,
4563 though, and it must be identical each time.
4565 For example, the perl EV module uses something like this:
4567 #define EV_COMMON \
4568 SV *self; /* contains this struct */ \
4569 SV *cb_sv, *fh /* note no trailing ";" */
4571 =item EV_CB_DECLARE (type)
4573 =item EV_CB_INVOKE (watcher, revents)
4575 =item ev_set_cb (ev, cb)
4577 Can be used to change the callback member declaration in each watcher,
4578 and the way callbacks are invoked and set. Must expand to a struct member
4579 definition and a statement, respectively. See the F<ev.h> header file for
4580 their default definitions. One possible use for overriding these is to
4581 avoid the C<struct ev_loop *> as first argument in all cases, or to use
4582 method calls instead of plain function calls in C++.
4584 =back
4588 If you need to re-export the API (e.g. via a DLL) and you need a list of
4589 exported symbols, you can use the provided F<Symbol.*> files which list
4590 all public symbols, one per line:
4592 Symbols.ev for libev proper
4593 Symbols.event for the libevent emulation
4595 This can also be used to rename all public symbols to avoid clashes with
4596 multiple versions of libev linked together (which is obviously bad in
4597 itself, but sometimes it is inconvenient to avoid this).
4599 A sed command like this will create wrapper C<#define>'s that you need to
4600 include before including F<ev.h>:
4602 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
4604 This would create a file F<wrap.h> which essentially looks like this:
4606 #define ev_backend myprefix_ev_backend
4607 #define ev_check_start myprefix_ev_check_start
4608 #define ev_check_stop myprefix_ev_check_stop
4609 ...
4611 =head2 EXAMPLES
4613 For a real-world example of a program the includes libev
4614 verbatim, you can have a look at the EV perl module
4615 (L<>). It has the libev files in
4616 the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
4617 interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
4618 will be compiled. It is pretty complex because it provides its own header
4619 file.
4621 The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
4622 that everybody includes and which overrides some configure choices:
4624 #define EV_FEATURES 8
4625 #define EV_USE_SELECT 1
4626 #define EV_PREPARE_ENABLE 1
4627 #define EV_IDLE_ENABLE 1
4628 #define EV_SIGNAL_ENABLE 1
4629 #define EV_CHILD_ENABLE 1
4630 #define EV_USE_STDEXCEPT 0
4631 #define EV_CONFIG_H <config.h>
4633 #include "ev++.h"
4635 And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4637 #include "ev_cpp.h"
4638 #include "ev.c"
4644 =head3 THREADS
4646 All libev functions are reentrant and thread-safe unless explicitly
4647 documented otherwise, but libev implements no locking itself. This means
4648 that you can use as many loops as you want in parallel, as long as there
4649 are no concurrent calls into any libev function with the same loop
4650 parameter (C<ev_default_*> calls have an implicit default loop parameter,
4651 of course): libev guarantees that different event loops share no data
4652 structures that need any locking.
4654 Or to put it differently: calls with different loop parameters can be done
4655 concurrently from multiple threads, calls with the same loop parameter
4656 must be done serially (but can be done from different threads, as long as
4657 only one thread ever is inside a call at any point in time, e.g. by using
4658 a mutex per loop).
4660 Specifically to support threads (and signal handlers), libev implements
4661 so-called C<ev_async> watchers, which allow some limited form of
4662 concurrency on the same event loop, namely waking it up "from the
4663 outside".
4665 If you want to know which design (one loop, locking, or multiple loops
4666 without or something else still) is best for your problem, then I cannot
4667 help you, but here is some generic advice:
4669 =over 4
4671 =item * most applications have a main thread: use the default libev loop
4672 in that thread, or create a separate thread running only the default loop.
4674 This helps integrating other libraries or software modules that use libev
4675 themselves and don't care/know about threading.
4677 =item * one loop per thread is usually a good model.
4679 Doing this is almost never wrong, sometimes a better-performance model
4680 exists, but it is always a good start.
4682 =item * other models exist, such as the leader/follower pattern, where one
4683 loop is handed through multiple threads in a kind of round-robin fashion.
4685 Choosing a model is hard - look around, learn, know that usually you can do
4686 better than you currently do :-)
4688 =item * often you need to talk to some other thread which blocks in the
4689 event loop.
4691 C<ev_async> watchers can be used to wake them up from other threads safely
4692 (or from signal contexts...).
4694 An example use would be to communicate signals or other events that only
4695 work in the default loop by registering the signal watcher with the
4696 default loop and triggering an C<ev_async> watcher from the default loop
4697 watcher callback into the event loop interested in the signal.
4699 =back
4703 =head3 COROUTINES
4705 Libev is very accommodating to coroutines ("cooperative threads"):
4706 libev fully supports nesting calls to its functions from different
4707 coroutines (e.g. you can call C<ev_run> on the same loop from two
4708 different coroutines, and switch freely between both coroutines running
4709 the loop, as long as you don't confuse yourself). The only exception is
4710 that you must not do this from C<ev_periodic> reschedule callbacks.
