ViewVC Help
View File | Revision Log | Show Annotations | Download File
Revision: 1.458
Committed: Fri Dec 20 20:51:46 2019 UTC (4 years, 6 months ago) by root
Branch: MAIN
CVS Tags: EV-rel-4_31, rel-4_31
Changes since 1.457: +34 -1 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =encoding utf-8
3 =head1 NAME
5 libev - a high performance full-featured event loop written in C
7 =head1 SYNOPSIS
9 #include <ev.h>
13 // a single header file is required
14 #include <ev.h>
16 #include <stdio.h> // for puts
18 // every watcher type has its own typedef'd struct
19 // with the name ev_TYPE
20 ev_io stdin_watcher;
21 ev_timer timeout_watcher;
23 // all watcher callbacks have a similar signature
24 // this callback is called when data is readable on stdin
25 static void
26 stdin_cb (EV_P_ ev_io *w, int revents)
27 {
28 puts ("stdin ready");
29 // for one-shot events, one must manually stop the watcher
30 // with its corresponding stop function.
31 ev_io_stop (EV_A_ w);
33 // this causes all nested ev_run's to stop iterating
34 ev_break (EV_A_ EVBREAK_ALL);
35 }
37 // another callback, this time for a time-out
38 static void
39 timeout_cb (EV_P_ ev_timer *w, int revents)
40 {
41 puts ("timeout");
42 // this causes the innermost ev_run to stop iterating
43 ev_break (EV_A_ EVBREAK_ONE);
44 }
46 int
47 main (void)
48 {
49 // use the default event loop unless you have special needs
50 struct ev_loop *loop = EV_DEFAULT;
52 // initialise an io watcher, then start it
53 // this one will watch for stdin to become readable
54 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
55 ev_io_start (loop, &stdin_watcher);
57 // initialise a timer watcher, then start it
58 // simple non-repeating 5.5 second timeout
59 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
60 ev_timer_start (loop, &timeout_watcher);
62 // now wait for events to arrive
63 ev_run (loop, 0);
65 // break was called, so exit
66 return 0;
67 }
71 This document documents the libev software package.
73 The newest version of this document is also available as an html-formatted
74 web page you might find easier to navigate when reading it for the first
75 time: L<>.
77 While this document tries to be as complete as possible in documenting
78 libev, its usage and the rationale behind its design, it is not a tutorial
79 on event-based programming, nor will it introduce event-based programming
80 with libev.
82 Familiarity with event based programming techniques in general is assumed
83 throughout this document.
87 This manual tries to be very detailed, but unfortunately, this also makes
88 it very long. If you just want to know the basics of libev, I suggest
89 reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
90 look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
91 C<ev_timer> sections in L</WATCHER TYPES>.
93 =head1 ABOUT LIBEV
95 Libev is an event loop: you register interest in certain events (such as a
96 file descriptor being readable or a timeout occurring), and it will manage
97 these event sources and provide your program with events.
99 To do this, it must take more or less complete control over your process
100 (or thread) by executing the I<event loop> handler, and will then
101 communicate events via a callback mechanism.
103 You register interest in certain events by registering so-called I<event
104 watchers>, which are relatively small C structures you initialise with the
105 details of the event, and then hand it over to libev by I<starting> the
106 watcher.
108 =head2 FEATURES
110 Libev supports C<select>, C<poll>, the Linux-specific aio and C<epoll>
111 interfaces, the BSD-specific C<kqueue> and the Solaris-specific event port
112 mechanisms for file descriptor events (C<ev_io>), the Linux C<inotify>
113 interface (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
114 inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
115 timers (C<ev_timer>), absolute timers with customised rescheduling
116 (C<ev_periodic>), synchronous signals (C<ev_signal>), process status
117 change events (C<ev_child>), and event watchers dealing with the event
118 loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
119 C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
120 limited support for fork events (C<ev_fork>).
122 It also is quite fast (see this
123 L<benchmark|> comparing it to libevent
124 for example).
126 =head2 CONVENTIONS
128 Libev is very configurable. In this manual the default (and most common)
129 configuration will be described, which supports multiple event loops. For
130 more info about various configuration options please have a look at
131 B<EMBED> section in this manual. If libev was configured without support
132 for multiple event loops, then all functions taking an initial argument of
133 name C<loop> (which is always of type C<struct ev_loop *>) will not have
134 this argument.
138 Libev represents time as a single floating point number, representing
139 the (fractional) number of seconds since the (POSIX) epoch (in practice
140 somewhere near the beginning of 1970, details are complicated, don't
141 ask). This type is called C<ev_tstamp>, which is what you should use
142 too. It usually aliases to the C<double> type in C. When you need to do
143 any calculations on it, you should treat it as some floating point value.
145 Unlike the name component C<stamp> might indicate, it is also used for
146 time differences (e.g. delays) throughout libev.
150 Libev knows three classes of errors: operating system errors, usage errors
151 and internal errors (bugs).
153 When libev catches an operating system error it cannot handle (for example
154 a system call indicating a condition libev cannot fix), it calls the callback
155 set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
156 abort. The default is to print a diagnostic message and to call C<abort
157 ()>.
159 When libev detects a usage error such as a negative timer interval, then
160 it will print a diagnostic message and abort (via the C<assert> mechanism,
161 so C<NDEBUG> will disable this checking): these are programming errors in
162 the libev caller and need to be fixed there.
164 Via the C<EV_FREQUENT> macro you can compile in and/or enable extensive
165 consistency checking code inside libev that can be used to check for
166 internal inconsistencies, suually caused by application bugs.
168 Libev also has a few internal error-checking C<assert>ions. These do not
169 trigger under normal circumstances, as they indicate either a bug in libev
170 or worse.
175 These functions can be called anytime, even before initialising the
176 library in any way.
178 =over 4
180 =item ev_tstamp ev_time ()
182 Returns the current time as libev would use it. Please note that the
183 C<ev_now> function is usually faster and also often returns the timestamp
184 you actually want to know. Also interesting is the combination of
185 C<ev_now_update> and C<ev_now>.
187 =item ev_sleep (ev_tstamp interval)
189 Sleep for the given interval: The current thread will be blocked
190 until either it is interrupted or the given time interval has
191 passed (approximately - it might return a bit earlier even if not
192 interrupted). Returns immediately if C<< interval <= 0 >>.
194 Basically this is a sub-second-resolution C<sleep ()>.
196 The range of the C<interval> is limited - libev only guarantees to work
197 with sleep times of up to one day (C<< interval <= 86400 >>).
199 =item int ev_version_major ()
201 =item int ev_version_minor ()
203 You can find out the major and minor ABI version numbers of the library
204 you linked against by calling the functions C<ev_version_major> and
205 C<ev_version_minor>. If you want, you can compare against the global
206 symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
207 version of the library your program was compiled against.
209 These version numbers refer to the ABI version of the library, not the
210 release version.
212 Usually, it's a good idea to terminate if the major versions mismatch,
213 as this indicates an incompatible change. Minor versions are usually
214 compatible to older versions, so a larger minor version alone is usually
215 not a problem.
217 Example: Make sure we haven't accidentally been linked against the wrong
218 version (note, however, that this will not detect other ABI mismatches,
219 such as LFS or reentrancy).
221 assert (("libev version mismatch",
222 ev_version_major () == EV_VERSION_MAJOR
223 && ev_version_minor () >= EV_VERSION_MINOR));
225 =item unsigned int ev_supported_backends ()
227 Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
228 value) compiled into this binary of libev (independent of their
229 availability on the system you are running on). See C<ev_default_loop> for
230 a description of the set values.
232 Example: make sure we have the epoll method, because yeah this is cool and
233 a must have and can we have a torrent of it please!!!11
235 assert (("sorry, no epoll, no sex",
236 ev_supported_backends () & EVBACKEND_EPOLL));
238 =item unsigned int ev_recommended_backends ()
240 Return the set of all backends compiled into this binary of libev and
241 also recommended for this platform, meaning it will work for most file
242 descriptor types. This set is often smaller than the one returned by
243 C<ev_supported_backends>, as for example kqueue is broken on most BSDs
244 and will not be auto-detected unless you explicitly request it (assuming
245 you know what you are doing). This is the set of backends that libev will
246 probe for if you specify no backends explicitly.
248 =item unsigned int ev_embeddable_backends ()
250 Returns the set of backends that are embeddable in other event loops. This
251 value is platform-specific but can include backends not available on the
252 current system. To find which embeddable backends might be supported on
253 the current system, you would need to look at C<ev_embeddable_backends ()
254 & ev_supported_backends ()>, likewise for recommended ones.
256 See the description of C<ev_embed> watchers for more info.
258 =item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
260 Sets the allocation function to use (the prototype is similar - the
261 semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
262 used to allocate and free memory (no surprises here). If it returns zero
263 when memory needs to be allocated (C<size != 0>), the library might abort
264 or take some potentially destructive action.
266 Since some systems (at least OpenBSD and Darwin) fail to implement
267 correct C<realloc> semantics, libev will use a wrapper around the system
268 C<realloc> and C<free> functions by default.
270 You could override this function in high-availability programs to, say,
271 free some memory if it cannot allocate memory, to use a special allocator,
272 or even to sleep a while and retry until some memory is available.
274 Example: The following is the C<realloc> function that libev itself uses
275 which should work with C<realloc> and C<free> functions of all kinds and
276 is probably a good basis for your own implementation.
278 static void *
279 ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
280 {
281 if (size)
282 return realloc (ptr, size);
284 free (ptr);
285 return 0;
286 }
288 Example: Replace the libev allocator with one that waits a bit and then
289 retries.
291 static void *
292 persistent_realloc (void *ptr, size_t size)
293 {
294 if (!size)
295 {
296 free (ptr);
297 return 0;
298 }
300 for (;;)
301 {
302 void *newptr = realloc (ptr, size);
304 if (newptr)
305 return newptr;
307 sleep (60);
308 }
309 }
311 ...
312 ev_set_allocator (persistent_realloc);
314 =item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
316 Set the callback function to call on a retryable system call error (such
317 as failed select, poll, epoll_wait). The message is a printable string
318 indicating the system call or subsystem causing the problem. If this
319 callback is set, then libev will expect it to remedy the situation, no
320 matter what, when it returns. That is, libev will generally retry the
321 requested operation, or, if the condition doesn't go away, do bad stuff
322 (such as abort).
324 Example: This is basically the same thing that libev does internally, too.
326 static void
327 fatal_error (const char *msg)
328 {
329 perror (msg);
330 abort ();
331 }
333 ...
334 ev_set_syserr_cb (fatal_error);
336 =item ev_feed_signal (int signum)
338 This function can be used to "simulate" a signal receive. It is completely
339 safe to call this function at any time, from any context, including signal
340 handlers or random threads.
342 Its main use is to customise signal handling in your process, especially
343 in the presence of threads. For example, you could block signals
344 by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
345 creating any loops), and in one thread, use C<sigwait> or any other
346 mechanism to wait for signals, then "deliver" them to libev by calling
347 C<ev_feed_signal>.
349 =back
353 An event loop is described by a C<struct ev_loop *> (the C<struct> is
354 I<not> optional in this case unless libev 3 compatibility is disabled, as
355 libev 3 had an C<ev_loop> function colliding with the struct name).
357 The library knows two types of such loops, the I<default> loop, which
358 supports child process events, and dynamically created event loops which
359 do not.
361 =over 4
363 =item struct ev_loop *ev_default_loop (unsigned int flags)
365 This returns the "default" event loop object, which is what you should
366 normally use when you just need "the event loop". Event loop objects and
367 the C<flags> parameter are described in more detail in the entry for
368 C<ev_loop_new>.
370 If the default loop is already initialised then this function simply
371 returns it (and ignores the flags. If that is troubling you, check
372 C<ev_backend ()> afterwards). Otherwise it will create it with the given
373 flags, which should almost always be C<0>, unless the caller is also the
374 one calling C<ev_run> or otherwise qualifies as "the main program".
376 If you don't know what event loop to use, use the one returned from this
377 function (or via the C<EV_DEFAULT> macro).
379 Note that this function is I<not> thread-safe, so if you want to use it
380 from multiple threads, you have to employ some kind of mutex (note also
381 that this case is unlikely, as loops cannot be shared easily between
382 threads anyway).
384 The default loop is the only loop that can handle C<ev_child> watchers,
385 and to do this, it always registers a handler for C<SIGCHLD>. If this is
386 a problem for your application you can either create a dynamic loop with
387 C<ev_loop_new> which doesn't do that, or you can simply overwrite the
388 C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
390 Example: This is the most typical usage.
392 if (!ev_default_loop (0))
393 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
395 Example: Restrict libev to the select and poll backends, and do not allow
396 environment settings to be taken into account:
400 =item struct ev_loop *ev_loop_new (unsigned int flags)
402 This will create and initialise a new event loop object. If the loop
403 could not be initialised, returns false.
405 This function is thread-safe, and one common way to use libev with
406 threads is indeed to create one loop per thread, and using the default
407 loop in the "main" or "initial" thread.
409 The flags argument can be used to specify special behaviour or specific
410 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
412 The following flags are supported:
414 =over 4
416 =item C<EVFLAG_AUTO>
418 The default flags value. Use this if you have no clue (it's the right
419 thing, believe me).
421 =item C<EVFLAG_NOENV>
423 If this flag bit is or'ed into the flag value (or the program runs setuid
424 or setgid) then libev will I<not> look at the environment variable
425 C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
426 override the flags completely if it is found in the environment. This is
427 useful to try out specific backends to test their performance, to work
428 around bugs, or to make libev threadsafe (accessing environment variables
429 cannot be done in a threadsafe way, but usually it works if no other
430 thread modifies them).
434 Instead of calling C<ev_loop_fork> manually after a fork, you can also
435 make libev check for a fork in each iteration by enabling this flag.
437 This works by calling C<getpid ()> on every iteration of the loop,
438 and thus this might slow down your event loop if you do a lot of loop
439 iterations and little real work, but is usually not noticeable (on my
440 GNU/Linux system for example, C<getpid> is actually a simple 5-insn
441 sequence without a system call and thus I<very> fast, but my GNU/Linux
442 system also has C<pthread_atfork> which is even faster). (Update: glibc
443 versions 2.25 apparently removed the C<getpid> optimisation again).
445 The big advantage of this flag is that you can forget about fork (and
446 forget about forgetting to tell libev about forking, although you still
447 have to ignore C<SIGPIPE>) when you use this flag.
449 This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
450 environment variable.
454 When this flag is specified, then libev will not attempt to use the
455 I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
456 testing, this flag can be useful to conserve inotify file descriptors, as
457 otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
461 When this flag is specified, then libev will attempt to use the
462 I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
463 delivers signals synchronously, which makes it both faster and might make
464 it possible to get the queued signal data. It can also simplify signal
465 handling with threads, as long as you properly block signals in your
466 threads that are not interested in handling them.
468 Signalfd will not be used by default as this changes your signal mask, and
469 there are a lot of shoddy libraries and programs (glib's threadpool for
470 example) that can't properly initialise their signal masks.
474 When this flag is specified, then libev will avoid to modify the signal
475 mask. Specifically, this means you have to make sure signals are unblocked
476 when you want to receive them.
478 This behaviour is useful when you want to do your own signal handling, or
479 want to handle signals only in specific threads and want to avoid libev
480 unblocking the signals.
482 It's also required by POSIX in a threaded program, as libev calls
483 C<sigprocmask>, whose behaviour is officially unspecified.
487 When this flag is specified, the libev will avoid using a C<timerfd> to
488 detect time jumps. It will still be able to detect time jumps, but takes
489 longer and has a lower accuracy in doing so, but saves a file descriptor
490 per loop.
492 The current implementation only tries to use a C<timerfd> when the first
493 C<ev_periodic> watcher is started and falls back on other methods if it
494 cannot be created, but this behaviour might change in the future.
496 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
498 This is your standard select(2) backend. Not I<completely> standard, as
499 libev tries to roll its own fd_set with no limits on the number of fds,
500 but if that fails, expect a fairly low limit on the number of fds when
501 using this backend. It doesn't scale too well (O(highest_fd)), but its
502 usually the fastest backend for a low number of (low-numbered :) fds.
504 To get good performance out of this backend you need a high amount of
505 parallelism (most of the file descriptors should be busy). If you are
506 writing a server, you should C<accept ()> in a loop to accept as many
507 connections as possible during one iteration. You might also want to have
508 a look at C<ev_set_io_collect_interval ()> to increase the amount of
509 readiness notifications you get per iteration.
511 This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
512 C<writefds> set (and to work around Microsoft Windows bugs, also onto the
513 C<exceptfds> set on that platform).
515 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
517 And this is your standard poll(2) backend. It's more complicated
518 than select, but handles sparse fds better and has no artificial
519 limit on the number of fds you can use (except it will slow down
520 considerably with a lot of inactive fds). It scales similarly to select,
521 i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
522 performance tips.
524 This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
527 =item C<EVBACKEND_EPOLL> (value 4, Linux)
529 Use the Linux-specific epoll(7) interface (for both pre- and post-2.6.9
530 kernels).
532 For few fds, this backend is a bit little slower than poll and select, but
533 it scales phenomenally better. While poll and select usually scale like
534 O(total_fds) where total_fds is the total number of fds (or the highest
535 fd), epoll scales either O(1) or O(active_fds).
537 The epoll mechanism deserves honorable mention as the most misdesigned
538 of the more advanced event mechanisms: mere annoyances include silently
539 dropping file descriptors, requiring a system call per change per file
540 descriptor (and unnecessary guessing of parameters), problems with dup,
541 returning before the timeout value, resulting in additional iterations
542 (and only giving 5ms accuracy while select on the same platform gives
543 0.1ms) and so on. The biggest issue is fork races, however - if a program
544 forks then I<both> parent and child process have to recreate the epoll
545 set, which can take considerable time (one syscall per file descriptor)
546 and is of course hard to detect.
548 Epoll is also notoriously buggy - embedding epoll fds I<should> work,
549 but of course I<doesn't>, and epoll just loves to report events for
550 totally I<different> file descriptors (even already closed ones, so
551 one cannot even remove them from the set) than registered in the set
552 (especially on SMP systems). Libev tries to counter these spurious
553 notifications by employing an additional generation counter and comparing
554 that against the events to filter out spurious ones, recreating the set
555 when required. Epoll also erroneously rounds down timeouts, but gives you
556 no way to know when and by how much, so sometimes you have to busy-wait
557 because epoll returns immediately despite a nonzero timeout. And last
558 not least, it also refuses to work with some file descriptors which work
559 perfectly fine with C<select> (files, many character devices...).
561 Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
562 cobbled together in a hurry, no thought to design or interaction with
563 others. Oh, the pain, will it ever stop...
565 While stopping, setting and starting an I/O watcher in the same iteration
566 will result in some caching, there is still a system call per such
567 incident (because the same I<file descriptor> could point to a different
568 I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
569 file descriptors might not work very well if you register events for both
570 file descriptors.
572 Best performance from this backend is achieved by not unregistering all
573 watchers for a file descriptor until it has been closed, if possible,
574 i.e. keep at least one watcher active per fd at all times. Stopping and
575 starting a watcher (without re-setting it) also usually doesn't cause
576 extra overhead. A fork can both result in spurious notifications as well
577 as in libev having to destroy and recreate the epoll object, which can
578 take considerable time and thus should be avoided.
580 All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
581 faster than epoll for maybe up to a hundred file descriptors, depending on
582 the usage. So sad.
584 While nominally embeddable in other event loops, this feature is broken in
585 a lot of kernel revisions, but probably(!) works in current versions.
587 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
590 =item C<EVBACKEND_LINUXAIO> (value 64, Linux)
592 Use the Linux-specific Linux AIO (I<not> C<< aio(7) >> but C<<
593 io_submit(2) >>) event interface available in post-4.18 kernels (but libev
594 only tries to use it in 4.19+).
596 This is another Linux train wreck of an event interface.
598 If this backend works for you (as of this writing, it was very
599 experimental), it is the best event interface available on Linux and might
600 be well worth enabling it - if it isn't available in your kernel this will
601 be detected and this backend will be skipped.
603 This backend can batch oneshot requests and supports a user-space ring
604 buffer to receive events. It also doesn't suffer from most of the design
605 problems of epoll (such as not being able to remove event sources from
606 the epoll set), and generally sounds too good to be true. Because, this
607 being the Linux kernel, of course it suffers from a whole new set of
608 limitations, forcing you to fall back to epoll, inheriting all its design
609 issues.
611 For one, it is not easily embeddable (but probably could be done using
612 an event fd at some extra overhead). It also is subject to a system wide
613 limit that can be configured in F</proc/sys/fs/aio-max-nr>. If no AIO
614 requests are left, this backend will be skipped during initialisation, and
615 will switch to epoll when the loop is active.
617 Most problematic in practice, however, is that not all file descriptors
618 work with it. For example, in Linux 5.1, TCP sockets, pipes, event fds,
619 files, F</dev/null> and many others are supported, but ttys do not work
620 properly (a known bug that the kernel developers don't care about, see
621 L<>), so this is not
622 (yet?) a generic event polling interface.
624 Overall, it seems the Linux developers just don't want it to have a
625 generic event handling mechanism other than C<select> or C<poll>.
627 To work around all these problem, the current version of libev uses its
628 epoll backend as a fallback for file descriptor types that do not work. Or
629 falls back completely to epoll if the kernel acts up.
631 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
634 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
636 Kqueue deserves special mention, as at the time this backend was
637 implemented, it was broken on all BSDs except NetBSD (usually it doesn't
638 work reliably with anything but sockets and pipes, except on Darwin,
639 where of course it's completely useless). Unlike epoll, however, whose
640 brokenness is by design, these kqueue bugs can be (and mostly have been)
641 fixed without API changes to existing programs. For this reason it's not
642 being "auto-detected" on all platforms unless you explicitly specify it
643 in the flags (i.e. using C<EVBACKEND_KQUEUE>) or libev was compiled on a
644 known-to-be-good (-enough) system like NetBSD.
646 You still can embed kqueue into a normal poll or select backend and use it
647 only for sockets (after having made sure that sockets work with kqueue on
648 the target platform). See C<ev_embed> watchers for more info.
650 It scales in the same way as the epoll backend, but the interface to the
651 kernel is more efficient (which says nothing about its actual speed, of
652 course). While stopping, setting and starting an I/O watcher does never
653 cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
654 two event changes per incident. Support for C<fork ()> is very bad (you
655 might have to leak fds on fork, but it's more sane than epoll) and it
656 drops fds silently in similarly hard-to-detect cases.
658 This backend usually performs well under most conditions.
660 While nominally embeddable in other event loops, this doesn't work
661 everywhere, so you might need to test for this. And since it is broken
662 almost everywhere, you should only use it when you have a lot of sockets
663 (for which it usually works), by embedding it into another event loop
664 (e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
665 also broken on OS X)) and, did I mention it, using it only for sockets.
667 This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
668 C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
669 C<NOTE_EOF>.
671 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
673 This is not implemented yet (and might never be, unless you send me an
674 implementation). According to reports, C</dev/poll> only supports sockets
675 and is not embeddable, which would limit the usefulness of this backend
676 immensely.
678 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
680 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
681 it's really slow, but it still scales very well (O(active_fds)).
683 While this backend scales well, it requires one system call per active
684 file descriptor per loop iteration. For small and medium numbers of file
685 descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
686 might perform better.
688 On the positive side, this backend actually performed fully to
689 specification in all tests and is fully embeddable, which is a rare feat
690 among the OS-specific backends (I vastly prefer correctness over speed
691 hacks).
693 On the negative side, the interface is I<bizarre> - so bizarre that
694 even sun itself gets it wrong in their code examples: The event polling
695 function sometimes returns events to the caller even though an error
696 occurred, but with no indication whether it has done so or not (yes, it's
697 even documented that way) - deadly for edge-triggered interfaces where you
698 absolutely have to know whether an event occurred or not because you have
699 to re-arm the watcher.
701 Fortunately libev seems to be able to work around these idiocies.
703 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
706 =item C<EVBACKEND_ALL>
708 Try all backends (even potentially broken ones that wouldn't be tried
709 with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
712 It is definitely not recommended to use this flag, use whatever
713 C<ev_recommended_backends ()> returns, or simply do not specify a backend
714 at all.
718 Not a backend at all, but a mask to select all backend bits from a
719 C<flags> value, in case you want to mask out any backends from a flags
720 value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
722 =back
724 If one or more of the backend flags are or'ed into the flags value,
725 then only these backends will be tried (in the reverse order as listed
726 here). If none are specified, all backends in C<ev_recommended_backends
727 ()> will be tried.
729 Example: Try to create a event loop that uses epoll and nothing else.
731 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
732 if (!epoller)
733 fatal ("no epoll found here, maybe it hides under your chair");
735 Example: Use whatever libev has to offer, but make sure that kqueue is
736 used if available.
738 struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
740 Example: Similarly, on linux, you mgiht want to take advantage of the
741 linux aio backend if possible, but fall back to something else if that
742 isn't available.
744 struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
746 =item ev_loop_destroy (loop)
748 Destroys an event loop object (frees all memory and kernel state
749 etc.). None of the active event watchers will be stopped in the normal
750 sense, so e.g. C<ev_is_active> might still return true. It is your
751 responsibility to either stop all watchers cleanly yourself I<before>
752 calling this function, or cope with the fact afterwards (which is usually
753 the easiest thing, you can just ignore the watchers and/or C<free ()> them
754 for example).
756 Note that certain global state, such as signal state (and installed signal
757 handlers), will not be freed by this function, and related watchers (such
758 as signal and child watchers) would need to be stopped manually.
760 This function is normally used on loop objects allocated by
761 C<ev_loop_new>, but it can also be used on the default loop returned by
762 C<ev_default_loop>, in which case it is not thread-safe.
764 Note that it is not advisable to call this function on the default loop
765 except in the rare occasion where you really need to free its resources.
766 If you need dynamically allocated loops it is better to use C<ev_loop_new>
767 and C<ev_loop_destroy>.
769 =item ev_loop_fork (loop)
771 This function sets a flag that causes subsequent C<ev_run> iterations
772 to reinitialise the kernel state for backends that have one. Despite
773 the name, you can call it anytime you are allowed to start or stop
774 watchers (except inside an C<ev_prepare> callback), but it makes most
775 sense after forking, in the child process. You I<must> call it (or use
776 C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
778 In addition, if you want to reuse a loop (via this function or
779 C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
781 Again, you I<have> to call it on I<any> loop that you want to re-use after
782 a fork, I<even if you do not plan to use the loop in the parent>. This is
783 because some kernel interfaces *cough* I<kqueue> *cough* do funny things
784 during fork.
786 On the other hand, you only need to call this function in the child
787 process if and only if you want to use the event loop in the child. If
788 you just fork+exec or create a new loop in the child, you don't have to
789 call it at all (in fact, C<epoll> is so badly broken that it makes a
790 difference, but libev will usually detect this case on its own and do a
791 costly reset of the backend).
793 The function itself is quite fast and it's usually not a problem to call
794 it just in case after a fork.
796 Example: Automate calling C<ev_loop_fork> on the default loop when
797 using pthreads.
799 static void
800 post_fork_child (void)
801 {
802 ev_loop_fork (EV_DEFAULT);
803 }
805 ...
806 pthread_atfork (0, 0, post_fork_child);
808 =item int ev_is_default_loop (loop)
810 Returns true when the given loop is, in fact, the default loop, and false
811 otherwise.
813 =item unsigned int ev_iteration (loop)
815 Returns the current iteration count for the event loop, which is identical
816 to the number of times libev did poll for new events. It starts at C<0>
817 and happily wraps around with enough iterations.
819 This value can sometimes be useful as a generation counter of sorts (it
820 "ticks" the number of loop iterations), as it roughly corresponds with
821 C<ev_prepare> and C<ev_check> calls - and is incremented between the
822 prepare and check phases.
824 =item unsigned int ev_depth (loop)
826 Returns the number of times C<ev_run> was entered minus the number of
827 times C<ev_run> was exited normally, in other words, the recursion depth.
829 Outside C<ev_run>, this number is zero. In a callback, this number is
830 C<1>, unless C<ev_run> was invoked recursively (or from another thread),
831 in which case it is higher.
833 Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
834 throwing an exception etc.), doesn't count as "exit" - consider this
835 as a hint to avoid such ungentleman-like behaviour unless it's really
836 convenient, in which case it is fully supported.
838 =item unsigned int ev_backend (loop)
840 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
841 use.
843 =item ev_tstamp ev_now (loop)
845 Returns the current "event loop time", which is the time the event loop
846 received events and started processing them. This timestamp does not
847 change as long as callbacks are being processed, and this is also the base
848 time used for relative timers. You can treat it as the timestamp of the
849 event occurring (or more correctly, libev finding out about it).
851 =item ev_now_update (loop)
853 Establishes the current time by querying the kernel, updating the time
854 returned by C<ev_now ()> in the progress. This is a costly operation and
855 is usually done automatically within C<ev_run ()>.
857 This function is rarely useful, but when some event callback runs for a
858 very long time without entering the event loop, updating libev's idea of
859 the current time is a good idea.
861 See also L</The special problem of time updates> in the C<ev_timer> section.
863 =item ev_suspend (loop)
865 =item ev_resume (loop)
867 These two functions suspend and resume an event loop, for use when the
868 loop is not used for a while and timeouts should not be processed.
870 A typical use case would be an interactive program such as a game: When
871 the user presses C<^Z> to suspend the game and resumes it an hour later it
872 would be best to handle timeouts as if no time had actually passed while
873 the program was suspended. This can be achieved by calling C<ev_suspend>
874 in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
875 C<ev_resume> directly afterwards to resume timer processing.
877 Effectively, all C<ev_timer> watchers will be delayed by the time spend
878 between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
879 will be rescheduled (that is, they will lose any events that would have
880 occurred while suspended).
882 After calling C<ev_suspend> you B<must not> call I<any> function on the
883 given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
884 without a previous call to C<ev_suspend>.
886 Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
887 event loop time (see C<ev_now_update>).
889 =item bool ev_run (loop, int flags)
891 Finally, this is it, the event handler. This function usually is called
892 after you have initialised all your watchers and you want to start
893 handling events. It will ask the operating system for any new events, call
894 the watcher callbacks, and then repeat the whole process indefinitely: This
895 is why event loops are called I<loops>.
897 If the flags argument is specified as C<0>, it will keep handling events
898 until either no event watchers are active anymore or C<ev_break> was
899 called.
901 The return value is false if there are no more active watchers (which
902 usually means "all jobs done" or "deadlock"), and true in all other cases
903 (which usually means " you should call C<ev_run> again").
905 Please note that an explicit C<ev_break> is usually better than
906 relying on all watchers to be stopped when deciding when a program has
907 finished (especially in interactive programs), but having a program
908 that automatically loops as long as it has to and no longer by virtue
909 of relying on its watchers stopping correctly, that is truly a thing of
910 beauty.
912 This function is I<mostly> exception-safe - you can break out of a
913 C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
914 exception and so on. This does not decrement the C<ev_depth> value, nor
915 will it clear any outstanding C<EVBREAK_ONE> breaks.
917 A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
918 those events and any already outstanding ones, but will not wait and
919 block your process in case there are no events and will return after one
920 iteration of the loop. This is sometimes useful to poll and handle new
921 events while doing lengthy calculations, to keep the program responsive.
923 A flags value of C<EVRUN_ONCE> will look for new events (waiting if
924 necessary) and will handle those and any already outstanding ones. It
925 will block your process until at least one new event arrives (which could
926 be an event internal to libev itself, so there is no guarantee that a
927 user-registered callback will be called), and will return after one
928 iteration of the loop.
930 This is useful if you are waiting for some external event in conjunction
931 with something not expressible using other libev watchers (i.e. "roll your
932 own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
933 usually a better approach for this kind of thing.
935 Here are the gory details of what C<ev_run> does (this is for your
936 understanding, not a guarantee that things will work exactly like this in
937 future versions):
939 - Increment loop depth.
940 - Reset the ev_break status.
941 - Before the first iteration, call any pending watchers.
942 LOOP:
943 - If EVFLAG_FORKCHECK was used, check for a fork.
944 - If a fork was detected (by any means), queue and call all fork watchers.
945 - Queue and call all prepare watchers.
946 - If ev_break was called, goto FINISH.
947 - If we have been forked, detach and recreate the kernel state
948 as to not disturb the other process.
949 - Update the kernel state with all outstanding changes.
950 - Update the "event loop time" (ev_now ()).
951 - Calculate for how long to sleep or block, if at all
952 (active idle watchers, EVRUN_NOWAIT or not having
953 any active watchers at all will result in not sleeping).
954 - Sleep if the I/O and timer collect interval say so.
955 - Increment loop iteration counter.
956 - Block the process, waiting for any events.
957 - Queue all outstanding I/O (fd) events.
958 - Update the "event loop time" (ev_now ()), and do time jump adjustments.
959 - Queue all expired timers.
960 - Queue all expired periodics.
961 - Queue all idle watchers with priority higher than that of pending events.
962 - Queue all check watchers.
963 - Call all queued watchers in reverse order (i.e. check watchers first).
964 Signals and child watchers are implemented as I/O watchers, and will
965 be handled here by queueing them when their watcher gets executed.
966 - If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
967 were used, or there are no active watchers, goto FINISH, otherwise
968 continue with step LOOP.
970 - Reset the ev_break status iff it was EVBREAK_ONE.
971 - Decrement the loop depth.
972 - Return.
974 Example: Queue some jobs and then loop until no events are outstanding
975 anymore.
977 ... queue jobs here, make sure they register event watchers as long
978 ... as they still have work to do (even an idle watcher will do..)
979 ev_run (my_loop, 0);
980 ... jobs done or somebody called break. yeah!
982 =item ev_break (loop, how)
984 Can be used to make a call to C<ev_run> return early (but only after it
985 has processed all outstanding events). The C<how> argument must be either
986 C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
987 C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
989 This "break state" will be cleared on the next call to C<ev_run>.
991 It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
992 which case it will have no effect.
994 =item ev_ref (loop)
996 =item ev_unref (loop)
998 Ref/unref can be used to add or remove a reference count on the event
999 loop: Every watcher keeps one reference, and as long as the reference
1000 count is nonzero, C<ev_run> will not return on its own.
1002 This is useful when you have a watcher that you never intend to
1003 unregister, but that nevertheless should not keep C<ev_run> from
1004 returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
1005 before stopping it.
1007 As an example, libev itself uses this for its internal signal pipe: It
1008 is not visible to the libev user and should not keep C<ev_run> from
1009 exiting if no event watchers registered by it are active. It is also an
1010 excellent way to do this for generic recurring timers or from within
1011 third-party libraries. Just remember to I<unref after start> and I<ref
1012 before stop> (but only if the watcher wasn't active before, or was active
1013 before, respectively. Note also that libev might stop watchers itself
1014 (e.g. non-repeating timers) in which case you have to C<ev_ref>
1015 in the callback).
1017 Example: Create a signal watcher, but keep it from keeping C<ev_run>
1018 running when nothing else is active.
1020 ev_signal exitsig;
1021 ev_signal_init (&exitsig, sig_cb, SIGINT);
1022 ev_signal_start (loop, &exitsig);
1023 ev_unref (loop);
1025 Example: For some weird reason, unregister the above signal handler again.
1027 ev_ref (loop);
1028 ev_signal_stop (loop, &exitsig);
1030 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
1032 =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
1034 These advanced functions influence the time that libev will spend waiting
1035 for events. Both time intervals are by default C<0>, meaning that libev
1036 will try to invoke timer/periodic callbacks and I/O callbacks with minimum
1037 latency.
1039 Setting these to a higher value (the C<interval> I<must> be >= C<0>)
1040 allows libev to delay invocation of I/O and timer/periodic callbacks
1041 to increase efficiency of loop iterations (or to increase power-saving
1042 opportunities).
1044 The idea is that sometimes your program runs just fast enough to handle
1045 one (or very few) event(s) per loop iteration. While this makes the
1046 program responsive, it also wastes a lot of CPU time to poll for new
1047 events, especially with backends like C<select ()> which have a high
1048 overhead for the actual polling but can deliver many events at once.
1050 By setting a higher I<io collect interval> you allow libev to spend more
1051 time collecting I/O events, so you can handle more events per iteration,
1052 at the cost of increasing latency. Timeouts (both C<ev_periodic> and
1053 C<ev_timer>) will not be affected. Setting this to a non-null value will
1054 introduce an additional C<ev_sleep ()> call into most loop iterations. The
1055 sleep time ensures that libev will not poll for I/O events more often then
1056 once per this interval, on average (as long as the host time resolution is
1057 good enough).
1059 Likewise, by setting a higher I<timeout collect interval> you allow libev
1060 to spend more time collecting timeouts, at the expense of increased
1061 latency/jitter/inexactness (the watcher callback will be called
1062 later). C<ev_io> watchers will not be affected. Setting this to a non-null
1063 value will not introduce any overhead in libev.
1065 Many (busy) programs can usually benefit by setting the I/O collect
1066 interval to a value near C<0.1> or so, which is often enough for
1067 interactive servers (of course not for games), likewise for timeouts. It
1068 usually doesn't make much sense to set it to a lower value than C<0.01>,
1069 as this approaches the timing granularity of most systems. Note that if
1070 you do transactions with the outside world and you can't increase the
1071 parallelity, then this setting will limit your transaction rate (if you
1072 need to poll once per transaction and the I/O collect interval is 0.01,
1073 then you can't do more than 100 transactions per second).
1075 Setting the I<timeout collect interval> can improve the opportunity for
1076 saving power, as the program will "bundle" timer callback invocations that
1077 are "near" in time together, by delaying some, thus reducing the number of
1078 times the process sleeps and wakes up again. Another useful technique to
1079 reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
1080 they fire on, say, one-second boundaries only.
1082 Example: we only need 0.1s timeout granularity, and we wish not to poll
1083 more often than 100 times per second:
1085 ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
1086 ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
1088 =item ev_invoke_pending (loop)
1090 This call will simply invoke all pending watchers while resetting their
1091 pending state. Normally, C<ev_run> does this automatically when required,
1092 but when overriding the invoke callback this call comes handy. This
1093 function can be invoked from a watcher - this can be useful for example
1094 when you want to do some lengthy calculation and want to pass further
1095 event handling to another thread (you still have to make sure only one
1096 thread executes within C<ev_invoke_pending> or C<ev_run> of course).
1098 =item int ev_pending_count (loop)
1100 Returns the number of pending watchers - zero indicates that no watchers
1101 are pending.
1103 =item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
1105 This overrides the invoke pending functionality of the loop: Instead of
1106 invoking all pending watchers when there are any, C<ev_run> will call
1107 this callback instead. This is useful, for example, when you want to
1108 invoke the actual watchers inside another context (another thread etc.).
1110 If you want to reset the callback, use C<ev_invoke_pending> as new
1111 callback.
1113 =item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
1115 Sometimes you want to share the same loop between multiple threads. This
1116 can be done relatively simply by putting mutex_lock/unlock calls around
1117 each call to a libev function.
1119 However, C<ev_run> can run an indefinite time, so it is not feasible
1120 to wait for it to return. One way around this is to wake up the event
1121 loop via C<ev_break> and C<ev_async_send>, another way is to set these
1122 I<release> and I<acquire> callbacks on the loop.
1124 When set, then C<release> will be called just before the thread is
1125 suspended waiting for new events, and C<acquire> is called just
1126 afterwards.
1128 Ideally, C<release> will just call your mutex_unlock function, and
1129 C<acquire> will just call the mutex_lock function again.
1131 While event loop modifications are allowed between invocations of
1132 C<release> and C<acquire> (that's their only purpose after all), no
1133 modifications done will affect the event loop, i.e. adding watchers will
1134 have no effect on the set of file descriptors being watched, or the time
1135 waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
1136 to take note of any changes you made.
1138 In theory, threads executing C<ev_run> will be async-cancel safe between
1139 invocations of C<release> and C<acquire>.
1141 See also the locking example in the C<THREADS> section later in this
1142 document.
1144 =item ev_set_userdata (loop, void *data)
1146 =item void *ev_userdata (loop)
1148 Set and retrieve a single C<void *> associated with a loop. When
1149 C<ev_set_userdata> has never been called, then C<ev_userdata> returns
1150 C<0>.
1152 These two functions can be used to associate arbitrary data with a loop,
1153 and are intended solely for the C<invoke_pending_cb>, C<release> and
1154 C<acquire> callbacks described above, but of course can be (ab-)used for
1155 any other purpose as well.
1157 =item ev_verify (loop)
1159 This function only does something when C<EV_VERIFY> support has been
1160 compiled in, which is the default for non-minimal builds. It tries to go
1161 through all internal structures and checks them for validity. If anything
1162 is found to be inconsistent, it will print an error message to standard
1163 error and call C<abort ()>.
1165 This can be used to catch bugs inside libev itself: under normal
1166 circumstances, this function will never abort as of course libev keeps its
1167 data structures consistent.
1169 =back
1174 In the following description, uppercase C<TYPE> in names stands for the
1175 watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
1176 watchers and C<ev_io_start> for I/O watchers.
1178 A watcher is an opaque structure that you allocate and register to record
1179 your interest in some event. To make a concrete example, imagine you want
1180 to wait for STDIN to become readable, you would create an C<ev_io> watcher
1181 for that:
1183 static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
1184 {
1185 ev_io_stop (w);
1186 ev_break (loop, EVBREAK_ALL);
1187 }
1189 struct ev_loop *loop = ev_default_loop (0);
1191 ev_io stdin_watcher;
1193 ev_init (&stdin_watcher, my_cb);
1194 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
1195 ev_io_start (loop, &stdin_watcher);
1197 ev_run (loop, 0);
1199 As you can see, you are responsible for allocating the memory for your
1200 watcher structures (and it is I<usually> a bad idea to do this on the
1201 stack).
1203 Each watcher has an associated watcher structure (called C<struct ev_TYPE>
1204 or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
1206 Each watcher structure must be initialised by a call to C<ev_init (watcher
1207 *, callback)>, which expects a callback to be provided. This callback is
1208 invoked each time the event occurs (or, in the case of I/O watchers, each
1209 time the event loop detects that the file descriptor given is readable
1210 and/or writable).