4712 Care has been taken to ensure that libev does not keep local state inside
4713 C<ev_run>, and other calls do not usually allow for coroutine switches as
4714 they do not call any callbacks.
4718 Depending on your compiler and compiler settings, you might get no or a
4719 lot of warnings when compiling libev code. Some people are apparently
4720 scared by this.
4722 However, these are unavoidable for many reasons. For one, each compiler
4723 has different warnings, and each user has different tastes regarding
4724 warning options. "Warn-free" code therefore cannot be a goal except when
4725 targeting a specific compiler and compiler-version.
4727 Another reason is that some compiler warnings require elaborate
4728 workarounds, or other changes to the code that make it less clear and less
4729 maintainable.
4731 And of course, some compiler warnings are just plain stupid, or simply
4732 wrong (because they don't actually warn about the condition their message
4733 seems to warn about). For example, certain older gcc versions had some
4734 warnings that resulted in an extreme number of false positives. These have
4735 been fixed, but some people still insist on making code warn-free with
4736 such buggy versions.
4738 While libev is written to generate as few warnings as possible,
4739 "warn-free" code is not a goal, and it is recommended not to build libev
4740 with any compiler warnings enabled unless you are prepared to cope with
4741 them (e.g. by ignoring them). Remember that warnings are just that:
4742 warnings, not errors, or proof of bugs.
4745 =head2 VALGRIND
4747 Valgrind has a special section here because it is a popular tool that is
4748 highly useful. Unfortunately, valgrind reports are very hard to interpret.
4750 If you think you found a bug (memory leak, uninitialised data access etc.)
4751 in libev, then check twice: If valgrind reports something like:
4753 ==2274== definitely lost: 0 bytes in 0 blocks.
4754 ==2274== possibly lost: 0 bytes in 0 blocks.
4755 ==2274== still reachable: 256 bytes in 1 blocks.
4757 Then there is no memory leak, just as memory accounted to global variables
4758 is not a memleak - the memory is still being referenced, and didn't leak.
4760 Similarly, under some circumstances, valgrind might report kernel bugs
4761 as if it were a bug in libev (e.g. in realloc or in the poll backend,
4762 although an acceptable workaround has been found here), or it might be
4763 confused.
4765 Keep in mind that valgrind is a very good tool, but only a tool. Don't
4766 make it into some kind of religion.
4768 If you are unsure about something, feel free to contact the mailing list
4769 with the full valgrind report and an explanation on why you think this
4770 is a bug in libev (best check the archives, too :). However, don't be
4771 annoyed when you get a brisk "this is no bug" answer and take the chance
4772 of learning how to interpret valgrind properly.
4774 If you need, for some reason, empty reports from valgrind for your project
4775 I suggest using suppression lists.
4782 GNU/Linux is the only common platform that supports 64 bit file/large file
4783 interfaces but I<disables> them by default.
4785 That means that libev compiled in the default environment doesn't support
4786 files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
4788 Unfortunately, many programs try to work around this GNU/Linux issue
4789 by enabling the large file API, which makes them incompatible with the
4790 standard libev compiled for their system.
4792 Likewise, libev cannot enable the large file API itself as this would
4793 suddenly make it incompatible to the default compile time environment,
4794 i.e. all programs not using special compile switches.
4796 =head2 OS/X AND DARWIN BUGS
4798 The whole thing is a bug if you ask me - basically any system interface
4799 you touch is broken, whether it is locales, poll, kqueue or even the
4800 OpenGL drivers.
4802 =head3 C<kqueue> is buggy
4804 The kqueue syscall is broken in all known versions - most versions support
4805 only sockets, many support pipes.
4807 Libev tries to work around this by not using C<kqueue> by default on this
4808 rotten platform, but of course you can still ask for it when creating a
4809 loop - embedding a socket-only kqueue loop into a select-based one is
4810 probably going to work well.
4812 =head3 C<poll> is buggy
4814 Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
4815 implementation by something calling C<kqueue> internally around the 10.5.6
4816 release, so now C<kqueue> I<and> C<poll> are broken.
4818 Libev tries to work around this by not using C<poll> by default on
4819 this rotten platform, but of course you can still ask for it when creating
4820 a loop.
4822 =head3 C<select> is buggy
4824 All that's left is C<select>, and of course Apple found a way to fuck this
4825 one up as well: On OS/X, C<select> actively limits the number of file
4826 descriptors you can pass in to 1024 - your program suddenly crashes when
4827 you use more.
4829 There is an undocumented "workaround" for this - defining
4830 C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
4831 work on OS/X.
4835 =head3 C<errno> reentrancy
4837 The default compile environment on Solaris is unfortunately so
4838 thread-unsafe that you can't even use components/libraries compiled
4839 without C<-D_REENTRANT> in a threaded program, which, of course, isn't
4840 defined by default. A valid, if stupid, implementation choice.
4842 If you want to use libev in threaded environments you have to make sure
4843 it's compiled with C<_REENTRANT> defined.