1212 Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
1213 macro to configure it, with arguments specific to the watcher type. There
1214 is also a macro to combine initialisation and setting in one call: C<<
1215 ev_TYPE_init (watcher *, callback, ...) >>.
1217 To make the watcher actually watch out for events, you have to start it
1218 with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
1219 *) >>), and you can stop watching for events at any time by calling the
1220 corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
1222 As long as your watcher is active (has been started but not stopped) you
1223 must not touch the values stored in it. Most specifically you must never
1224 reinitialise it or call its C<ev_TYPE_set> macro.
1226 Each and every callback receives the event loop pointer as first, the
1227 registered watcher structure as second, and a bitset of received events as
1228 third argument.
1230 The received events usually include a single bit per event type received
1231 (you can receive multiple events at the same time). The possible bit masks
1232 are:
1234 =over 4
1236 =item C<EV_READ>
1238 =item C<EV_WRITE>
1240 The file descriptor in the C<ev_io> watcher has become readable and/or
1241 writable.
1243 =item C<EV_TIMER>
1245 The C<ev_timer> watcher has timed out.
1247 =item C<EV_PERIODIC>
1249 The C<ev_periodic> watcher has timed out.
1251 =item C<EV_SIGNAL>
1253 The signal specified in the C<ev_signal> watcher has been received by a thread.
1255 =item C<EV_CHILD>
1257 The pid specified in the C<ev_child> watcher has received a status change.
1259 =item C<EV_STAT>
1261 The path specified in the C<ev_stat> watcher changed its attributes somehow.
1263 =item C<EV_IDLE>
1265 The C<ev_idle> watcher has determined that you have nothing better to do.
1267 =item C<EV_PREPARE>
1269 =item C<EV_CHECK>
1271 All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
1272 gather new events, and all C<ev_check> watchers are queued (not invoked)
1273 just after C<ev_run> has gathered them, but before it queues any callbacks
1274 for any received events. That means C<ev_prepare> watchers are the last
1275 watchers invoked before the event loop sleeps or polls for new events, and
1276 C<ev_check> watchers will be invoked before any other watchers of the same
1277 or lower priority within an event loop iteration.
1279 Callbacks of both watcher types can start and stop as many watchers as
1280 they want, and all of them will be taken into account (for example, a
1281 C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
1282 blocking).
1284 =item C<EV_EMBED>
1286 The embedded event loop specified in the C<ev_embed> watcher needs attention.
1288 =item C<EV_FORK>
1290 The event loop has been resumed in the child process after fork (see
1291 C<ev_fork>).
1293 =item C<EV_CLEANUP>
1295 The event loop is about to be destroyed (see C<ev_cleanup>).
1297 =item C<EV_ASYNC>
1299 The given async watcher has been asynchronously notified (see C<ev_async>).
1301 =item C<EV_CUSTOM>
1303 Not ever sent (or otherwise used) by libev itself, but can be freely used
1304 by libev users to signal watchers (e.g. via C<ev_feed_event>).
1306 =item C<EV_ERROR>
1308 An unspecified error has occurred, the watcher has been stopped. This might
1309 happen because the watcher could not be properly started because libev
1310 ran out of memory, a file descriptor was found to be closed or any other
1311 problem. Libev considers these application bugs.
1313 You best act on it by reporting the problem and somehow coping with the
1314 watcher being stopped. Note that well-written programs should not receive
1315 an error ever, so when your watcher receives it, this usually indicates a
1316 bug in your program.
1318 Libev will usually signal a few "dummy" events together with an error, for
1319 example it might indicate that a fd is readable or writable, and if your
1320 callbacks is well-written it can just attempt the operation and cope with
1321 the error from read() or write(). This will not work in multi-threaded
1322 programs, though, as the fd could already be closed and reused for another
1323 thing, so beware.
1325 =back
1329 =over 4
1331 =item C<ev_init> (ev_TYPE *watcher, callback)
1333 This macro initialises the generic portion of a watcher. The contents
1334 of the watcher object can be arbitrary (so C<malloc> will do). Only
1335 the generic parts of the watcher are initialised, you I<need> to call
1336 the type-specific C<ev_TYPE_set> macro afterwards to initialise the
1337 type-specific parts. For each type there is also a C<ev_TYPE_init> macro
1338 which rolls both calls into one.
1340 You can reinitialise a watcher at any time as long as it has been stopped
1341 (or never started) and there are no pending events outstanding.
1343 The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
1344 int revents)>.
1346 Example: Initialise an C<ev_io> watcher in two steps.
1348 ev_io w;
1349 ev_init (&w, my_cb);
1350 ev_io_set (&w, STDIN_FILENO, EV_READ);
1352 =item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
1354 This macro initialises the type-specific parts of a watcher. You need to
1355 call C<ev_init> at least once before you call this macro, but you can
1356 call C<ev_TYPE_set> any number of times. You must not, however, call this
1357 macro on a watcher that is active (it can be pending, however, which is a
1358 difference to the C<ev_init> macro).
1360 Although some watcher types do not have type-specific arguments
1361 (e.g. C<ev_prepare>) you still need to call its C<set> macro.
1363 See C<ev_init>, above, for an example.
1365 =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
1367 This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
1368 calls into a single call. This is the most convenient method to initialise
1369 a watcher. The same limitations apply, of course.
1371 Example: Initialise and set an C<ev_io> watcher in one step.
1373 ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
1375 =item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
1377 Starts (activates) the given watcher. Only active watchers will receive
1378 events. If the watcher is already active nothing will happen.
1380 Example: Start the C<ev_io> watcher that is being abused as example in this
1381 whole section.
1383 ev_io_start (EV_DEFAULT_UC, &w);
1385 =item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
1387 Stops the given watcher if active, and clears the pending status (whether
1388 the watcher was active or not).
1390 It is possible that stopped watchers are pending - for example,
1391 non-repeating timers are being stopped when they become pending - but
1392 calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
1393 pending. If you want to free or reuse the memory used by the watcher it is
1394 therefore a good idea to always call its C<ev_TYPE_stop> function.
1396 =item bool ev_is_active (ev_TYPE *watcher)
1398 Returns a true value iff the watcher is active (i.e. it has been started
1399 and not yet been stopped). As long as a watcher is active you must not modify
1400 it.
1402 =item bool ev_is_pending (ev_TYPE *watcher)
1404 Returns a true value iff the watcher is pending, (i.e. it has outstanding
1405 events but its callback has not yet been invoked). As long as a watcher
1406 is pending (but not active) you must not call an init function on it (but
1407 C<ev_TYPE_set> is safe), you must not change its priority, and you must
1408 make sure the watcher is available to libev (e.g. you cannot C<free ()>
1409 it).
1411 =item callback ev_cb (ev_TYPE *watcher)
1413 Returns the callback currently set on the watcher.
1415 =item ev_set_cb (ev_TYPE *watcher, callback)
1417 Change the callback. You can change the callback at virtually any time
1418 (modulo threads).
1420 =item ev_set_priority (ev_TYPE *watcher, int priority)
1422 =item int ev_priority (ev_TYPE *watcher)
1424 Set and query the priority of the watcher. The priority is a small
1425 integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
1426 (default: C<-2>). Pending watchers with higher priority will be invoked
1427 before watchers with lower priority, but priority will not keep watchers
1428 from being executed (except for C<ev_idle> watchers).
1430 If you need to suppress invocation when higher priority events are pending
1431 you need to look at C<ev_idle> watchers, which provide this functionality.
1433 You I<must not> change the priority of a watcher as long as it is active or
1434 pending.
1436 Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
1437 fine, as long as you do not mind that the priority value you query might
1438 or might not have been clamped to the valid range.
1440 The default priority used by watchers when no priority has been set is
1441 always C<0>, which is supposed to not be too high and not be too low :).
1443 See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1444 priorities.
1446 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
1448 Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1449 C<loop> nor C<revents> need to be valid as long as the watcher callback
1450 can deal with that fact, as both are simply passed through to the
1451 callback.
1453 =item int ev_clear_pending (loop, ev_TYPE *watcher)
1455 If the watcher is pending, this function clears its pending status and
1456 returns its C<revents> bitset (as if its callback was invoked). If the
1457 watcher isn't pending it does nothing and returns C<0>.
1459 Sometimes it can be useful to "poll" a watcher instead of waiting for its
1460 callback to be invoked, which can be accomplished with this function.
1462 =item ev_feed_event (loop, ev_TYPE *watcher, int revents)
1464 Feeds the given event set into the event loop, as if the specified event
1465 had happened for the specified watcher (which must be a pointer to an
1466 initialised but not necessarily started event watcher). Obviously you must
1467 not free the watcher as long as it has pending events.
1469 Stopping the watcher, letting libev invoke it, or calling
1470 C<ev_clear_pending> will clear the pending event, even if the watcher was
1471 not started in the first place.
1473 See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1474 functions that do not need a watcher.
1476 =back
1481 =head2 WATCHER STATES
1483 There are various watcher states mentioned throughout this manual -
1484 active, pending and so on. In this section these states and the rules to
1485 transition between them will be described in more detail - and while these
1486 rules might look complicated, they usually do "the right thing".
1488 =over 4
1490 =item initialised
1492 Before a watcher can be registered with the event loop it has to be
1493 initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1494 C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1496 In this state it is simply some block of memory that is suitable for
1497 use in an event loop. It can be moved around, freed, reused etc. at
1498 will - as long as you either keep the memory contents intact, or call
1499 C<ev_TYPE_init> again.
1501 =item started/running/active
1503 Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1504 property of the event loop, and is actively waiting for events. While in
1505 this state it cannot be accessed (except in a few documented ways), moved,
1506 freed or anything else - the only legal thing is to keep a pointer to it,
1507 and call libev functions on it that are documented to work on active watchers.
1509 =item pending
1511 If a watcher is active and libev determines that an event it is interested
1512 in has occurred (such as a timer expiring), it will become pending. It will
1513 stay in this pending state until either it is stopped or its callback is
1514 about to be invoked, so it is not normally pending inside the watcher
1515 callback.
1517 The watcher might or might not be active while it is pending (for example,
1518 an expired non-repeating timer can be pending but no longer active). If it
1519 is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
1520 but it is still property of the event loop at this time, so cannot be
1521 moved, freed or reused. And if it is active the rules described in the
1522 previous item still apply.
1524 It is also possible to feed an event on a watcher that is not active (e.g.
1525 via C<ev_feed_event>), in which case it becomes pending without being
1526 active.
1528 =item stopped
1530 A watcher can be stopped implicitly by libev (in which case it might still
1531 be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
1532 latter will clear any pending state the watcher might be in, regardless
1533 of whether it was active or not, so stopping a watcher explicitly before
1534 freeing it is often a good idea.
1536 While stopped (and not pending) the watcher is essentially in the
1537 initialised state, that is, it can be reused, moved, modified in any way
1538 you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
1539 it again).
1541 =back
1545 Many event loops support I<watcher priorities>, which are usually small
1546 integers that influence the ordering of event callback invocation
1547 between watchers in some way, all else being equal.
1549 In libev, watcher priorities can be set using C<ev_set_priority>. See its
1550 description for the more technical details such as the actual priority
1551 range.
1553 There are two common ways how these these priorities are being interpreted
1554 by event loops:
1556 In the more common lock-out model, higher priorities "lock out" invocation
1557 of lower priority watchers, which means as long as higher priority
1558 watchers receive events, lower priority watchers are not being invoked.
1560 The less common only-for-ordering model uses priorities solely to order
1561 callback invocation within a single event loop iteration: Higher priority
1562 watchers are invoked before lower priority ones, but they all get invoked
1563 before polling for new events.
1565 Libev uses the second (only-for-ordering) model for all its watchers
1566 except for idle watchers (which use the lock-out model).
1568 The rationale behind this is that implementing the lock-out model for
1569 watchers is not well supported by most kernel interfaces, and most event
1570 libraries will just poll for the same events again and again as long as
1571 their callbacks have not been executed, which is very inefficient in the
1572 common case of one high-priority watcher locking out a mass of lower
1573 priority ones.
1575 Static (ordering) priorities are most useful when you have two or more
1576 watchers handling the same resource: a typical usage example is having an
1577 C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
1578 timeouts. Under load, data might be received while the program handles
1579 other jobs, but since timers normally get invoked first, the timeout
1580 handler will be executed before checking for data. In that case, giving
1581 the timer a lower priority than the I/O watcher ensures that I/O will be
1582 handled first even under adverse conditions (which is usually, but not
1583 always, what you want).
1585 Since idle watchers use the "lock-out" model, meaning that idle watchers
1586 will only be executed when no same or higher priority watchers have
1587 received events, they can be used to implement the "lock-out" model when
1588 required.
1590 For example, to emulate how many other event libraries handle priorities,
1591 you can associate an C<ev_idle> watcher to each such watcher, and in
1592 the normal watcher callback, you just start the idle watcher. The real
1593 processing is done in the idle watcher callback. This causes libev to
1594 continuously poll and process kernel event data for the watcher, but when
1595 the lock-out case is known to be rare (which in turn is rare :), this is
1596 workable.
1598 Usually, however, the lock-out model implemented that way will perform
1599 miserably under the type of load it was designed to handle. In that case,
1600 it might be preferable to stop the real watcher before starting the
1601 idle watcher, so the kernel will not have to process the event in case
1602 the actual processing will be delayed for considerable time.
1604 Here is an example of an I/O watcher that should run at a strictly lower
1605 priority than the default, and which should only process data when no
1606 other events are pending:
1608 ev_idle idle; // actual processing watcher
1609 ev_io io; // actual event watcher
1611 static void
1612 io_cb (EV_P_ ev_io *w, int revents)
1613 {
1614 // stop the I/O watcher, we received the event, but
1615 // are not yet ready to handle it.
1616 ev_io_stop (EV_A_ w);
1618 // start the idle watcher to handle the actual event.
1619 // it will not be executed as long as other watchers
1620 // with the default priority are receiving events.
1621 ev_idle_start (EV_A_ &idle);
1622 }
1624 static void
1625 idle_cb (EV_P_ ev_idle *w, int revents)
1626 {
1627 // actual processing
1628 read (STDIN_FILENO, ...);
1630 // have to start the I/O watcher again, as
1631 // we have handled the event
1632 ev_io_start (EV_P_ &io);
1633 }
1635 // initialisation
1636 ev_idle_init (&idle, idle_cb);
1637 ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
1638 ev_io_start (EV_DEFAULT_ &io);
1640 In the "real" world, it might also be beneficial to start a timer, so that
1641 low-priority connections can not be locked out forever under load. This
1642 enables your program to keep a lower latency for important connections
1643 during short periods of high load, while not completely locking out less
1644 important ones.
1647 =head1 WATCHER TYPES
1649 This section describes each watcher in detail, but will not repeat
1650 information given in the last section. Any initialisation/set macros,
1651 functions and members specific to the watcher type are explained.
1653 Members are additionally marked with either I<[read-only]>, meaning that,
1654 while the watcher is active, you can look at the member and expect some
1655 sensible content, but you must not modify it (you can modify it while the
1656 watcher is stopped to your hearts content), or I<[read-write]>, which
1657 means you can expect it to have some sensible content while the watcher
1658 is active, but you can also modify it. Modifying it may not do something
1659 sensible or take immediate effect (or do anything at all), but libev will
1660 not crash or malfunction in any way.
1663 =head2 C<ev_io> - is this file descriptor readable or writable?
1665 I/O watchers check whether a file descriptor is readable or writable
1666 in each iteration of the event loop, or, more precisely, when reading
1667 would not block the process and writing would at least be able to write
1668 some data. This behaviour is called level-triggering because you keep
1669 receiving events as long as the condition persists. Remember you can stop
1670 the watcher if you don't want to act on the event and neither want to
1671 receive future events.
1673 In general you can register as many read and/or write event watchers per
1674 fd as you want (as long as you don't confuse yourself). Setting all file
1675 descriptors to non-blocking mode is also usually a good idea (but not
1676 required if you know what you are doing).
1678 Another thing you have to watch out for is that it is quite easy to
1679 receive "spurious" readiness notifications, that is, your callback might
1680 be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1681 because there is no data. It is very easy to get into this situation even
1682 with a relatively standard program structure. Thus it is best to always
1683 use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1684 preferable to a program hanging until some data arrives.
1686 If you cannot run the fd in non-blocking mode (for example you should
1687 not play around with an Xlib connection), then you have to separately
1688 re-test whether a file descriptor is really ready with a known-to-be good
1689 interface such as poll (fortunately in the case of Xlib, it already does
1690 this on its own, so its quite safe to use). Some people additionally
1691 use C<SIGALRM> and an interval timer, just to be sure you won't block
1692 indefinitely.
1694 But really, best use non-blocking mode.
1696 =head3 The special problem of disappearing file descriptors
1698 Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
1699 a file descriptor (either due to calling C<close> explicitly or any other
1700 means, such as C<dup2>). The reason is that you register interest in some
1701 file descriptor, but when it goes away, the operating system will silently
1702 drop this interest. If another file descriptor with the same number then
1703 is registered with libev, there is no efficient way to see that this is,
1704 in fact, a different file descriptor.
1706 To avoid having to explicitly tell libev about such cases, libev follows
1707 the following policy: Each time C<ev_io_set> is being called, libev
1708 will assume that this is potentially a new file descriptor, otherwise
1709 it is assumed that the file descriptor stays the same. That means that
1710 you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1711 descriptor even if the file descriptor number itself did not change.
1713 This is how one would do it normally anyway, the important point is that
1714 the libev application should not optimise around libev but should leave
1715 optimisations to libev.
1717 =head3 The special problem of dup'ed file descriptors
1719 Some backends (e.g. epoll), cannot register events for file descriptors,
1720 but only events for the underlying file descriptions. That means when you
1721 have C<dup ()>'ed file descriptors or weirder constellations, and register
1722 events for them, only one file descriptor might actually receive events.
1724 There is no workaround possible except not registering events
1725 for potentially C<dup ()>'ed file descriptors, or to resort to
1728 =head3 The special problem of files
1730 Many people try to use C<select> (or libev) on file descriptors
1731 representing files, and expect it to become ready when their program
1732 doesn't block on disk accesses (which can take a long time on their own).
1734 However, this cannot ever work in the "expected" way - you get a readiness
1735 notification as soon as the kernel knows whether and how much data is
1736 there, and in the case of open files, that's always the case, so you
1737 always get a readiness notification instantly, and your read (or possibly
1738 write) will still block on the disk I/O.
1740 Another way to view it is that in the case of sockets, pipes, character
1741 devices and so on, there is another party (the sender) that delivers data
1742 on its own, but in the case of files, there is no such thing: the disk
1743 will not send data on its own, simply because it doesn't know what you
1744 wish to read - you would first have to request some data.
1746 Since files are typically not-so-well supported by advanced notification
1747 mechanism, libev tries hard to emulate POSIX behaviour with respect
1748 to files, even though you should not use it. The reason for this is
1749 convenience: sometimes you want to watch STDIN or STDOUT, which is
1750 usually a tty, often a pipe, but also sometimes files or special devices
1751 (for example, C<epoll> on Linux works with F</dev/random> but not with
1752 F</dev/urandom>), and even though the file might better be served with
1753 asynchronous I/O instead of with non-blocking I/O, it is still useful when
1754 it "just works" instead of freezing.
1756 So avoid file descriptors pointing to files when you know it (e.g. use
1757 libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
1758 when you rarely read from a file instead of from a socket, and want to
1759 reuse the same code path.
1761 =head3 The special problem of fork
1763 Some backends (epoll, kqueue, linuxaio, iouring) do not support C<fork ()>
1764 at all or exhibit useless behaviour. Libev fully supports fork, but needs
1765 to be told about it in the child if you want to continue to use it in the
1766 child.
1768 To support fork in your child processes, you have to call C<ev_loop_fork
1769 ()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1772 =head3 The special problem of SIGPIPE
1774 While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1775 when writing to a pipe whose other end has been closed, your program gets
1776 sent a SIGPIPE, which, by default, aborts your program. For most programs
1777 this is sensible behaviour, for daemons, this is usually undesirable.
1779 So when you encounter spurious, unexplained daemon exits, make sure you
1780 ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1781 somewhere, as that would have given you a big clue).
1783 =head3 The special problem of accept()ing when you can't
1785 Many implementations of the POSIX C<accept> function (for example,
1786 found in post-2004 Linux) have the peculiar behaviour of not removing a
1787 connection from the pending queue in all error cases.
1789 For example, larger servers often run out of file descriptors (because
1790 of resource limits), causing C<accept> to fail with C<ENFILE> but not
1791 rejecting the connection, leading to libev signalling readiness on
1792 the next iteration again (the connection still exists after all), and
1793 typically causing the program to loop at 100% CPU usage.
1795 Unfortunately, the set of errors that cause this issue differs between
1796 operating systems, there is usually little the app can do to remedy the
1797 situation, and no known thread-safe method of removing the connection to
1798 cope with overload is known (to me).
1800 One of the easiest ways to handle this situation is to just ignore it
1801 - when the program encounters an overload, it will just loop until the
1802 situation is over. While this is a form of busy waiting, no OS offers an
1803 event-based way to handle this situation, so it's the best one can do.
1805 A better way to handle the situation is to log any errors other than
1806 C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
1807 messages, and continue as usual, which at least gives the user an idea of
1808 what could be wrong ("raise the ulimit!"). For extra points one could stop
1809 the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
1810 usage.
1812 If your program is single-threaded, then you could also keep a dummy file
1813 descriptor for overload situations (e.g. by opening F</dev/null>), and
1814 when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
1815 close that fd, and create a new dummy fd. This will gracefully refuse
1816 clients under typical overload conditions.
1818 The last way to handle it is to simply log the error and C<exit>, as
1819 is often done with C<malloc> failures, but this results in an easy
1820 opportunity for a DoS attack.
1822 =head3 Watcher-Specific Functions
1824 =over 4
1826 =item ev_io_init (ev_io *, callback, int fd, int events)
1828 =item ev_io_set (ev_io *, int fd, int events)
1830 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1831 receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
1832 C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
1834 =item int fd [read-only]
1836 The file descriptor being watched.
1838 =item int events [read-only]
1840 The events being watched.
1842 =back
1844 =head3 Examples
1846 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1847 readable, but only once. Since it is likely line-buffered, you could
1848 attempt to read a whole line in the callback.
1850 static void
1851 stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
1852 {
1853 ev_io_stop (loop, w);
1854 .. read from stdin here (or from w->fd) and handle any I/O errors
1855 }
1857 ...
1858 struct ev_loop *loop = ev_default_init (0);
1859 ev_io stdin_readable;
1860 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1861 ev_io_start (loop, &stdin_readable);
1862 ev_run (loop, 0);
1865 =head2 C<ev_timer> - relative and optionally repeating timeouts
1867 Timer watchers are simple relative timers that generate an event after a
1868 given time, and optionally repeating in regular intervals after that.
1870 The timers are based on real time, that is, if you register an event that
1871 times out after an hour and you reset your system clock to January last
1872 year, it will still time out after (roughly) one hour. "Roughly" because
1873 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1874 monotonic clock option helps a lot here).
1876 The callback is guaranteed to be invoked only I<after> its timeout has
1877 passed (not I<at>, so on systems with very low-resolution clocks this
1878 might introduce a small delay, see "the special problem of being too
1879 early", below). If multiple timers become ready during the same loop
1880 iteration then the ones with earlier time-out values are invoked before
1881 ones of the same priority with later time-out values (but this is no
1882 longer true when a callback calls C<ev_run> recursively).
1884 =head3 Be smart about timeouts
1886 Many real-world problems involve some kind of timeout, usually for error
1887 recovery. A typical example is an HTTP request - if the other side hangs,
1888 you want to raise some error after a while.
1890 What follows are some ways to handle this problem, from obvious and
1891 inefficient to smart and efficient.
1893 In the following, a 60 second activity timeout is assumed - a timeout that
1894 gets reset to 60 seconds each time there is activity (e.g. each time some
1895 data or other life sign was received).
1897 =over 4
1899 =item 1. Use a timer and stop, reinitialise and start it on activity.
1901 This is the most obvious, but not the most simple way: In the beginning,
1902 start the watcher:
1904 ev_timer_init (timer, callback, 60., 0.);
1905 ev_timer_start (loop, timer);
1907 Then, each time there is some activity, C<ev_timer_stop> it, initialise it
1908 and start it again:
1910 ev_timer_stop (loop, timer);
1911 ev_timer_set (timer, 60., 0.);
1912 ev_timer_start (loop, timer);
1914 This is relatively simple to implement, but means that each time there is
1915 some activity, libev will first have to remove the timer from its internal
1916 data structure and then add it again. Libev tries to be fast, but it's
1917 still not a constant-time operation.
1919 =item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
1921 This is the easiest way, and involves using C<ev_timer_again> instead of
1922 C<ev_timer_start>.
1924 To implement this, configure an C<ev_timer> with a C<repeat> value
1925 of C<60> and then call C<ev_timer_again> at start and each time you
1926 successfully read or write some data. If you go into an idle state where
1927 you do not expect data to travel on the socket, you can C<ev_timer_stop>
1928 the timer, and C<ev_timer_again> will automatically restart it if need be.
1930 That means you can ignore both the C<ev_timer_start> function and the
1931 C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
1932 member and C<ev_timer_again>.
1934 At start:
1936 ev_init (timer, callback);
1937 timer->repeat = 60.;
1938 ev_timer_again (loop, timer);
1940 Each time there is some activity:
1942 ev_timer_again (loop, timer);
1944 It is even possible to change the time-out on the fly, regardless of
1945 whether the watcher is active or not:
1947 timer->repeat = 30.;
1948 ev_timer_again (loop, timer);
1950 This is slightly more efficient then stopping/starting the timer each time
1951 you want to modify its timeout value, as libev does not have to completely
1952 remove and re-insert the timer from/into its internal data structure.
1954 It is, however, even simpler than the "obvious" way to do it.
1956 =item 3. Let the timer time out, but then re-arm it as required.
1958 This method is more tricky, but usually most efficient: Most timeouts are
1959 relatively long compared to the intervals between other activity - in
1960 our example, within 60 seconds, there are usually many I/O events with
1961 associated activity resets.
1963 In this case, it would be more efficient to leave the C<ev_timer> alone,
1964 but remember the time of last activity, and check for a real timeout only
1965 within the callback:
1967 ev_tstamp timeout = 60.;
1968 ev_tstamp last_activity; // time of last activity
1969 ev_timer timer;
1971 static void
1972 callback (EV_P_ ev_timer *w, int revents)
1973 {
1974 // calculate when the timeout would happen
1975 ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
1977 // if negative, it means we the timeout already occurred
1978 if (after < 0.)
1979 {
1980 // timeout occurred, take action
1981 }
1982 else
1983 {
1984 // callback was invoked, but there was some recent
1985 // activity. simply restart the timer to time out
1986 // after "after" seconds, which is the earliest time
1987 // the timeout can occur.
1988 ev_timer_set (w, after, 0.);
1989 ev_timer_start (EV_A_ w);
1990 }
1991 }
1993 To summarise the callback: first calculate in how many seconds the
1994 timeout will occur (by calculating the absolute time when it would occur,
1995 C<last_activity + timeout>, and subtracting the current time, C<ev_now
1996 (EV_A)> from that).
1998 If this value is negative, then we are already past the timeout, i.e. we
1999 timed out, and need to do whatever is needed in this case.
2001 Otherwise, we now the earliest time at which the timeout would trigger,
2002 and simply start the timer with this timeout value.
2004 In other words, each time the callback is invoked it will check whether
2005 the timeout occurred. If not, it will simply reschedule itself to check
2006 again at the earliest time it could time out. Rinse. Repeat.
2008 This scheme causes more callback invocations (about one every 60 seconds
2009 minus half the average time between activity), but virtually no calls to
2010 libev to change the timeout.
2012 To start the machinery, simply initialise the watcher and set
2013 C<last_activity> to the current time (meaning there was some activity just
2014 now), then call the callback, which will "do the right thing" and start
2015 the timer:
2017 last_activity = ev_now (EV_A);
2018 ev_init (&timer, callback);
2019 callback (EV_A_ &timer, 0);
2021 When there is some activity, simply store the current time in
2022 C<last_activity>, no libev calls at all:
2024 if (activity detected)
2025 last_activity = ev_now (EV_A);
2027 When your timeout value changes, then the timeout can be changed by simply
2028 providing a new value, stopping the timer and calling the callback, which
2029 will again do the right thing (for example, time out immediately :).
2031 timeout = new_value;
2032 ev_timer_stop (EV_A_ &timer);
2033 callback (EV_A_ &timer, 0);
2035 This technique is slightly more complex, but in most cases where the
2036 time-out is unlikely to be triggered, much more efficient.
2038 =item 4. Wee, just use a double-linked list for your timeouts.
2040 If there is not one request, but many thousands (millions...), all
2041 employing some kind of timeout with the same timeout value, then one can
2042 do even better:
2044 When starting the timeout, calculate the timeout value and put the timeout
2045 at the I<end> of the list.
2047 Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
2048 the list is expected to fire (for example, using the technique #3).
2050 When there is some activity, remove the timer from the list, recalculate
2051 the timeout, append it to the end of the list again, and make sure to
2052 update the C<ev_timer> if it was taken from the beginning of the list.
2054 This way, one can manage an unlimited number of timeouts in O(1) time for
2055 starting, stopping and updating the timers, at the expense of a major
2056 complication, and having to use a constant timeout. The constant timeout
2057 ensures that the list stays sorted.
2059 =back
2061 So which method the best?
2063 Method #2 is a simple no-brain-required solution that is adequate in most
2064 situations. Method #3 requires a bit more thinking, but handles many cases
2065 better, and isn't very complicated either. In most case, choosing either
2066 one is fine, with #3 being better in typical situations.
2068 Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
2069 rather complicated, but extremely efficient, something that really pays
2070 off after the first million or so of active timers, i.e. it's usually
2071 overkill :)
2073 =head3 The special problem of being too early
2075 If you ask a timer to call your callback after three seconds, then
2076 you expect it to be invoked after three seconds - but of course, this
2077 cannot be guaranteed to infinite precision. Less obviously, it cannot be
2078 guaranteed to any precision by libev - imagine somebody suspending the
2079 process with a STOP signal for a few hours for example.
2081 So, libev tries to invoke your callback as soon as possible I<after> the
2082 delay has occurred, but cannot guarantee this.
2084 A less obvious failure mode is calling your callback too early: many event
2085 loops compare timestamps with a "elapsed delay >= requested delay", but
2086 this can cause your callback to be invoked much earlier than you would
2087 expect.
2089 To see why, imagine a system with a clock that only offers full second
2090 resolution (think windows if you can't come up with a broken enough OS
2091 yourself). If you schedule a one-second timer at the time 500.9, then the
2092 event loop will schedule your timeout to elapse at a system time of 500
2093 (500.9 truncated to the resolution) + 1, or 501.
2095 If an event library looks at the timeout 0.1s later, it will see "501 >=
2096 501" and invoke the callback 0.1s after it was started, even though a
2097 one-second delay was requested - this is being "too early", despite best
2098 intentions.
2100 This is the reason why libev will never invoke the callback if the elapsed
2101 delay equals the requested delay, but only when the elapsed delay is
2102 larger than the requested delay. In the example above, libev would only invoke
2103 the callback at system time 502, or 1.1s after the timer was started.
2105 So, while libev cannot guarantee that your callback will be invoked
2106 exactly when requested, it I<can> and I<does> guarantee that the requested
2107 delay has actually elapsed, or in other words, it always errs on the "too
2108 late" side of things.
2110 =head3 The special problem of time updates
2112 Establishing the current time is a costly operation (it usually takes
2113 at least one system call): EV therefore updates its idea of the current
2114 time only before and after C<ev_run> collects new events, which causes a
2115 growing difference between C<ev_now ()> and C<ev_time ()> when handling
2116 lots of events in one iteration.
2118 The relative timeouts are calculated relative to the C<ev_now ()>
2119 time. This is usually the right thing as this timestamp refers to the time
2120 of the event triggering whatever timeout you are modifying/starting. If
2121 you suspect event processing to be delayed and you I<need> to base the
2122 timeout on the current time, use something like the following to adjust
2123 for it:
2125 ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
2127 If the event loop is suspended for a long time, you can also force an
2128 update of the time returned by C<ev_now ()> by calling C<ev_now_update
2129 ()>, although that will push the event time of all outstanding events
2130 further into the future.
2132 =head3 The special problem of unsynchronised clocks
2134 Modern systems have a variety of clocks - libev itself uses the normal
2135 "wall clock" clock and, if available, the monotonic clock (to avoid time
2136 jumps).
2138 Neither of these clocks is synchronised with each other or any other clock
2139 on the system, so C<ev_time ()> might return a considerably different time
2140 than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
2141 a call to C<gettimeofday> might return a second count that is one higher
2142 than a directly following call to C<time>.
2144 The moral of this is to only compare libev-related timestamps with
2145 C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
2146 a second or so.
2148 One more problem arises due to this lack of synchronisation: if libev uses
2149 the system monotonic clock and you compare timestamps from C<ev_time>
2150 or C<ev_now> from when you started your timer and when your callback is
2151 invoked, you will find that sometimes the callback is a bit "early".
2153 This is because C<ev_timer>s work in real time, not wall clock time, so
2154 libev makes sure your callback is not invoked before the delay happened,
2155 I<measured according to the real time>, not the system clock.
2157 If your timeouts are based on a physical timescale (e.g. "time out this
2158 connection after 100 seconds") then this shouldn't bother you as it is
2159 exactly the right behaviour.
2161 If you want to compare wall clock/system timestamps to your timers, then
2162 you need to use C<ev_periodic>s, as these are based on the wall clock
2163 time, where your comparisons will always generate correct results.
2165 =head3 The special problems of suspended animation
2167 When you leave the server world it is quite customary to hit machines that
2168 can suspend/hibernate - what happens to the clocks during such a suspend?
2170 Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
2171 all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
2172 to run until the system is suspended, but they will not advance while the
2173 system is suspended. That means, on resume, it will be as if the program
2174 was frozen for a few seconds, but the suspend time will not be counted
2175 towards C<ev_timer> when a monotonic clock source is used. The real time
2176 clock advanced as expected, but if it is used as sole clocksource, then a
2177 long suspend would be detected as a time jump by libev, and timers would
2178 be adjusted accordingly.
2180 I would not be surprised to see different behaviour in different between
2181 operating systems, OS versions or even different hardware.
2183 The other form of suspend (job control, or sending a SIGSTOP) will see a
2184 time jump in the monotonic clocks and the realtime clock. If the program
2185 is suspended for a very long time, and monotonic clock sources are in use,
2186 then you can expect C<ev_timer>s to expire as the full suspension time
2187 will be counted towards the timers. When no monotonic clock source is in
2188 use, then libev will again assume a timejump and adjust accordingly.
2190 It might be beneficial for this latter case to call C<ev_suspend>
2191 and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
2192 deterministic behaviour in this case (you can do nothing against
2193 C<SIGSTOP>).
2195 =head3 Watcher-Specific Functions and Data Members
2197 =over 4
2199 =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
2201 =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
2203 Configure the timer to trigger after C<after> seconds (fractional and
2204 negative values are supported). If C<repeat> is C<0.>, then it will
2205 automatically be stopped once the timeout is reached. If it is positive,
2206 then the timer will automatically be configured to trigger again C<repeat>
2207 seconds later, again, and again, until stopped manually.
2209 The timer itself will do a best-effort at avoiding drift, that is, if
2210 you configure a timer to trigger every 10 seconds, then it will normally
2211 trigger at exactly 10 second intervals. If, however, your program cannot
2212 keep up with the timer (because it takes longer than those 10 seconds to
2213 do stuff) the timer will not fire more than once per event loop iteration.
2215 =item ev_timer_again (loop, ev_timer *)
2217 This will act as if the timer timed out, and restarts it again if it is
2218 repeating. It basically works like calling C<ev_timer_stop>, updating the
2219 timeout to the C<repeat> value and calling C<ev_timer_start>.
2221 The exact semantics are as in the following rules, all of which will be
2222 applied to the watcher:
2224 =over 4
2226 =item If the timer is pending, the pending status is always cleared.
2228 =item If the timer is started but non-repeating, stop it (as if it timed
2229 out, without invoking it).
2231 =item If the timer is repeating, make the C<repeat> value the new timeout
2232 and start the timer, if necessary.
2234 =back
2236 This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
2237 usage example.
2239 =item ev_tstamp ev_timer_remaining (loop, ev_timer *)
2241 Returns the remaining time until a timer fires. If the timer is active,
2242 then this time is relative to the current event loop time, otherwise it's
2243 the timeout value currently configured.
2245 That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
2246 C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
2247 will return C<4>. When the timer expires and is restarted, it will return
2248 roughly C<7> (likely slightly less as callback invocation takes some time,
2249 too), and so on.
2251 =item ev_tstamp repeat [read-write]
2253 The current C<repeat> value. Will be used each time the watcher times out
2254 or C<ev_timer_again> is called, and determines the next timeout (if any),
2255 which is also when any modifications are taken into account.
2257 =back
2259 =head3 Examples
2261 Example: Create a timer that fires after 60 seconds.
2263 static void
2264 one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
2265 {
2266 .. one minute over, w is actually stopped right here
2267 }
2269 ev_timer mytimer;
2270 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
2271 ev_timer_start (loop, &mytimer);
2273 Example: Create a timeout timer that times out after 10 seconds of
2274 inactivity.
2276 static void
2277 timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
2278 {
2279 .. ten seconds without any activity
2280 }
2282 ev_timer mytimer;
2283 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
2284 ev_timer_again (&mytimer); /* start timer */
2285 ev_run (loop, 0);
2287 // and in some piece of code that gets executed on any "activity":
2288 // reset the timeout to start ticking again at 10 seconds
2289 ev_timer_again (&mytimer);
2292 =head2 C<ev_periodic> - to cron or not to cron?
2294 Periodic watchers are also timers of a kind, but they are very versatile
2295 (and unfortunately a bit complex).
2297 Unlike C<ev_timer>, periodic watchers are not based on real time (or
2298 relative time, the physical time that passes) but on wall clock time
2299 (absolute time, the thing you can read on your calendar or clock). The
2300 difference is that wall clock time can run faster or slower than real
2301 time, and time jumps are not uncommon (e.g. when you adjust your
2302 wrist-watch).
2304 You can tell a periodic watcher to trigger after some specific point
2305 in time: for example, if you tell a periodic watcher to trigger "in 10
2306 seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
2307 not a delay) and then reset your system clock to January of the previous
2308 year, then it will take a year or more to trigger the event (unlike an
2309 C<ev_timer>, which would still trigger roughly 10 seconds after starting
2310 it, as it uses a relative timeout).
2312 C<ev_periodic> watchers can also be used to implement vastly more complex
2313 timers, such as triggering an event on each "midnight, local time", or
2314 other complicated rules. This cannot easily be done with C<ev_timer>
2315 watchers, as those cannot react to time jumps.
2317 As with timers, the callback is guaranteed to be invoked only when the
2318 point in time where it is supposed to trigger has passed. If multiple
2319 timers become ready during the same loop iteration then the ones with
2320 earlier time-out values are invoked before ones with later time-out values
2321 (but this is no longer true when a callback calls C<ev_run> recursively).
2323 =head3 Watcher-Specific Functions and Data Members
2325 =over 4
2327 =item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
2329 =item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
2331 Lots of arguments, let's sort it out... There are basically three modes of
2332 operation, and we will explain them from simplest to most complex:
2334 =over 4
2336 =item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
2338 In this configuration the watcher triggers an event after the wall clock
2339 time C<offset> has passed. It will not repeat and will not adjust when a
2340 time jump occurs, that is, if it is to be run at January 1st 2011 then it
2341 will be stopped and invoked when the system clock reaches or surpasses
2342 this point in time.
2344 =item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
2346 In this mode the watcher will always be scheduled to time out at the next
2347 C<offset + N * interval> time (for some integer N, which can also be
2348 negative) and then repeat, regardless of any time jumps. The C<offset>
2349 argument is merely an offset into the C<interval> periods.
2351 This can be used to create timers that do not drift with respect to the
2352 system clock, for example, here is an C<ev_periodic> that triggers each
2353 hour, on the hour (with respect to UTC):
2355 ev_periodic_set (&periodic, 0., 3600., 0);
2357 This doesn't mean there will always be 3600 seconds in between triggers,
2358 but only that the callback will be called when the system time shows a
2359 full hour (UTC), or more correctly, when the system time is evenly divisible
2360 by 3600.
2362 Another way to think about it (for the mathematically inclined) is that
2363 C<ev_periodic> will try to run the callback in this mode at the next possible
2364 time where C<time = offset (mod interval)>, regardless of any time jumps.
2366 The C<interval> I<MUST> be positive, and for numerical stability, the
2367 interval value should be higher than C<1/8192> (which is around 100
2368 microseconds) and C<offset> should be higher than C<0> and should have
2369 at most a similar magnitude as the current time (say, within a factor of
2370 ten). Typical values for offset are, in fact, C<0> or something between
2371 C<0> and C<interval>, which is also the recommended range.
2373 Note also that there is an upper limit to how often a timer can fire (CPU
2374 speed for example), so if C<interval> is very small then timing stability
2375 will of course deteriorate. Libev itself tries to be exact to be about one
2376 millisecond (if the OS supports it and the machine is fast enough).
2378 =item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
2380 In this mode the values for C<interval> and C<offset> are both being
2381 ignored. Instead, each time the periodic watcher gets scheduled, the
2382 reschedule callback will be called with the watcher as first, and the
2383 current time as second argument.
2385 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
2386 or make ANY other event loop modifications whatsoever, unless explicitly
2387 allowed by documentation here>.
2389 If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
2390 it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
2391 only event loop modification you are allowed to do).