4845 =head3 Event port backend
4847 The scalable event interface for Solaris is called "event
4848 ports". Unfortunately, this mechanism is very buggy in all major
4849 releases. If you run into high CPU usage, your program freezes or you get
4850 a large number of spurious wakeups, make sure you have all the relevant
4851 and latest kernel patches applied. No, I don't know which ones, but there
4852 are multiple ones to apply, and afterwards, event ports actually work
4853 great.
4855 If you can't get it to work, you can try running the program by setting
4856 the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
4857 C<select> backends.
4859 =head2 AIX POLL BUG
4861 AIX unfortunately has a broken C<poll.h> header. Libev works around
4862 this by trying to avoid the poll backend altogether (i.e. it's not even
4863 compiled in), which normally isn't a big problem as C<select> works fine
4864 with large bitsets on AIX, and AIX is dead anyway.
4868 =head3 General issues
4870 Win32 doesn't support any of the standards (e.g. POSIX) that libev
4871 requires, and its I/O model is fundamentally incompatible with the POSIX
4872 model. Libev still offers limited functionality on this platform in
4873 the form of the C<EVBACKEND_SELECT> backend, and only supports socket
4874 descriptors. This only applies when using Win32 natively, not when using
4875 e.g. cygwin. Actually, it only applies to the microsofts own compilers,
4876 as every compielr comes with a slightly differently broken/incompatible
4877 environment.
4879 Lifting these limitations would basically require the full
4880 re-implementation of the I/O system. If you are into this kind of thing,
4881 then note that glib does exactly that for you in a very portable way (note
4882 also that glib is the slowest event library known to man).
4884 There is no supported compilation method available on windows except
4885 embedding it into other applications.
4887 Sensible signal handling is officially unsupported by Microsoft - libev
4888 tries its best, but under most conditions, signals will simply not work.
4890 Not a libev limitation but worth mentioning: windows apparently doesn't
4891 accept large writes: instead of resulting in a partial write, windows will
4892 either accept everything or return C<ENOBUFS> if the buffer is too large,
4893 so make sure you only write small amounts into your sockets (less than a
4894 megabyte seems safe, but this apparently depends on the amount of memory
4895 available).
4897 Due to the many, low, and arbitrary limits on the win32 platform and
4898 the abysmal performance of winsockets, using a large number of sockets
4899 is not recommended (and not reasonable). If your program needs to use
4900 more than a hundred or so sockets, then likely it needs to use a totally
4901 different implementation for windows, as libev offers the POSIX readiness
4902 notification model, which cannot be implemented efficiently on windows
4903 (due to Microsoft monopoly games).
4905 A typical way to use libev under windows is to embed it (see the embedding
4906 section for details) and use the following F<evwrap.h> header file instead
4907 of F<ev.h>:
4909 #define EV_STANDALONE /* keeps ev from requiring config.h */
4910 #define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
4912 #include "ev.h"
4914 And compile the following F<evwrap.c> file into your project (make sure
4915 you do I<not> compile the F<ev.c> or any other embedded source files!):
4917 #include "evwrap.h"
4918 #include "ev.c"
4920 =head3 The winsocket C<select> function
4922 The winsocket C<select> function doesn't follow POSIX in that it
4923 requires socket I<handles> and not socket I<file descriptors> (it is
4924 also extremely buggy). This makes select very inefficient, and also
4925 requires a mapping from file descriptors to socket handles (the Microsoft
4926 C runtime provides the function C<_open_osfhandle> for this). See the
4927 discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
4928 C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
4930 The configuration for a "naked" win32 using the Microsoft runtime
4931 libraries and raw winsocket select is:
4933 #define EV_USE_SELECT 1
4934 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
4936 Note that winsockets handling of fd sets is O(n), so you can easily get a
4937 complexity in the O(n²) range when using win32.
4939 =head3 Limited number of file descriptors
4941 Windows has numerous arbitrary (and low) limits on things.
4943 Early versions of winsocket's select only supported waiting for a maximum
4944 of C<64> handles (probably owning to the fact that all windows kernels
4945 can only wait for C<64> things at the same time internally; Microsoft
4946 recommends spawning a chain of threads and wait for 63 handles and the
4947 previous thread in each. Sounds great!).
4949 Newer versions support more handles, but you need to define C<FD_SETSIZE>
4950 to some high number (e.g. C<2048>) before compiling the winsocket select
4951 call (which might be in libev or elsewhere, for example, perl and many
4952 other interpreters do their own select emulation on windows).
4954 Another limit is the number of file descriptors in the Microsoft runtime
4955 libraries, which by default is C<64> (there must be a hidden I<64>
4956 fetish or something like this inside Microsoft). You can increase this
4957 by calling C<_setmaxstdio>, which can increase this limit to C<2048>
4958 (another arbitrary limit), but is broken in many versions of the Microsoft
4959 runtime libraries. This might get you to about C<512> or C<2048> sockets
4960 (depending on windows version and/or the phase of the moon). To get more,
4961 you need to wrap