2393 The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
2394 *w, ev_tstamp now)>, e.g.:
2396 static ev_tstamp
2397 my_rescheduler (ev_periodic *w, ev_tstamp now)
2398 {
2399 return now + 60.;
2400 }
2402 It must return the next time to trigger, based on the passed time value
2403 (that is, the lowest time value larger than to the second argument). It
2404 will usually be called just before the callback will be triggered, but
2405 might be called at other times, too.
2407 NOTE: I<< This callback must always return a time that is higher than or
2408 equal to the passed C<now> value >>.
2410 This can be used to create very complex timers, such as a timer that
2411 triggers on "next midnight, local time". To do this, you would calculate
2412 the next midnight after C<now> and return the timestamp value for
2413 this. Here is a (completely untested, no error checking) example on how to
2414 do this:
2416 #include <time.h>
2418 static ev_tstamp
2419 my_rescheduler (ev_periodic *w, ev_tstamp now)
2420 {
2421 time_t tnow = (time_t)now;
2422 struct tm tm;
2423 localtime_r (&tnow, &tm);
2425 tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
2426 ++tm.tm_mday; // midnight next day
2428 return mktime (&tm);
2429 }
2431 Note: this code might run into trouble on days that have more then two
2432 midnights (beginning and end).
2434 =back
2436 =item ev_periodic_again (loop, ev_periodic *)
2438 Simply stops and restarts the periodic watcher again. This is only useful
2439 when you changed some parameters or the reschedule callback would return
2440 a different time than the last time it was called (e.g. in a crond like
2441 program when the crontabs have changed).
2443 =item ev_tstamp ev_periodic_at (ev_periodic *)
2445 When active, returns the absolute time that the watcher is supposed
2446 to trigger next. This is not the same as the C<offset> argument to
2447 C<ev_periodic_set>, but indeed works even in interval and manual
2448 rescheduling modes.
2450 =item ev_tstamp offset [read-write]
2452 When repeating, this contains the offset value, otherwise this is the
2453 absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
2454 although libev might modify this value for better numerical stability).
2456 Can be modified any time, but changes only take effect when the periodic
2457 timer fires or C<ev_periodic_again> is being called.
2459 =item ev_tstamp interval [read-write]
2461 The current interval value. Can be modified any time, but changes only
2462 take effect when the periodic timer fires or C<ev_periodic_again> is being
2463 called.
2465 =item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
2467 The current reschedule callback, or C<0>, if this functionality is
2468 switched off. Can be changed any time, but changes only take effect when
2469 the periodic timer fires or C<ev_periodic_again> is being called.
2471 =back
2473 =head3 Examples
2475 Example: Call a callback every hour, or, more precisely, whenever the
2476 system time is divisible by 3600. The callback invocation times have
2477 potentially a lot of jitter, but good long-term stability.
2479 static void
2480 clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
2481 {
2482 ... its now a full hour (UTC, or TAI or whatever your clock follows)
2483 }
2485 ev_periodic hourly_tick;
2486 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
2487 ev_periodic_start (loop, &hourly_tick);
2489 Example: The same as above, but use a reschedule callback to do it:
2491 #include <math.h>
2493 static ev_tstamp
2494 my_scheduler_cb (ev_periodic *w, ev_tstamp now)
2495 {
2496 return now + (3600. - fmod (now, 3600.));
2497 }
2499 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
2501 Example: Call a callback every hour, starting now:
2503 ev_periodic hourly_tick;
2504 ev_periodic_init (&hourly_tick, clock_cb,
2505 fmod (ev_now (loop), 3600.), 3600., 0);
2506 ev_periodic_start (loop, &hourly_tick);
2509 =head2 C<ev_signal> - signal me when a signal gets signalled!
2511 Signal watchers will trigger an event when the process receives a specific
2512 signal one or more times. Even though signals are very asynchronous, libev
2513 will try its best to deliver signals synchronously, i.e. as part of the
2514 normal event processing, like any other event.
2516 If you want signals to be delivered truly asynchronously, just use
2517 C<sigaction> as you would do without libev and forget about sharing
2518 the signal. You can even use C<ev_async> from a signal handler to
2519 synchronously wake up an event loop.
2521 You can configure as many watchers as you like for the same signal, but
2522 only within the same loop, i.e. you can watch for C<SIGINT> in your
2523 default loop and for C<SIGIO> in another loop, but you cannot watch for
2524 C<SIGINT> in both the default loop and another loop at the same time. At
2525 the moment, C<SIGCHLD> is permanently tied to the default loop.
2527 Only after the first watcher for a signal is started will libev actually
2528 register something with the kernel. It thus coexists with your own signal
2529 handlers as long as you don't register any with libev for the same signal.
2531 If possible and supported, libev will install its handlers with
2532 C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
2533 not be unduly interrupted. If you have a problem with system calls getting
2534 interrupted by signals you can block all signals in an C<ev_check> watcher
2535 and unblock them in an C<ev_prepare> watcher.
2537 =head3 The special problem of inheritance over fork/execve/pthread_create
2539 Both the signal mask (C<sigprocmask>) and the signal disposition
2540 (C<sigaction>) are unspecified after starting a signal watcher (and after
2541 stopping it again), that is, libev might or might not block the signal,
2542 and might or might not set or restore the installed signal handler (but
2545 While this does not matter for the signal disposition (libev never
2546 sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2547 C<execve>), this matters for the signal mask: many programs do not expect
2548 certain signals to be blocked.
2550 This means that before calling C<exec> (from the child) you should reset
2551 the signal mask to whatever "default" you expect (all clear is a good
2552 choice usually).
2554 The simplest way to ensure that the signal mask is reset in the child is
2555 to install a fork handler with C<pthread_atfork> that resets it. That will
2556 catch fork calls done by libraries (such as the libc) as well.
2558 In current versions of libev, the signal will not be blocked indefinitely
2559 unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
2560 the window of opportunity for problems, it will not go away, as libev
2561 I<has> to modify the signal mask, at least temporarily.
2563 So I can't stress this enough: I<If you do not reset your signal mask when
2564 you expect it to be empty, you have a race condition in your code>. This
2565 is not a libev-specific thing, this is true for most event libraries.
2567 =head3 The special problem of threads signal handling
2569 POSIX threads has problematic signal handling semantics, specifically,
2570 a lot of functionality (sigfd, sigwait etc.) only really works if all
2571 threads in a process block signals, which is hard to achieve.
2573 When you want to use sigwait (or mix libev signal handling with your own
2574 for the same signals), you can tackle this problem by globally blocking
2575 all signals before creating any threads (or creating them with a fully set
2576 sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
2577 loops. Then designate one thread as "signal receiver thread" which handles
2578 these signals. You can pass on any signals that libev might be interested
2579 in by calling C<ev_feed_signal>.
2581 =head3 Watcher-Specific Functions and Data Members
2583 =over 4
2585 =item ev_signal_init (ev_signal *, callback, int signum)
2587 =item ev_signal_set (ev_signal *, int signum)
2589 Configures the watcher to trigger on the given signal number (usually one
2590 of the C<SIGxxx> constants).
2592 =item int signum [read-only]
2594 The signal the watcher watches out for.
2596 =back
2598 =head3 Examples
2600 Example: Try to exit cleanly on SIGINT.
2602 static void
2603 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
2604 {
2605 ev_break (loop, EVBREAK_ALL);
2606 }
2608 ev_signal signal_watcher;
2609 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
2610 ev_signal_start (loop, &signal_watcher);
2613 =head2 C<ev_child> - watch out for process status changes
2615 Child watchers trigger when your process receives a SIGCHLD in response to
2616 some child status changes (most typically when a child of yours dies or
2617 exits). It is permissible to install a child watcher I<after> the child
2618 has been forked (which implies it might have already exited), as long
2619 as the event loop isn't entered (or is continued from a watcher), i.e.,
2620 forking and then immediately registering a watcher for the child is fine,
2621 but forking and registering a watcher a few event loop iterations later or
2622 in the next callback invocation is not.
2624 Only the default event loop is capable of handling signals, and therefore
2625 you can only register child watchers in the default event loop.
2627 Due to some design glitches inside libev, child watchers will always be
2628 handled at maximum priority (their priority is set to C<EV_MAXPRI> by
2629 libev)
2631 =head3 Process Interaction
2633 Libev grabs C<SIGCHLD> as soon as the default event loop is
2634 initialised. This is necessary to guarantee proper behaviour even if the
2635 first child watcher is started after the child exits. The occurrence
2636 of C<SIGCHLD> is recorded asynchronously, but child reaping is done
2637 synchronously as part of the event loop processing. Libev always reaps all
2638 children, even ones not watched.
2640 =head3 Overriding the Built-In Processing
2642 Libev offers no special support for overriding the built-in child
2643 processing, but if your application collides with libev's default child
2644 handler, you can override it easily by installing your own handler for
2645 C<SIGCHLD> after initialising the default loop, and making sure the
2646 default loop never gets destroyed. You are encouraged, however, to use an
2647 event-based approach to child reaping and thus use libev's support for
2648 that, so other libev users can use C<ev_child> watchers freely.
2650 =head3 Stopping the Child Watcher
2652 Currently, the child watcher never gets stopped, even when the
2653 child terminates, so normally one needs to stop the watcher in the
2654 callback. Future versions of libev might stop the watcher automatically
2655 when a child exit is detected (calling C<ev_child_stop> twice is not a
2656 problem).
2658 =head3 Watcher-Specific Functions and Data Members
2660 =over 4
2662 =item ev_child_init (ev_child *, callback, int pid, int trace)
2664 =item ev_child_set (ev_child *, int pid, int trace)
2666 Configures the watcher to wait for status changes of process C<pid> (or
2667 I<any> process if C<pid> is specified as C<0>). The callback can look
2668 at the C<rstatus> member of the C<ev_child> watcher structure to see
2669 the status word (use the macros from C<sys/wait.h> and see your systems
2670 C<waitpid> documentation). The C<rpid> member contains the pid of the
2671 process causing the status change. C<trace> must be either C<0> (only
2672 activate the watcher when the process terminates) or C<1> (additionally
2673 activate the watcher when the process is stopped or continued).
2675 =item int pid [read-only]
2677 The process id this watcher watches out for, or C<0>, meaning any process id.
2679 =item int rpid [read-write]
2681 The process id that detected a status change.
2683 =item int rstatus [read-write]
2685 The process exit/trace status caused by C<rpid> (see your systems
2686 C<waitpid> and C<sys/wait.h> documentation for details).
2688 =back
2690 =head3 Examples
2692 Example: C<fork()> a new process and install a child handler to wait for
2693 its completion.
2695 ev_child cw;
2697 static void
2698 child_cb (EV_P_ ev_child *w, int revents)
2699 {
2700 ev_child_stop (EV_A_ w);
2701 printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
2702 }
2704 pid_t pid = fork ();
2706 if (pid < 0)
2707 // error
2708 else if (pid == 0)
2709 {
2710 // the forked child executes here
2711 exit (1);
2712 }
2713 else
2714 {
2715 ev_child_init (&cw, child_cb, pid, 0);
2716 ev_child_start (EV_DEFAULT_ &cw);
2717 }
2720 =head2 C<ev_stat> - did the file attributes just change?
2722 This watches a file system path for attribute changes. That is, it calls
2723 C<stat> on that path in regular intervals (or when the OS says it changed)
2724 and sees if it changed compared to the last time, invoking the callback
2725 if it did. Starting the watcher C<stat>'s the file, so only changes that
2726 happen after the watcher has been started will be reported.
2728 The path does not need to exist: changing from "path exists" to "path does
2729 not exist" is a status change like any other. The condition "path does not
2730 exist" (or more correctly "path cannot be stat'ed") is signified by the
2731 C<st_nlink> field being zero (which is otherwise always forced to be at
2732 least one) and all the other fields of the stat buffer having unspecified
2733 contents.
2735 The path I<must not> end in a slash or contain special components such as
2736 C<.> or C<..>. The path I<should> be absolute: If it is relative and
2737 your working directory changes, then the behaviour is undefined.
2739 Since there is no portable change notification interface available, the
2740 portable implementation simply calls C<stat(2)> regularly on the path
2741 to see if it changed somehow. You can specify a recommended polling
2742 interval for this case. If you specify a polling interval of C<0> (highly
2743 recommended!) then a I<suitable, unspecified default> value will be used
2744 (which you can expect to be around five seconds, although this might
2745 change dynamically). Libev will also impose a minimum interval which is
2746 currently around C<0.1>, but that's usually overkill.
2748 This watcher type is not meant for massive numbers of stat watchers,
2749 as even with OS-supported change notifications, this can be
2750 resource-intensive.
2752 At the time of this writing, the only OS-specific interface implemented
2753 is the Linux inotify interface (implementing kqueue support is left as an
2754 exercise for the reader. Note, however, that the author sees no way of
2755 implementing C<ev_stat> semantics with kqueue, except as a hint).
2757 =head3 ABI Issues (Largefile Support)
2759 Libev by default (unless the user overrides this) uses the default
2760 compilation environment, which means that on systems with large file
2761 support disabled by default, you get the 32 bit version of the stat
2762 structure. When using the library from programs that change the ABI to
2763 use 64 bit file offsets the programs will fail. In that case you have to
2764 compile libev with the same flags to get binary compatibility. This is
2765 obviously the case with any flags that change the ABI, but the problem is
2766 most noticeably displayed with ev_stat and large file support.
2768 The solution for this is to lobby your distribution maker to make large
2769 file interfaces available by default (as e.g. FreeBSD does) and not
2770 optional. Libev cannot simply switch on large file support because it has
2771 to exchange stat structures with application programs compiled using the
2772 default compilation environment.
2774 =head3 Inotify and Kqueue
2776 When C<inotify (7)> support has been compiled into libev and present at
2777 runtime, it will be used to speed up change detection where possible. The
2778 inotify descriptor will be created lazily when the first C<ev_stat>
2779 watcher is being started.
2781 Inotify presence does not change the semantics of C<ev_stat> watchers
2782 except that changes might be detected earlier, and in some cases, to avoid
2783 making regular C<stat> calls. Even in the presence of inotify support
2784 there are many cases where libev has to resort to regular C<stat> polling,
2785 but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
2786 many bugs), the path exists (i.e. stat succeeds), and the path resides on
2787 a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
2788 xfs are fully working) libev usually gets away without polling.
2790 There is no support for kqueue, as apparently it cannot be used to
2791 implement this functionality, due to the requirement of having a file
2792 descriptor open on the object at all times, and detecting renames, unlinks
2793 etc. is difficult.
2795 =head3 C<stat ()> is a synchronous operation
2797 Libev doesn't normally do any kind of I/O itself, and so is not blocking
2798 the process. The exception are C<ev_stat> watchers - those call C<stat
2799 ()>, which is a synchronous operation.
2801 For local paths, this usually doesn't matter: unless the system is very
2802 busy or the intervals between stat's are large, a stat call will be fast,
2803 as the path data is usually in memory already (except when starting the
2804 watcher).
2806 For networked file systems, calling C<stat ()> can block an indefinite
2807 time due to network issues, and even under good conditions, a stat call
2808 often takes multiple milliseconds.
2810 Therefore, it is best to avoid using C<ev_stat> watchers on networked
2811 paths, although this is fully supported by libev.
2813 =head3 The special problem of stat time resolution
2815 The C<stat ()> system call only supports full-second resolution portably,
2816 and even on systems where the resolution is higher, most file systems
2817 still only support whole seconds.
2819 That means that, if the time is the only thing that changes, you can
2820 easily miss updates: on the first update, C<ev_stat> detects a change and
2821 calls your callback, which does something. When there is another update
2822 within the same second, C<ev_stat> will be unable to detect unless the
2823 stat data does change in other ways (e.g. file size).
2825 The solution to this is to delay acting on a change for slightly more
2826 than a second (or till slightly after the next full second boundary), using
2827 a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
2828 ev_timer_again (loop, w)>).
2830 The C<.02> offset is added to work around small timing inconsistencies
2831 of some operating systems (where the second counter of the current time
2832 might be be delayed. One such system is the Linux kernel, where a call to
2833 C<gettimeofday> might return a timestamp with a full second later than
2834 a subsequent C<time> call - if the equivalent of C<time ()> is used to
2835 update file times then there will be a small window where the kernel uses
2836 the previous second to update file times but libev might already execute
2837 the timer callback).
2839 =head3 Watcher-Specific Functions and Data Members
2841 =over 4
2843 =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
2845 =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
2847 Configures the watcher to wait for status changes of the given
2848 C<path>. The C<interval> is a hint on how quickly a change is expected to
2849 be detected and should normally be specified as C<0> to let libev choose
2850 a suitable value. The memory pointed to by C<path> must point to the same
2851 path for as long as the watcher is active.
2853 The callback will receive an C<EV_STAT> event when a change was detected,
2854 relative to the attributes at the time the watcher was started (or the
2855 last change was detected).
2857 =item ev_stat_stat (loop, ev_stat *)
2859 Updates the stat buffer immediately with new values. If you change the
2860 watched path in your callback, you could call this function to avoid
2861 detecting this change (while introducing a race condition if you are not
2862 the only one changing the path). Can also be useful simply to find out the
2863 new values.
2865 =item ev_statdata attr [read-only]
2867 The most-recently detected attributes of the file. Although the type is
2868 C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
2869 suitable for your system, but you can only rely on the POSIX-standardised
2870 members to be present. If the C<st_nlink> member is C<0>, then there was
2871 some error while C<stat>ing the file.
2873 =item ev_statdata prev [read-only]
2875 The previous attributes of the file. The callback gets invoked whenever
2876 C<prev> != C<attr>, or, more precisely, one or more of these members
2877 differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
2878 C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
2880 =item ev_tstamp interval [read-only]
2882 The specified interval.
2884 =item const char *path [read-only]
2886 The file system path that is being watched.
2888 =back
2890 =head3 Examples
2892 Example: Watch C</etc/passwd> for attribute changes.
2894 static void
2895 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
2896 {
2897 /* /etc/passwd changed in some way */
2898 if (w->attr.st_nlink)
2899 {
2900 printf ("passwd current size %ld\n", (long)w->attr.st_size);
2901 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
2902 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
2903 }
2904 else
2905 /* you shalt not abuse printf for puts */
2906 puts ("wow, /etc/passwd is not there, expect problems. "
2907 "if this is windows, they already arrived\n");
2908 }
2910 ...
2911 ev_stat passwd;
2913 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
2914 ev_stat_start (loop, &passwd);
2916 Example: Like above, but additionally use a one-second delay so we do not
2917 miss updates (however, frequent updates will delay processing, too, so
2918 one might do the work both on C<ev_stat> callback invocation I<and> on
2919 C<ev_timer> callback invocation).
2921 static ev_stat passwd;
2922 static ev_timer timer;
2924 static void
2925 timer_cb (EV_P_ ev_timer *w, int revents)
2926 {
2927 ev_timer_stop (EV_A_ w);
2929 /* now it's one second after the most recent passwd change */
2930 }
2932 static void
2933 stat_cb (EV_P_ ev_stat *w, int revents)
2934 {
2935 /* reset the one-second timer */
2936 ev_timer_again (EV_A_ &timer);
2937 }
2939 ...
2940 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2941 ev_stat_start (loop, &passwd);
2942 ev_timer_init (&timer, timer_cb, 0., 1.02);
2945 =head2 C<ev_idle> - when you've got nothing better to do...
2947 Idle watchers trigger events when no other events of the same or higher
2948 priority are pending (prepare, check and other idle watchers do not count
2949 as receiving "events").
2951 That is, as long as your process is busy handling sockets or timeouts
2952 (or even signals, imagine) of the same or higher priority it will not be
2953 triggered. But when your process is idle (or only lower-priority watchers
2954 are pending), the idle watchers are being called once per event loop
2955 iteration - until stopped, that is, or your process receives more events
2956 and becomes busy again with higher priority stuff.
2958 The most noteworthy effect is that as long as any idle watchers are
2959 active, the process will not block when waiting for new events.
2961 Apart from keeping your process non-blocking (which is a useful
2962 effect on its own sometimes), idle watchers are a good place to do
2963 "pseudo-background processing", or delay processing stuff to after the
2964 event loop has handled all outstanding events.
2966 =head3 Abusing an C<ev_idle> watcher for its side-effect
2968 As long as there is at least one active idle watcher, libev will never
2969 sleep unnecessarily. Or in other words, it will loop as fast as possible.
2970 For this to work, the idle watcher doesn't need to be invoked at all - the
2971 lowest priority will do.
2973 This mode of operation can be useful together with an C<ev_check> watcher,
2974 to do something on each event loop iteration - for example to balance load
2975 between different connections.
2977 See L</Abusing an ev_check watcher for its side-effect> for a longer
2978 example.
2980 =head3 Watcher-Specific Functions and Data Members
2982 =over 4
2984 =item ev_idle_init (ev_idle *, callback)
2986 Initialises and configures the idle watcher - it has no parameters of any
2987 kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
2988 believe me.
2990 =back
2992 =head3 Examples
2994 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
2995 callback, free it. Also, use no error checking, as usual.
2997 static void
2998 idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
2999 {
3000 // stop the watcher
3001 ev_idle_stop (loop, w);
3003 // now we can free it
3004 free (w);
3006 // now do something you wanted to do when the program has
3007 // no longer anything immediate to do.
3008 }
3010 ev_idle *idle_watcher = malloc (sizeof (ev_idle));
3011 ev_idle_init (idle_watcher, idle_cb);
3012 ev_idle_start (loop, idle_watcher);
3015 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
3017 Prepare and check watchers are often (but not always) used in pairs:
3018 prepare watchers get invoked before the process blocks and check watchers
3019 afterwards.
3021 You I<must not> call C<ev_run> (or similar functions that enter the
3022 current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
3023 C<ev_check> watchers. Other loops than the current one are fine,
3024 however. The rationale behind this is that you do not need to check
3025 for recursion in those watchers, i.e. the sequence will always be
3026 C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
3027 kind they will always be called in pairs bracketing the blocking call.
3029 Their main purpose is to integrate other event mechanisms into libev and
3030 their use is somewhat advanced. They could be used, for example, to track
3031 variable changes, implement your own watchers, integrate net-snmp or a
3032 coroutine library and lots more. They are also occasionally useful if
3033 you cache some data and want to flush it before blocking (for example,
3034 in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
3035 watcher).
3037 This is done by examining in each prepare call which file descriptors
3038 need to be watched by the other library, registering C<ev_io> watchers
3039 for them and starting an C<ev_timer> watcher for any timeouts (many
3040 libraries provide exactly this functionality). Then, in the check watcher,
3041 you check for any events that occurred (by checking the pending status
3042 of all watchers and stopping them) and call back into the library. The
3043 I/O and timer callbacks will never actually be called (but must be valid
3044 nevertheless, because you never know, you know?).
3046 As another example, the Perl Coro module uses these hooks to integrate
3047 coroutines into libev programs, by yielding to other active coroutines
3048 during each prepare and only letting the process block if no coroutines
3049 are ready to run (it's actually more complicated: it only runs coroutines
3050 with priority higher than or equal to the event loop and one coroutine
3051 of lower priority, but only once, using idle watchers to keep the event
3052 loop from blocking if lower-priority coroutines are active, thus mapping
3053 low-priority coroutines to idle/background tasks).
3055 When used for this purpose, it is recommended to give C<ev_check> watchers
3056 highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
3057 any other watchers after the poll (this doesn't matter for C<ev_prepare>
3058 watchers).
3060 Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
3061 activate ("feed") events into libev. While libev fully supports this, they
3062 might get executed before other C<ev_check> watchers did their job. As
3063 C<ev_check> watchers are often used to embed other (non-libev) event
3064 loops those other event loops might be in an unusable state until their
3065 C<ev_check> watcher ran (always remind yourself to coexist peacefully with
3066 others).
3068 =head3 Abusing an C<ev_check> watcher for its side-effect
3070 C<ev_check> (and less often also C<ev_prepare>) watchers can also be
3071 useful because they are called once per event loop iteration. For
3072 example, if you want to handle a large number of connections fairly, you
3073 normally only do a bit of work for each active connection, and if there
3074 is more work to do, you wait for the next event loop iteration, so other
3075 connections have a chance of making progress.
3077 Using an C<ev_check> watcher is almost enough: it will be called on the
3078 next event loop iteration. However, that isn't as soon as possible -
3079 without external events, your C<ev_check> watcher will not be invoked.
3081 This is where C<ev_idle> watchers come in handy - all you need is a
3082 single global idle watcher that is active as long as you have one active
3083 C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
3084 will not sleep, and the C<ev_check> watcher makes sure a callback gets
3085 invoked. Neither watcher alone can do that.
3087 =head3 Watcher-Specific Functions and Data Members
3089 =over 4
3091 =item ev_prepare_init (ev_prepare *, callback)
3093 =item ev_check_init (ev_check *, callback)
3095 Initialises and configures the prepare or check watcher - they have no
3096 parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
3097 macros, but using them is utterly, utterly, utterly and completely
3098 pointless.
3100 =back
3102 =head3 Examples
3104 There are a number of principal ways to embed other event loops or modules
3105 into libev. Here are some ideas on how to include libadns into libev
3106 (there is a Perl module named C<EV::ADNS> that does this, which you could
3107 use as a working example. Another Perl module named C<EV::Glib> embeds a
3108 Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
3109 Glib event loop).
3111 Method 1: Add IO watchers and a timeout watcher in a prepare handler,
3112 and in a check watcher, destroy them and call into libadns. What follows
3113 is pseudo-code only of course. This requires you to either use a low
3114 priority for the check watcher or use C<ev_clear_pending> explicitly, as
3115 the callbacks for the IO/timeout watchers might not have been called yet.
3117 static ev_io iow [nfd];
3118 static ev_timer tw;
3120 static void
3121 io_cb (struct ev_loop *loop, ev_io *w, int revents)
3122 {
3123 }
3125 // create io watchers for each fd and a timer before blocking
3126 static void
3127 adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
3128 {
3129 int timeout = 3600000;
3130 struct pollfd fds [nfd];
3131 // actual code will need to loop here and realloc etc.
3132 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
3134 /* the callback is illegal, but won't be called as we stop during check */
3135 ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
3136 ev_timer_start (loop, &tw);
3138 // create one ev_io per pollfd
3139 for (int i = 0; i < nfd; ++i)
3140 {
3141 ev_io_init (iow + i, io_cb, fds [i].fd,
3142 ((fds [i].events & POLLIN ? EV_READ : 0)
3143 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
3145 fds [i].revents = 0;
3146 ev_io_start (loop, iow + i);
3147 }
3148 }
3150 // stop all watchers after blocking
3151 static void
3152 adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
3153 {
3154 ev_timer_stop (loop, &tw);
3156 for (int i = 0; i < nfd; ++i)
3157 {
3158 // set the relevant poll flags
3159 // could also call adns_processreadable etc. here
3160 struct pollfd *fd = fds + i;
3161 int revents = ev_clear_pending (iow + i);
3162 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
3163 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
3165 // now stop the watcher
3166 ev_io_stop (loop, iow + i);
3167 }
3169 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
3170 }
3172 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
3173 in the prepare watcher and would dispose of the check watcher.
3175 Method 3: If the module to be embedded supports explicit event
3176 notification (libadns does), you can also make use of the actual watcher
3177 callbacks, and only destroy/create the watchers in the prepare watcher.
3179 static void
3180 timer_cb (EV_P_ ev_timer *w, int revents)
3181 {
3182 adns_state ads = (adns_state)w->data;
3183 update_now (EV_A);
3185 adns_processtimeouts (ads, &tv_now);
3186 }
3188 static void
3189 io_cb (EV_P_ ev_io *w, int revents)
3190 {
3191 adns_state ads = (adns_state)w->data;
3192 update_now (EV_A);
3194 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
3195 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
3196 }
3198 // do not ever call adns_afterpoll
3200 Method 4: Do not use a prepare or check watcher because the module you
3201 want to embed is not flexible enough to support it. Instead, you can
3202 override their poll function. The drawback with this solution is that the
3203 main loop is now no longer controllable by EV. The C<Glib::EV> module uses
3204 this approach, effectively embedding EV as a client into the horrible
3205 libglib event loop.
3207 static gint
3208 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
3209 {
3210 int got_events = 0;
3212 for (n = 0; n < nfds; ++n)
3213 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
3215 if (timeout >= 0)
3216 // create/start timer
3218 // poll
3219 ev_run (EV_A_ 0);
3221 // stop timer again
3222 if (timeout >= 0)
3223 ev_timer_stop (EV_A_ &to);
3225 // stop io watchers again - their callbacks should have set
3226 for (n = 0; n < nfds; ++n)
3227 ev_io_stop (EV_A_ iow [n]);
3229 return got_events;
3230 }
3233 =head2 C<ev_embed> - when one backend isn't enough...
3235 This is a rather advanced watcher type that lets you embed one event loop
3236 into another (currently only C<ev_io> events are supported in the embedded
3237 loop, other types of watchers might be handled in a delayed or incorrect
3238 fashion and must not be used).
3240 There are primarily two reasons you would want that: work around bugs and
3241 prioritise I/O.
3243 As an example for a bug workaround, the kqueue backend might only support
3244 sockets on some platform, so it is unusable as generic backend, but you
3245 still want to make use of it because you have many sockets and it scales
3246 so nicely. In this case, you would create a kqueue-based loop and embed
3247 it into your default loop (which might use e.g. poll). Overall operation
3248 will be a bit slower because first libev has to call C<poll> and then
3249 C<kevent>, but at least you can use both mechanisms for what they are
3250 best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
3252 As for prioritising I/O: under rare circumstances you have the case where
3253 some fds have to be watched and handled very quickly (with low latency),
3254 and even priorities and idle watchers might have too much overhead. In
3255 this case you would put all the high priority stuff in one loop and all
3256 the rest in a second one, and embed the second one in the first.
3258 As long as the watcher is active, the callback will be invoked every
3259 time there might be events pending in the embedded loop. The callback
3260 must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
3261 sweep and invoke their callbacks (the callback doesn't need to invoke the
3262 C<ev_embed_sweep> function directly, it could also start an idle watcher
3263 to give the embedded loop strictly lower priority for example).
3265 You can also set the callback to C<0>, in which case the embed watcher
3266 will automatically execute the embedded loop sweep whenever necessary.
3268 Fork detection will be handled transparently while the C<ev_embed> watcher
3269 is active, i.e., the embedded loop will automatically be forked when the
3270 embedding loop forks. In other cases, the user is responsible for calling
3271 C<ev_loop_fork> on the embedded loop.
3273 Unfortunately, not all backends are embeddable: only the ones returned by
3274 C<ev_embeddable_backends> are, which, unfortunately, does not include any
3275 portable one.
3277 So when you want to use this feature you will always have to be prepared
3278 that you cannot get an embeddable loop. The recommended way to get around
3279 this is to have a separate variables for your embeddable loop, try to
3280 create it, and if that fails, use the normal loop for everything.
3282 =head3 C<ev_embed> and fork
3284 While the C<ev_embed> watcher is running, forks in the embedding loop will
3285 automatically be applied to the embedded loop as well, so no special
3286 fork handling is required in that case. When the watcher is not running,
3287 however, it is still the task of the libev user to call C<ev_loop_fork ()>
3288 as applicable.
3290 =head3 Watcher-Specific Functions and Data Members
3292 =over 4
3294 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
3296 =item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
3298 Configures the watcher to embed the given loop, which must be
3299 embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
3300 invoked automatically, otherwise it is the responsibility of the callback
3301 to invoke it (it will continue to be called until the sweep has been done,
3302 if you do not want that, you need to temporarily stop the embed watcher).
3304 =item ev_embed_sweep (loop, ev_embed *)
3306 Make a single, non-blocking sweep over the embedded loop. This works
3307 similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
3308 appropriate way for embedded loops.
3310 =item struct ev_loop *other [read-only]
3312 The embedded event loop.
3314 =back
3316 =head3 Examples
3318 Example: Try to get an embeddable event loop and embed it into the default
3319 event loop. If that is not possible, use the default loop. The default
3320 loop is stored in C<loop_hi>, while the embeddable loop is stored in
3321 C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
3322 used).
3324 struct ev_loop *loop_hi = ev_default_init (0);
3325 struct ev_loop *loop_lo = 0;
3326 ev_embed embed;
3328 // see if there is a chance of getting one that works
3329 // (remember that a flags value of 0 means autodetection)
3330 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
3331 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
3332 : 0;
3334 // if we got one, then embed it, otherwise default to loop_hi
3335 if (loop_lo)
3336 {
3337 ev_embed_init (&embed, 0, loop_lo);
3338 ev_embed_start (loop_hi, &embed);
3339 }
3340 else
3341 loop_lo = loop_hi;
3343 Example: Check if kqueue is available but not recommended and create
3344 a kqueue backend for use with sockets (which usually work with any
3345 kqueue implementation). Store the kqueue/socket-only event loop in
3346 C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
3348 struct ev_loop *loop = ev_default_init (0);
3349 struct ev_loop *loop_socket = 0;
3350 ev_embed embed;
3352 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
3353 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
3354 {
3355 ev_embed_init (&embed, 0, loop_socket);
3356 ev_embed_start (loop, &embed);
3357 }
3359 if (!loop_socket)
3360 loop_socket = loop;
3362 // now use loop_socket for all sockets, and loop for everything else
3365 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
3367 Fork watchers are called when a C<fork ()> was detected (usually because
3368 whoever is a good citizen cared to tell libev about it by calling
3369 C<ev_loop_fork>). The invocation is done before the event loop blocks next
3370 and before C<ev_check> watchers are being called, and only in the child
3371 after the fork. If whoever good citizen calling C<ev_default_fork> cheats
3372 and calls it in the wrong process, the fork handlers will be invoked, too,
3373 of course.
3375 =head3 The special problem of life after fork - how is it possible?
3377 Most uses of C<fork ()> consist of forking, then some simple calls to set
3378 up/change the process environment, followed by a call to C<exec()>. This
3379 sequence should be handled by libev without any problems.
3381 This changes when the application actually wants to do event handling
3382 in the child, or both parent in child, in effect "continuing" after the
3383 fork.
3385 The default mode of operation (for libev, with application help to detect
3386 forks) is to duplicate all the state in the child, as would be expected
3387 when I<either> the parent I<or> the child process continues.
3389 When both processes want to continue using libev, then this is usually the
3390 wrong result. In that case, usually one process (typically the parent) is
3391 supposed to continue with all watchers in place as before, while the other
3392 process typically wants to start fresh, i.e. without any active watchers.
3394 The cleanest and most efficient way to achieve that with libev is to
3395 simply create a new event loop, which of course will be "empty", and
3396 use that for new watchers. This has the advantage of not touching more
3397 memory than necessary, and thus avoiding the copy-on-write, and the
3398 disadvantage of having to use multiple event loops (which do not support
3399 signal watchers).
3401 When this is not possible, or you want to use the default loop for
3402 other reasons, then in the process that wants to start "fresh", call
3403 C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
3404 Destroying the default loop will "orphan" (not stop) all registered
3405 watchers, so you have to be careful not to execute code that modifies
3406 those watchers. Note also that in that case, you have to re-register any
3407 signal watchers.
3409 =head3 Watcher-Specific Functions and Data Members
3411 =over 4
3413 =item ev_fork_init (ev_fork *, callback)
3415 Initialises and configures the fork watcher - it has no parameters of any
3416 kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
3417 really.
3419 =back
3422 =head2 C<ev_cleanup> - even the best things end
3424 Cleanup watchers are called just before the event loop is being destroyed
3425 by a call to C<ev_loop_destroy>.
3427 While there is no guarantee that the event loop gets destroyed, cleanup
3428 watchers provide a convenient method to install cleanup hooks for your
3429 program, worker threads and so on - you just to make sure to destroy the
3430 loop when you want them to be invoked.
3432 Cleanup watchers are invoked in the same way as any other watcher. Unlike
3433 all other watchers, they do not keep a reference to the event loop (which
3434 makes a lot of sense if you think about it). Like all other watchers, you
3435 can call libev functions in the callback, except C<ev_cleanup_start>.
3437 =head3 Watcher-Specific Functions and Data Members
3439 =over 4
3441 =item ev_cleanup_init (ev_cleanup *, callback)
3443 Initialises and configures the cleanup watcher - it has no parameters of
3444 any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
3445 pointless, I assure you.
3447 =back
3449 Example: Register an atexit handler to destroy the default loop, so any
3450 cleanup functions are called.
3452 static void
3453 program_exits (void)
3454 {
3455 ev_loop_destroy (EV_DEFAULT_UC);
3456 }
3458 ...
3459 atexit (program_exits);
3462 =head2 C<ev_async> - how to wake up an event loop
3464 In general, you cannot use an C<ev_loop> from multiple threads or other
3465 asynchronous sources such as signal handlers (as opposed to multiple event
3466 loops - those are of course safe to use in different threads).
3468 Sometimes, however, you need to wake up an event loop you do not control,
3469 for example because it belongs to another thread. This is what C<ev_async>
3470 watchers do: as long as the C<ev_async> watcher is active, you can signal
3471 it by calling C<ev_async_send>, which is thread- and signal safe.
3473 This functionality is very similar to C<ev_signal> watchers, as signals,
3474 too, are asynchronous in nature, and signals, too, will be compressed
3475 (i.e. the number of callback invocations may be less than the number of
3476 C<ev_async_send> calls). In fact, you could use signal watchers as a kind
3477 of "global async watchers" by using a watcher on an otherwise unused
3478 signal, and C<ev_feed_signal> to signal this watcher from another thread,
3479 even without knowing which loop owns the signal.
3481 =head3 Queueing
3483 C<ev_async> does not support queueing of data in any way. The reason
3484 is that the author does not know of a simple (or any) algorithm for a
3485 multiple-writer-single-reader queue that works in all cases and doesn't
3486 need elaborate support such as pthreads or unportable memory access
3487 semantics.
3489 That means that if you want to queue data, you have to provide your own
3490 queue. But at least I can tell you how to implement locking around your
3491 queue:
3493 =over 4
3495 =item queueing from a signal handler context
3497 To implement race-free queueing, you simply add to the queue in the signal
3498 handler but you block the signal handler in the watcher callback. Here is
3499 an example that does that for some fictitious SIGUSR1 handler:
3501 static ev_async mysig;
3503 static void
3504 sigusr1_handler (void)
3505 {
3506 sometype data;
3508 // no locking etc.
3509 queue_put (data);
3510 ev_async_send (EV_DEFAULT_ &mysig);
3511 }
3513 static void
3514 mysig_cb (EV_P_ ev_async *w, int revents)
3515 {
3516 sometype data;
3517 sigset_t block, prev;
3519 sigemptyset (&block);
3520 sigaddset (&block, SIGUSR1);
3521 sigprocmask (SIG_BLOCK, &block, &prev);
3523 while (queue_get (&data))
3524 process (data);
3526 if (sigismember (&prev, SIGUSR1)
3527 sigprocmask (SIG_UNBLOCK, &block, 0);
3528 }
3530 (Note: pthreads in theory requires you to use C<pthread_setmask>
3531 instead of C<sigprocmask> when you use threads, but libev doesn't do it
3532 either...).
3534 =item queueing from a thread context
3536 The strategy for threads is different, as you cannot (easily) block
3537 threads but you can easily preempt them, so to queue safely you need to
3538 employ a traditional mutex lock, such as in this pthread example:
3540 static ev_async mysig;
3541 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
3543 static void
3544 otherthread (void)
3545 {
3546 // only need to lock the actual queueing operation
3547 pthread_mutex_lock (&mymutex);
3548 queue_put (data);
3549 pthread_mutex_unlock (&mymutex);
3551 ev_async_send (EV_DEFAULT_ &mysig);
3552 }
3554 static void
3555 mysig_cb (EV_P_ ev_async *w, int revents)
3556 {
3557 pthread_mutex_lock (&mymutex);
3559 while (queue_get (&data))
3560 process (data);
3562 pthread_mutex_unlock (&mymutex);
3563 }
3565 =back
3568 =head3 Watcher-Specific Functions and Data Members
3570 =over 4
3572 =item ev_async_init (ev_async *, callback)
3574 Initialises and configures the async watcher - it has no parameters of any
3575 kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
3576 trust me.
3578 =item ev_async_send (loop, ev_async *)
3580 Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3581 an C<EV_ASYNC> event on the watcher into the event loop, and instantly
3582 returns.
3584 Unlike C<ev_feed_event>, this call is safe to do from other threads,
3585 signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3586 embedding section below on what exactly this means).
3588 Note that, as with other watchers in libev, multiple events might get
3589 compressed into a single callback invocation (another way to look at
3590 this is that C<ev_async> watchers are level-triggered: they are set on
3591 C<ev_async_send>, reset when the event loop detects that).
3593 This call incurs the overhead of at most one extra system call per event
3594 loop iteration, if the event loop is blocked, and no syscall at all if
3595 the event loop (or your program) is processing events. That means that
3596 repeated calls are basically free (there is no need to avoid calls for
3597 performance reasons) and that the overhead becomes smaller (typically
3598 zero) under load.
3600 =item bool = ev_async_pending (ev_async *)
3602 Returns a non-zero value when C<ev_async_send> has been called on the
3603 watcher but the event has not yet been processed (or even noted) by the
3604 event loop.
3606 C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
3607 the loop iterates next and checks for the watcher to have become active,
3608 it will reset the flag again. C<ev_async_pending> can be used to very
3609 quickly check whether invoking the loop might be a good idea.
3611 Not that this does I<not> check whether the watcher itself is pending,
3612 only whether it has been requested to make this watcher pending: there
3613 is a time window between the event loop checking and resetting the async
3614 notification, and the callback being invoked.
3616 =back
3621 There are some other functions of possible interest. Described. Here. Now.
3623 =over 4
3625 =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
3627 This function combines a simple timer and an I/O watcher, calls your
3628 callback on whichever event happens first and automatically stops both
3629 watchers. This is useful if you want to wait for a single event on an fd
3630 or timeout without having to allocate/configure/start/stop/free one or
3631 more watchers yourself.
3633 If C<fd> is less than 0, then no I/O watcher will be started and the
3634 C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
3635 the given C<fd> and C<events> set will be created and started.
3637 If C<timeout> is less than 0, then no timeout watcher will be
3638 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
3639 repeat = 0) will be started. C<0> is a valid timeout.
3641 The callback has the type C<void (*cb)(int revents, void *arg)> and is
3642 passed an C<revents> set like normal event callbacks (a combination of
3643 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
3644 value passed to C<ev_once>. Note that it is possible to receive I<both>
3645 a timeout and an io event at the same time - you probably should give io
3646 events precedence.
3648 Example: wait up to ten seconds for data to appear on STDIN_FILENO.
3650 static void stdin_ready (int revents, void *arg)
3651 {
3652 if (revents & EV_READ)
3653 /* stdin might have data for us, joy! */;
3654 else if (revents & EV_TIMER)
3655 /* doh, nothing entered */;
3656 }
3658 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3660 =item ev_feed_fd_event (loop, int fd, int revents)
3662 Feed an event on the given fd, as if a file descriptor backend detected
3663 the given events.
3665 =item ev_feed_signal_event (loop, int signum)
3667 Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3668 which is async-safe.
3670 =back
3675 This section explains some common idioms that are not immediately
3676 obvious. Note that examples are sprinkled over the whole manual, and this
3677 section only contains stuff that wouldn't fit anywhere else.
3681 Each watcher has, by default, a C<void *data> member that you can read
3682 or modify at any time: libev will completely ignore it. This can be used
3683 to associate arbitrary data with your watcher. If you need more data and
3684 don't want to allocate memory separately and store a pointer to it in that
3685 data member, you can also "subclass" the watcher type and provide your own
3686 data:
3688 struct my_io
3689 {
3690 ev_io io;
3691 int otherfd;
3692 void *somedata;
3693 struct whatever *mostinteresting;
3694 };
3696 ...
3697 struct my_io w;
3698 ev_io_init (&, my_cb, fd, EV_READ);
3700 And since your callback will be called with a pointer to the watcher, you
3701 can cast it back to your own type:
3703 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
3704 {
3705 struct my_io *w = (struct my_io *)w_;
3706 ...
3707 }
3709 More interesting and less C-conformant ways of casting your callback
3710 function type instead have been omitted.
3714 Another common scenario is to use some data structure with multiple
3715 embedded watchers, in effect creating your own watcher that combines
3716 multiple libev event sources into one "super-watcher":
3718 struct my_biggy
3719 {
3720 int some_data;
3721 ev_timer t1;
3722 ev_timer t2;
3723 }
3725 In this case getting the pointer to C<my_biggy> is a bit more
3726 complicated: Either you store the address of your C<my_biggy> struct in
3727 the C<data> member of the watcher (for woozies or C++ coders), or you need
3728 to use some pointer arithmetic using C<offsetof> inside your watchers (for
3729 real programmers):
3731 #include <stddef.h>
3733 static void
3734 t1_cb (EV_P_ ev_timer *w, int revents)
3735 {
3736 struct my_biggy big = (struct my_biggy *)
3737 (((char *)w) - offsetof (struct my_biggy, t1));
3738 }
3740 static void
3741 t2_cb (EV_P_ ev_timer *w, int revents)
3742 {
3743 struct my_biggy big = (struct my_biggy *)
3744 (((char *)w) - offsetof (struct my_biggy, t2));
3745 }
3749 Often you have structures like this in event-based programs:
3751 callback ()
3752 {
3753 free (request);
3754 }
3756 request = start_new_request (..., callback);
3758 The intent is to start some "lengthy" operation. The C<request> could be
3759 used to cancel the operation, or do other things with it.
3761 It's not uncommon to have code paths in C<start_new_request> that
3762 immediately invoke the callback, for example, to report errors. Or you add
3763 some caching layer that finds that it can skip the lengthy aspects of the
3764 operation and simply invoke the callback with the result.
3766 The problem here is that this will happen I<before> C<start_new_request>
3767 has returned, so C<request> is not set.
3769 Even if you pass the request by some safer means to the callback, you
3770 might want to do something to the request after starting it, such as
3771 canceling it, which probably isn't working so well when the callback has
3772 already been invoked.
3774 A common way around all these issues is to make sure that
3775 C<start_new_request> I<always> returns before the callback is invoked. If
3776 C<start_new_request> immediately knows the result, it can artificially
3777 delay invoking the callback by using a C<prepare> or C<idle> watcher for
3778 example, or more sneakily, by reusing an existing (stopped) watcher and
3779 pushing it into the pending queue:
3781 ev_set_cb (watcher, callback);
3782 ev_feed_event (EV_A_ watcher, 0);
3784 This way, C<start_new_request> can safely return before the callback is
3785 invoked, while not delaying callback invocation too much.
3789 Often (especially in GUI toolkits) there are places where you have
3790 I<modal> interaction, which is most easily implemented by recursively
3791 invoking C<ev_run>.
3793 This brings the problem of exiting - a callback might want to finish the
3794 main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
3795 a modal "Are you sure?" dialog is still waiting), or just the nested one
3796 and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
3797 other combination: In these cases, a simple C<ev_break> will not work.
3799 The solution is to maintain "break this loop" variable for each C<ev_run>
3800 invocation, and use a loop around C<ev_run> until the condition is
3801 triggered, using C<EVRUN_ONCE>:
3803 // main loop
3804 int exit_main_loop = 0;
3806 while (!exit_main_loop)
3807 ev_run (EV_DEFAULT_ EVRUN_ONCE);
3809 // in a modal watcher
3810 int exit_nested_loop = 0;
3812 while (!exit_nested_loop)
3813 ev_run (EV_A_ EVRUN_ONCE);
3815 To exit from any of these loops, just set the corresponding exit variable:
3817 // exit modal loop
3818 exit_nested_loop = 1;
3820 // exit main program, after modal loop is finished
3821 exit_main_loop = 1;
3823 // exit both
3824 exit_main_loop = exit_nested_loop = 1;
3828 Here is a fictitious example of how to run an event loop in a different
3829 thread from where callbacks are being invoked and watchers are
3830 created/added/removed.
3832 For a real-world example, see the C<EV::Loop::Async> perl module,
3833 which uses exactly this technique (which is suited for many high-level
3834 languages).
3836 The example uses a pthread mutex to protect the loop data, a condition
3837 variable to wait for callback invocations, an async watcher to notify the
3838 event loop thread and an unspecified mechanism to wake up the main thread.
3840 First, you need to associate some data with the event loop:
3842 typedef struct {
3843 mutex_t lock; /* global loop lock */
3844 ev_async async_w;
3845 thread_t tid;
3846 cond_t invoke_cv;
3847 } userdata;
3849 void prepare_loop (EV_P)
3850 {
3851 // for simplicity, we use a static userdata struct.
3852 static userdata u;
3854 ev_async_init (&u->async_w, async_cb);
3855 ev_async_start (EV_A_ &u->async_w);
3857 pthread_mutex_init (&u->lock, 0);
3858 pthread_cond_init (&u->invoke_cv, 0);
3860 // now associate this with the loop
3861 ev_set_userdata (EV_A_ u);
3862 ev_set_invoke_pending_cb (EV_A_ l_invoke);
3863 ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
3865 // then create the thread running ev_run
3866 pthread_create (&u->tid, 0, l_run, EV_A);
3867 }
3869 The callback for the C<ev_async> watcher does nothing: the watcher is used
3870 solely to wake up the event loop so it takes notice of any new watchers
3871 that might have been added:
3873 static void
3874 async_cb (EV_P_ ev_async *w, int revents)
3875 {
3876 // just used for the side effects
3877 }
3879 The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
3880 protecting the loop data, respectively.
3882 static void
3883 l_release (EV_P)
3884 {
3885 userdata *u = ev_userdata (EV_A);
3886 pthread_mutex_unlock (&u->lock);
3887 }
3889 static void
3890 l_acquire (EV_P)
3891 {
3892 userdata *u = ev_userdata (EV_A);
3893 pthread_mutex_lock (&u->lock);
3894 }
3896 The event loop thread first acquires the mutex, and then jumps straight
3897 into C<ev_run>:
3899 void *
3900 l_run (void *thr_arg)
3901 {
3902 struct ev_loop *loop = (struct ev_loop *)thr_arg;
3904 l_acquire (EV_A);
3905 pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
3906 ev_run (EV_A_ 0);
3907 l_release (EV_A);
3909 return 0;
3910 }
3912 Instead of invoking all pending watchers, the C<l_invoke> callback will
3913 signal the main thread via some unspecified mechanism (signals? pipe
3914 writes? C<Async::Interrupt>?) and then waits until all pending watchers
3915 have been called (in a while loop because a) spurious wakeups are possible
3916 and b) skipping inter-thread-communication when there are no pending
3917 watchers is very beneficial):
3919 static void
3920 l_invoke (EV_P)
3921 {
3922 userdata *u = ev_userdata (EV_A);
3924 while (ev_pending_count (EV_A))
3925 {
3926 wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
3927 pthread_cond_wait (&u->invoke_cv, &u->lock);
3928 }
3929 }
3931 Now, whenever the main thread gets told to invoke pending watchers, it
3932 will grab the lock, call C<ev_invoke_pending> and then signal the loop
3933 thread to continue:
3935 static void
3936 real_invoke_pending (EV_P)
3937 {
3938 userdata *u = ev_userdata (EV_A);
3940 pthread_mutex_lock (&u->lock);
3941 ev_invoke_pending (EV_A);
3942 pthread_cond_signal (&u->invoke_cv);
3943 pthread_mutex_unlock (&u->lock);
3944 }
3946 Whenever you want to start/stop a watcher or do other modifications to an
3947 event loop, you will now have to lock:
3949 ev_timer timeout_watcher;
3950 userdata *u = ev_userdata (EV_A);
3952 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
3954 pthread_mutex_lock (&u->lock);
3955 ev_timer_start (EV_A_ &timeout_watcher);
3956 ev_async_send (EV_A_ &u->async_w);
3957 pthread_mutex_unlock (&u->lock);
3959 Note that sending the C<ev_async> watcher is required because otherwise
3960 an event loop currently blocking in the kernel will have no knowledge
3961 about the newly added timer. By waking up the loop it will pick up any new
3962 watchers in the next event loop iteration.
3966 While the overhead of a callback that e.g. schedules a thread is small, it
3967 is still an overhead. If you embed libev, and your main usage is with some
3968 kind of threads or coroutines, you might want to customise libev so that
3969 doesn't need callbacks anymore.
3971 Imagine you have coroutines that you can switch to using a function
3972 C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
3973 and that due to some magic, the currently active coroutine is stored in a
3974 global called C<current_coro>. Then you can build your own "wait for libev
3975 event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
3976 the differing C<;> conventions):
3978 #define EV_CB_DECLARE(type) struct my_coro *cb;
3979 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
3981 That means instead of having a C callback function, you store the
3982 coroutine to switch to in each watcher, and instead of having libev call
3983 your callback, you instead have it switch to that coroutine.
3985 A coroutine might now wait for an event with a function called
3986 C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
3987 matter when, or whether the watcher is active or not when this function is
3988 called):
3990 void
3991 wait_for_event (ev_watcher *w)
3992 {
3993 ev_set_cb (w, current_coro);
3994 switch_to (libev_coro);
3995 }
3997 That basically suspends the coroutine inside C<wait_for_event> and
3998 continues the libev coroutine, which, when appropriate, switches back to
3999 this or any other coroutine.
4001 You can do similar tricks if you have, say, threads with an event queue -
4002 instead of storing a coroutine, you store the queue object and instead of
4003 switching to a coroutine, you push the watcher onto the queue and notify
4004 any waiters.
4006 To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
4007 files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
4009 // my_ev.h
4010 #define EV_CB_DECLARE(type) struct my_coro *cb;
4011 #define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
4012 #include "../libev/ev.h"
4014 // my_ev.c
4015 #define EV_H "my_ev.h"
4016 #include "../libev/ev.c"
4018 And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
4019 F<my_ev.c> into your project. When properly specifying include paths, you
4020 can even use F<ev.h> as header file name directly.
4025 Libev offers a compatibility emulation layer for libevent. It cannot
4026 emulate the internals of libevent, so here are some usage hints:
4028 =over 4
4030 =item * Only the libevent-1.4.1-beta API is being emulated.
4032 This was the newest libevent version available when libev was implemented,
4033 and is still mostly unchanged in 2010.
4035 =item * Use it by including <event.h>, as usual.
4037 =item * The following members are fully supported: ev_base, ev_callback,
4038 ev_arg, ev_fd, ev_res, ev_events.
4040 =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
4041 maintained by libev, it does not work exactly the same way as in libevent (consider
4042 it a private API).
4044 =item * Priorities are not currently supported. Initialising priorities
4045 will fail and all watchers will have the same priority, even though there
4046 is an ev_pri field.
4048 =item * In libevent, the last base created gets the signals, in libev, the
4049 base that registered the signal gets the signals.
4051 =item * Other members are not supported.
4053 =item * The libev emulation is I<not> ABI compatible to libevent, you need
4054 to use the libev header file and library.
4056 =back
4058 =head1 C++ SUPPORT
4060 =head2 C API
4062 The normal C API should work fine when used from C++: both ev.h and the
4063 libev sources can be compiled as C++. Therefore, code that uses the C API
4064 will work fine.
4066 Proper exception specifications might have to be added to callbacks passed
4067 to libev: exceptions may be thrown only from watcher callbacks, all other
4068 callbacks (allocator, syserr, loop acquire/release and periodic reschedule
4069 callbacks) must not throw exceptions, and might need a C<noexcept>
4070 specification. If you have code that needs to be compiled as both C and
4071 C++ you can use the C<EV_NOEXCEPT> macro for this:
4073 static void
4074 fatal_error (const char *msg) EV_NOEXCEPT
4075 {
4076 perror (msg);
4077 abort ();
4078 }
4080 ...
4081 ev_set_syserr_cb (fatal_error);
4083 The only API functions that can currently throw exceptions are C<ev_run>,
4084 C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
4085 because it runs cleanup watchers).
4087 Throwing exceptions in watcher callbacks is only supported if libev itself
4088 is compiled with a C++ compiler or your C and C++ environments allow
4089 throwing exceptions through C libraries (most do).
4091 =head2 C++ API
4093 Libev comes with some simplistic wrapper classes for C++ that mainly allow
4094 you to use some convenience methods to start/stop watchers and also change
4095 the callback model to a model using method callbacks on objects.
4097 To use it,
4099 #include <ev++.h>
4101 This automatically includes F<ev.h> and puts all of its definitions (many
4102 of them macros) into the global namespace. All C++ specific things are
4103 put into the C<ev> namespace. It should support all the same embedding
4104 options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
4106 Care has been taken to keep the overhead low. The only data member the C++
4107 classes add (compared to plain C-style watchers) is the event loop pointer
4108 that the watcher is associated with (or no additional members at all if
4109 you disable C<EV_MULTIPLICITY> when embedding libev).
4111 Currently, functions, static and non-static member functions and classes
4112 with C<operator ()> can be used as callbacks. Other types should be easy
4113 to add as long as they only need one additional pointer for context. If
4114 you need support for other types of functors please contact the author
4115 (preferably after implementing it).
4117 For all this to work, your C++ compiler either has to use the same calling
4118 conventions as your C compiler (for static member functions), or you have
4119 to embed libev and compile libev itself as C++.
4121 Here is a list of things available in the C<ev> namespace:
4123 =over 4
4125 =item C<ev::READ>, C<ev::WRITE> etc.
4127 These are just enum values with the same values as the C<EV_READ> etc.
4128 macros from F<ev.h>.
4130 =item C<ev::tstamp>, C<ev::now>
4132 Aliases to the same types/functions as with the C<ev_> prefix.
4134 =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
4136 For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
4137 the same name in the C<ev> namespace, with the exception of C<ev_signal>
4138 which is called C<ev::sig> to avoid clashes with the C<signal> macro
4139 defined by many implementations.
4141 All of those classes have these methods:
4143 =over 4
4145 =item ev::TYPE::TYPE ()
4147 =item ev::TYPE::TYPE (loop)
4149 =item ev::TYPE::~TYPE
4151 The constructor (optionally) takes an event loop to associate the watcher
4152 with. If it is omitted, it will use C<EV_DEFAULT>.
4154 The constructor calls C<ev_init> for you, which means you have to call the
4155 C<set> method before starting it.
4157 It will not set a callback, however: You have to call the templated C<set>
4158 method to set a callback before you can start the watcher.
4160 (The reason why you have to use a method is a limitation in C++ which does
4161 not allow explicit template arguments for constructors).
4163 The destructor automatically stops the watcher if it is active.
4165 =item w->set<class, &class::method> (object *)
4167 This method sets the callback method to call. The method has to have a
4168 signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
4169 first argument and the C<revents> as second. The object must be given as
4170 parameter and is stored in the C<data> member of the watcher.
4172 This method synthesizes efficient thunking code to call your method from
4173 the C callback that libev requires. If your compiler can inline your
4174 callback (i.e. it is visible to it at the place of the C<set> call and
4175 your compiler is good :), then the method will be fully inlined into the
4176 thunking function, making it as fast as a direct C callback.
4178 Example: simple class declaration and watcher initialisation
4180 struct myclass
4181 {
4182 void io_cb (ev::io &w, int revents) { }
4183 }
4185 myclass obj;
4186 ev::io iow;
4187 iow.set <myclass, &myclass::io_cb> (&obj);
4189 =item w->set (object *)
4191 This is a variation of a method callback - leaving out the method to call
4192 will default the method to C<operator ()>, which makes it possible to use
4193 functor objects without having to manually specify the C<operator ()> all
4194 the time. Incidentally, you can then also leave out the template argument
4195 list.
4197 The C<operator ()> method prototype must be C<void operator ()(watcher &w,
4198 int revents)>.
4200 See the method-C<set> above for more details.
4202 Example: use a functor object as callback.
4204 struct myfunctor
4205 {
4206 void operator() (ev::io &w, int revents)
4207 {
4208 ...
4209 }
4210 }
4212 myfunctor f;
4214 ev::io w;
4215 w.set (&f);
4217 =item w->set<function> (void *data = 0)
4219 Also sets a callback, but uses a static method or plain function as
4220 callback. The optional C<data> argument will be stored in the watcher's
4221 C<data> member and is free for you to use.
4223 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
4225 See the method-C<set> above for more details.
4227 Example: Use a plain function as callback.
4229 static void io_cb (ev::io &w, int revents) { }
4230 iow.set <io_cb> ();
4232 =item w->set (loop)
4234 Associates a different C<struct ev_loop> with this watcher. You can only
4235 do this when the watcher is inactive (and not pending either).
4237 =item w->set ([arguments])
4239 Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
4240 with the same arguments. Either this method or a suitable start method
4241 must be called at least once. Unlike the C counterpart, an active watcher
4242 gets automatically stopped and restarted when reconfiguring it with this
4243 method.
4245 For C<ev::embed> watchers this method is called C<set_embed>, to avoid
4246 clashing with the C<set (loop)> method.
4248 =item w->start ()
4250 Starts the watcher. Note that there is no C<loop> argument, as the
4251 constructor already stores the event loop.
4253 =item w->start ([arguments])
4255 Instead of calling C<set> and C<start> methods separately, it is often
4256 convenient to wrap them in one call. Uses the same type of arguments as
4257 the configure C<set> method of the watcher.
4259 =item w->stop ()
4261 Stops the watcher if it is active. Again, no C<loop> argument.
4263 =item w->again () (C<ev::timer>, C<ev::periodic> only)
4265 For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
4266 C<ev_TYPE_again> function.
4268 =item w->sweep () (C<ev::embed> only)
4270 Invokes C<ev_embed_sweep>.
4272 =item w->update () (C<ev::stat> only)
4274 Invokes C<ev_stat_stat>.
4276 =back
4278 =back
4280 Example: Define a class with two I/O and idle watchers, start the I/O
4281 watchers in the constructor.
4283 class myclass
4284 {
4285 ev::io io ; void io_cb (ev::io &w, int revents);
4286 ev::io io2 ; void io2_cb (ev::io &w, int revents);
4287 ev::idle idle; void idle_cb (ev::idle &w, int revents);
4289 myclass (int fd)
4290 {
4291 io .set <myclass, &myclass::io_cb > (this);
4292 io2 .set <myclass, &myclass::io2_cb > (this);
4293 idle.set <myclass, &myclass::idle_cb> (this);
4295 io.set (fd, ev::WRITE); // configure the watcher
4296 io.start (); // start it whenever convenient
4298 io2.start (fd, ev::READ); // set + start in one call
4299 }
4300 };
4305 Libev does not offer other language bindings itself, but bindings for a
4306 number of languages exist in the form of third-party packages. If you know
4307 any interesting language binding in addition to the ones listed here, drop
4308 me a note.
4310 =over 4
4312 =item Perl
4314 The EV module implements the full libev API and is actually used to test
4315 libev. EV is developed together with libev. Apart from the EV core module,
4316 there are additional modules that implement libev-compatible interfaces
4317 to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
4318 C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
4319 and C<EV::Glib>).
4321 It can be found and installed via CPAN, its homepage is at
4322 L<>.
4324 =item Python
4326 Python bindings can be found at L<>. It
4327 seems to be quite complete and well-documented.
4329 =item Ruby
4331 Tony Arcieri has written a ruby extension that offers access to a subset
4332 of the libev API and adds file handle abstractions, asynchronous DNS and
4333 more on top of it. It can be found via gem servers. Its homepage is at
4334 L<>.
4336 Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
4337 makes rev work even on mingw.
4339 =item Haskell
4341 A haskell binding to libev is available at
4342 L<>.
4344 =item D
4346 Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
4347 be found at L<>.
4349 =item Ocaml
4351 Erkki Seppala has written Ocaml bindings for libev, to be found at
4352 L<>.
4354 =item Lua
4356 Brian Maher has written a partial interface to libev for lua (at the
4357 time of this writing, only C<ev_io> and C<ev_timer>), to be found at
4358 L<>.
4360 =item Javascript
4362 Node.js (L<>) uses libev as the underlying event library.
4364 =item Others
4366 There are others, and I stopped counting.
4368 =back
4371 =head1 MACRO MAGIC
4373 Libev can be compiled with a variety of options, the most fundamental
4374 of which is C<EV_MULTIPLICITY>. This option determines whether (most)
4375 functions and callbacks have an initial C<struct ev_loop *> argument.
4377 To make it easier to write programs that cope with either variant, the
4378 following macros are defined:
4380 =over 4
4382 =item C<EV_A>, C<EV_A_>
4384 This provides the loop I<argument> for functions, if one is required ("ev
4385 loop argument"). The C<EV_A> form is used when this is the sole argument,
4386 C<EV_A_> is used when other arguments are following. Example:
4388 ev_unref (EV_A);
4389 ev_timer_add (EV_A_ watcher);
4390 ev_run (EV_A_ 0);
4392 It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
4393 which is often provided by the following macro.
4395 =item C<EV_P>, C<EV_P_>
4397 This provides the loop I<parameter> for functions, if one is required ("ev
4398 loop parameter"). The C<EV_P> form is used when this is the sole parameter,
4399 C<EV_P_> is used when other parameters are following. Example:
4401 // this is how ev_unref is being declared
4402 static void ev_unref (EV_P);
4404 // this is how you can declare your typical callback
4405 static void cb (EV_P_ ev_timer *w, int revents)
4407 It declares a parameter C<loop> of type C<struct ev_loop *>, quite
4408 suitable for use with C<EV_A>.
4410 =item C<EV_DEFAULT>, C<EV_DEFAULT_>
4412 Similar to the other two macros, this gives you the value of the default
4413 loop, if multiple loops are supported ("ev loop default"). The default loop
4414 will be initialised if it isn't already initialised.
4416 For non-multiplicity builds, these macros do nothing, so you always have
4417 to initialise the loop somewhere.
4421 Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
4422 default loop has been initialised (C<UC> == unchecked). Their behaviour
4423 is undefined when the default loop has not been initialised by a previous
4424 execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
4426 It is often prudent to use C<EV_DEFAULT> when initialising the first
4427 watcher in a function but use C<EV_DEFAULT_UC> afterwards.
4429 =back
4431 Example: Declare and initialise a check watcher, utilising the above
4432 macros so it will work regardless of whether multiple loops are supported
4433 or not.
4435 static void
4436 check_cb (EV_P_ ev_timer *w, int revents)
4437 {
4438 ev_check_stop (EV_A_ w);
4439 }
4441 ev_check check;
4442 ev_check_init (&check, check_cb);
4443 ev_check_start (EV_DEFAULT_ &check);
4444 ev_run (EV_DEFAULT_ 0);
4446 =head1 EMBEDDING
4448 Libev can (and often is) directly embedded into host
4449 applications. Examples of applications that embed it include the Deliantra
4450 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
4451 and rxvt-unicode.
4453 The goal is to enable you to just copy the necessary files into your
4454 source directory without having to change even a single line in them, so
4455 you can easily upgrade by simply copying (or having a checked-out copy of
4456 libev somewhere in your source tree).
4458 =head2 FILESETS
4460 Depending on what features you need you need to include one or more sets of files
4461 in your application.
4463 =head3 CORE EVENT LOOP
4465 To include only the libev core (all the C<ev_*> functions), with manual
4466 configuration (no autoconf):
4468 #define EV_STANDALONE 1
4469 #include "ev.c"
4471 This will automatically include F<ev.h>, too, and should be done in a
4472 single C source file only to provide the function implementations. To use
4473 it, do the same for F<ev.h> in all files wishing to use this API (best
4474 done by writing a wrapper around F<ev.h> that you can include instead and
4475 where you can put other configuration options):
4477 #define EV_STANDALONE 1
4478 #include "ev.h"
4480 Both header files and implementation files can be compiled with a C++
4481 compiler (at least, that's a stated goal, and breakage will be treated
4482 as a bug).
4484 You need the following files in your source tree, or in a directory
4485 in your include path (e.g. in libev/ when using -Ilibev):
4487 ev.h
4488 ev.c
4489 ev_vars.h
4490 ev_wrap.h
4492 ev_win32.c required on win32 platforms only
4494 ev_select.c only when select backend is enabled
4495 ev_poll.c only when poll backend is enabled
4496 ev_epoll.c only when the epoll backend is enabled
4497 ev_linuxaio.c only when the linux aio backend is enabled
4498 ev_iouring.c only when the linux io_uring backend is enabled
4499 ev_kqueue.c only when the kqueue backend is enabled
4500 ev_port.c only when the solaris port backend is enabled
4502 F<ev.c> includes the backend files directly when enabled, so you only need
4503 to compile this single file.
4507 To include the libevent compatibility API, also include:
4509 #include "event.c"
4511 in the file including F<ev.c>, and:
4513 #include "event.h"
4515 in the files that want to use the libevent API. This also includes F<ev.h>.
4517 You need the following additional files for this:
4519 event.h
4520 event.c
4524 Instead of using C<EV_STANDALONE=1> and providing your configuration in
4525 whatever way you want, you can also C<m4_include([libev.m4])> in your
4526 F<> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
4527 include F<config.h> and configure itself accordingly.
4529 For this of course you need the m4 file:
4531 libev.m4
4535 Libev can be configured via a variety of preprocessor symbols you have to
4536 define before including (or compiling) any of its files. The default in
4537 the absence of autoconf is documented for every option.
4539 Symbols marked with "(h)" do not change the ABI, and can have different
4540 values when compiling libev vs. including F<ev.h>, so it is permissible
4541 to redefine them before including F<ev.h> without breaking compatibility
4542 to a compiled library. All other symbols change the ABI, which means all
4543 users of libev and the libev code itself must be compiled with compatible
4544 settings.
4546 =over 4
4548 =item EV_COMPAT3 (h)
4550 Backwards compatibility is a major concern for libev. This is why this
4551 release of libev comes with wrappers for the functions and symbols that
4552 have been renamed between libev version 3 and 4.
4554 You can disable these wrappers (to test compatibility with future
4555 versions) by defining C<EV_COMPAT3> to C<0> when compiling your
4556 sources. This has the additional advantage that you can drop the C<struct>
4557 from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
4558 typedef in that case.
4560 In some future version, the default for C<EV_COMPAT3> will become C<0>,
4561 and in some even more future version the compatibility code will be
4562 removed completely.
4564 =item EV_STANDALONE (h)
4566 Must always be C<1> if you do not use autoconf configuration, which
4567 keeps libev from including F<config.h>, and it also defines dummy
4568 implementations for some libevent functions (such as logging, which is not
4569 supported). It will also not define any of the structs usually found in
4570 F<event.h> that are not directly supported by the libev core alone.
4572 In standalone mode, libev will still try to automatically deduce the
4573 configuration, but has to be more conservative.
4575 =item EV_USE_FLOOR
4577 If defined to be C<1>, libev will use the C<floor ()> function for its
4578 periodic reschedule calculations, otherwise libev will fall back on a
4579 portable (slower) implementation. If you enable this, you usually have to
4580 link against libm or something equivalent. Enabling this when the C<floor>
4581 function is not available will fail, so the safe default is to not enable
4582 this.
4586 If defined to be C<1>, libev will try to detect the availability of the
4587 monotonic clock option at both compile time and runtime. Otherwise no
4588 use of the monotonic clock option will be attempted. If you enable this,
4589 you usually have to link against librt or something similar. Enabling it
4590 when the functionality isn't available is safe, though, although you have
4591 to make sure you link against any libraries where the C<clock_gettime>
4592 function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
4594 =item EV_USE_REALTIME
4596 If defined to be C<1>, libev will try to detect the availability of the
4597 real-time clock option at compile time (and assume its availability
4598 at runtime if successful). Otherwise no use of the real-time clock
4599 option will be attempted. This effectively replaces C<gettimeofday>
4600 by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
4601 correctness. See the note about libraries in the description of
4602 C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
4607 If defined to be C<1>, libev will try to use a direct syscall instead
4608 of calling the system-provided C<clock_gettime> function. This option
4609 exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
4610 unconditionally pulls in C<libpthread>, slowing down single-threaded
4611 programs needlessly. Using a direct syscall is slightly slower (in
4612 theory), because no optimised vdso implementation can be used, but avoids
4613 the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
4614 higher, as it simplifies linking (no need for C<-lrt>).
4618 If defined to be C<1>, libev will assume that C<nanosleep ()> is available
4619 and will use it for delays. Otherwise it will use C<select ()>.
4621 =item EV_USE_EVENTFD
4623 If defined to be C<1>, then libev will assume that C<eventfd ()> is
4624 available and will probe for kernel support at runtime. This will improve
4625 C<ev_signal> and C<ev_async> performance and reduce resource consumption.
4626 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
4627 2.7 or newer, otherwise disabled.
4629 =item EV_USE_SIGNALFD
4631 If defined to be C<1>, then libev will assume that C<signalfd ()> is
4632 available and will probe for kernel support at runtime. This enables
4633 the use of EVFLAG_SIGNALFD for faster and simpler signal handling. If
4634 undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
4635 2.7 or newer, otherwise disabled.
4637 =item EV_USE_TIMERFD
4639 If defined to be C<1>, then libev will assume that C<timerfd ()> is
4640 available and will probe for kernel support at runtime. This allows
4641 libev to detect time jumps accurately. If undefined, it will be enabled
4642 if the headers indicate GNU/Linux + Glibc 2.8 or newer and define
4643 C<TFD_TIMER_CANCEL_ON_SET>, otherwise disabled.
4645 =item EV_USE_EVENTFD
4647 If defined to be C<1>, then libev will assume that C<eventfd ()> is
4648 available and will probe for kernel support at runtime. This will improve
4649 C<ev_signal> and C<ev_async> performance and reduce resource consumption.
4650 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
4651 2.7 or newer, otherwise disabled.
4653 =item EV_USE_SELECT
4655 If undefined or defined to be C<1>, libev will compile in support for the
4656 C<select>(2) backend. No attempt at auto-detection will be done: if no
4657 other method takes over, select will be it. Otherwise the select backend
4658 will not be compiled in.
4662 If defined to C<1>, then the select backend will use the system C<fd_set>
4663 structure. This is useful if libev doesn't compile due to a missing
4664 C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
4665 on exotic systems. This usually limits the range of file descriptors to
4666 some low limit such as 1024 or might have other limitations (winsocket
4667 only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
4668 configures the maximum size of the C<fd_set>.
4672 When defined to C<1>, the select backend will assume that
4673 select/socket/connect etc. don't understand file descriptors but
4674 wants osf handles on win32 (this is the case when the select to
4675 be used is the winsock select). This means that it will call
4676 C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
4677 it is assumed that all these functions actually work on fds, even
4678 on win32. Should not be defined on non-win32 platforms.
4680 =item EV_FD_TO_WIN32_HANDLE(fd)
4682 If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
4683 file descriptors to socket handles. When not defining this symbol (the
4684 default), then libev will call C<_get_osfhandle>, which is usually
4685 correct. In some cases, programs use their own file descriptor management,
4686 in which case they can provide this function to map fds to socket handles.
4688 =item EV_WIN32_HANDLE_TO_FD(handle)
4690 If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
4691 using the standard C<_open_osfhandle> function. For programs implementing
4692 their own fd to handle mapping, overwriting this function makes it easier
4693 to do so. This can be done by defining this macro to an appropriate value.
4695 =item EV_WIN32_CLOSE_FD(fd)
4697 If programs implement their own fd to handle mapping on win32, then this
4698 macro can be used to override the C<close> function, useful to unregister
4699 file descriptors again. Note that the replacement function has to close
4700 the underlying OS handle.
4704 If defined to be C<1>, libev will use C<WSASocket> to create its internal
4705 communication socket, which works better in some environments. Otherwise,
4706 the normal C<socket> function will be used, which works better in other
4707 environments.
4709 =item EV_USE_POLL
4711 If defined to be C<1>, libev will compile in support for the C<poll>(2)
4712 backend. Otherwise it will be enabled on non-win32 platforms. It
4713 takes precedence over select.
4715 =item EV_USE_EPOLL
4717 If defined to be C<1>, libev will compile in support for the Linux
4718 C<epoll>(7) backend. Its availability will be detected at runtime,
4719 otherwise another method will be used as fallback. This is the preferred
4720 backend for GNU/Linux systems. If undefined, it will be enabled if the
4721 headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4723 =item EV_USE_LINUXAIO
4725 If defined to be C<1>, libev will compile in support for the Linux aio
4726 backend (C<EV_USE_EPOLL> must also be enabled). If undefined, it will be
4727 enabled on linux, otherwise disabled.
4729 =item EV_USE_IOURING
4731 If defined to be C<1>, libev will compile in support for the Linux
4732 io_uring backend (C<EV_USE_EPOLL> must also be enabled). Due to it's
4733 current limitations it has to be requested explicitly. If undefined, it
4734 will be enabled on linux, otherwise disabled.
4736 =item EV_USE_KQUEUE
4738 If defined to be C<1>, libev will compile in support for the BSD style
4739 C<kqueue>(2) backend. Its actual availability will be detected at runtime,
4740 otherwise another method will be used as fallback. This is the preferred
4741 backend for BSD and BSD-like systems, although on most BSDs kqueue only
4742 supports some types of fds correctly (the only platform we found that
4743 supports ptys for example was NetBSD), so kqueue might be compiled in, but
4744 not be used unless explicitly requested. The best way to use it is to find
4745 out whether kqueue supports your type of fd properly and use an embedded
4746 kqueue loop.
4748 =item EV_USE_PORT
4750 If defined to be C<1>, libev will compile in support for the Solaris
4751 10 port style backend. Its availability will be detected at runtime,
4752 otherwise another method will be used as fallback. This is the preferred
4753 backend for Solaris 10 systems.
4755 =item EV_USE_DEVPOLL
4757 Reserved for future expansion, works like the USE symbols above.
4759 =item EV_USE_INOTIFY
4761 If defined to be C<1>, libev will compile in support for the Linux inotify
4762 interface to speed up C<ev_stat> watchers. Its actual availability will
4763 be detected at runtime. If undefined, it will be enabled if the headers
4764 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4766 =item EV_NO_SMP
4768 If defined to be C<1>, libev will assume that memory is always coherent
4769 between threads, that is, threads can be used, but threads never run on
4770 different cpus (or different cpu cores). This reduces dependencies
4771 and makes libev faster.
4773 =item EV_NO_THREADS
4775 If defined to be C<1>, libev will assume that it will never be called from
4776 different threads (that includes signal handlers), which is a stronger
4777 assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
4778 libev faster.
4780 =item EV_ATOMIC_T
4782 Libev requires an integer type (suitable for storing C<0> or C<1>) whose
4783 access is atomic with respect to other threads or signal contexts. No
4784 such type is easily found in the C language, so you can provide your own
4785 type that you know is safe for your purposes. It is used both for signal
4786 handler "locking" as well as for signal and thread safety in C<ev_async>
4787 watchers.
4789 In the absence of this define, libev will use C<sig_atomic_t volatile>
4790 (from F<signal.h>), which is usually good enough on most platforms.
4792 =item EV_H (h)
4794 The name of the F<ev.h> header file used to include it. The default if
4795 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
4796 used to virtually rename the F<ev.h> header file in case of conflicts.
4798 =item EV_CONFIG_H (h)
4800 If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
4801 F<ev.c>'s idea of where to find the F<config.h> file, similarly to
4802 C<EV_H>, above.
4804 =item EV_EVENT_H (h)
4806 Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
4807 of how the F<event.h> header can be found, the default is C<"event.h">.
4809 =item EV_PROTOTYPES (h)
4811 If defined to be C<0>, then F<ev.h> will not define any function
4812 prototypes, but still define all the structs and other symbols. This is
4813 occasionally useful if you want to provide your own wrapper functions
4814 around libev functions.
4818 If undefined or defined to C<1>, then all event-loop-specific functions
4819 will have the C<struct ev_loop *> as first argument, and you can create
4820 additional independent event loops. Otherwise there will be no support
4821 for multiple event loops and there is no first event loop pointer
4822 argument. Instead, all functions act on the single default loop.
4824 Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
4825 default loop when multiplicity is switched off - you always have to
4826 initialise the loop manually in this case.
4828 =item EV_MINPRI
4830 =item EV_MAXPRI
4832 The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
4833 C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
4834 provide for more priorities by overriding those symbols (usually defined
4835 to be C<-2> and C<2>, respectively).
4837 When doing priority-based operations, libev usually has to linearly search
4838 all the priorities, so having many of them (hundreds) uses a lot of space
4839 and time, so using the defaults of five priorities (-2 .. +2) is usually
4840 fine.
4842 If your embedding application does not need any priorities, defining these
4843 both to C<0> will save some memory and CPU.
4849 If undefined or defined to be C<1> (and the platform supports it), then
4850 the respective watcher type is supported. If defined to be C<0>, then it
4851 is not. Disabling watcher types mainly saves code size.
4853 =item EV_FEATURES
4855 If you need to shave off some kilobytes of code at the expense of some
4856 speed (but with the full API), you can define this symbol to request
4857 certain subsets of functionality. The default is to enable all features
4858 that can be enabled on the platform.
4860 A typical way to use this symbol is to define it to C<0> (or to a bitset
4861 with some broad features you want) and then selectively re-enable
4862 additional parts you want, for example if you want everything minimal,
4863 but multiple event loop support, async and child watchers and the poll
4864 backend, use this:
4866 #define EV_FEATURES 0
4867 #define EV_MULTIPLICITY 1
4868 #define EV_USE_POLL 1
4869 #define EV_CHILD_ENABLE 1
4870 #define EV_ASYNC_ENABLE 1
4872 The actual value is a bitset, it can be a combination of the following
4873 values (by default, all of these are enabled):
4875 =over 4
4877 =item C<1> - faster/larger code
4879 Use larger code to speed up some operations.
4881 Currently this is used to override some inlining decisions (enlarging the
4882 code size by roughly 30% on amd64).
4884 When optimising for size, use of compiler flags such as C<-Os> with
4885 gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
4886 assertions.
4888 The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
4889 (e.g. gcc with C<-Os>).
4891 =item C<2> - faster/larger data structures
4893 Replaces the small 2-heap for timer management by a faster 4-heap, larger
4894 hash table sizes and so on. This will usually further increase code size
4895 and can additionally have an effect on the size of data structures at
4896 runtime.
4898 The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
4899 (e.g. gcc with C<-Os>).
4901 =item C<4> - full API configuration
4903 This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
4904 enables multiplicity (C<EV_MULTIPLICITY>=1).
4906 =item C<8> - full API
4908 This enables a lot of the "lesser used" API functions. See C<ev.h> for
4909 details on which parts of the API are still available without this
4910 feature, and do not complain if this subset changes over time.
4912 =item C<16> - enable all optional watcher types
4914 Enables all optional watcher types. If you want to selectively enable
4915 only some watcher types other than I/O and timers (e.g. prepare,
4916 embed, async, child...) you can enable them manually by defining
4917 C<EV_watchertype_ENABLE> to C<1> instead.
4919 =item C<32> - enable all backends
4921 This enables all backends - without this feature, you need to enable at
4922 least one backend manually (C<EV_USE_SELECT> is a good choice).
4924 =item C<64> - enable OS-specific "helper" APIs
4926 Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
4927 default.
4929 =back
4931 Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
4932 reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
4933 code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
4934 watchers, timers and monotonic clock support.
4936 With an intelligent-enough linker (gcc+binutils are intelligent enough
4937 when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4938 your program might be left out as well - a binary starting a timer and an
4939 I/O watcher then might come out at only 5Kb.
4941 =item EV_API_STATIC
4943 If this symbol is defined (by default it is not), then all identifiers
4944 will have static linkage. This means that libev will not export any
4945 identifiers, and you cannot link against libev anymore. This can be useful
4946 when you embed libev, only want to use libev functions in a single file,
4947 and do not want its identifiers to be visible.
4949 To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
4950 wants to use libev.
4952 This option only works when libev is compiled with a C compiler, as C++
4953 doesn't support the required declaration syntax.
4955 =item EV_AVOID_STDIO
4957 If this is set to C<1> at compiletime, then libev will avoid using stdio
4958 functions (printf, scanf, perror etc.). This will increase the code size
4959 somewhat, but if your program doesn't otherwise depend on stdio and your
4960 libc allows it, this avoids