ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.pod
Revision: 1.235
Committed: Thu Apr 16 07:50:39 2009 UTC (15 years, 1 month ago) by root
Branch: MAIN
Changes since 1.234: +1 -1 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 NAME
2
3 libev - a high performance full-featured event loop written in C
4
5 =head1 SYNOPSIS
6
7 #include <ev.h>
8
9 =head2 EXAMPLE PROGRAM
10
11 // a single header file is required
12 #include <ev.h>
13
14 #include <stdio.h> // for puts
15
16 // every watcher type has its own typedef'd struct
17 // with the name ev_TYPE
18 ev_io stdin_watcher;
19 ev_timer timeout_watcher;
20
21 // all watcher callbacks have a similar signature
22 // this callback is called when data is readable on stdin
23 static void
24 stdin_cb (EV_P_ ev_io *w, int revents)
25 {
26 puts ("stdin ready");
27 // for one-shot events, one must manually stop the watcher
28 // with its corresponding stop function.
29 ev_io_stop (EV_A_ w);
30
31 // this causes all nested ev_loop's to stop iterating
32 ev_unloop (EV_A_ EVUNLOOP_ALL);
33 }
34
35 // another callback, this time for a time-out
36 static void
37 timeout_cb (EV_P_ ev_timer *w, int revents)
38 {
39 puts ("timeout");
40 // this causes the innermost ev_loop to stop iterating
41 ev_unloop (EV_A_ EVUNLOOP_ONE);
42 }
43
44 int
45 main (void)
46 {
47 // use the default event loop unless you have special needs
48 struct ev_loop *loop = ev_default_loop (0);
49
50 // initialise an io watcher, then start it
51 // this one will watch for stdin to become readable
52 ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
53 ev_io_start (loop, &stdin_watcher);
54
55 // initialise a timer watcher, then start it
56 // simple non-repeating 5.5 second timeout
57 ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
58 ev_timer_start (loop, &timeout_watcher);
59
60 // now wait for events to arrive
61 ev_loop (loop, 0);
62
63 // unloop was called, so exit
64 return 0;
65 }
66
67 =head1 DESCRIPTION
68
69 The newest version of this document is also available as an html-formatted
70 web page you might find easier to navigate when reading it for the first
71 time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
72
73 Libev is an event loop: you register interest in certain events (such as a
74 file descriptor being readable or a timeout occurring), and it will manage
75 these event sources and provide your program with events.
76
77 To do this, it must take more or less complete control over your process
78 (or thread) by executing the I<event loop> handler, and will then
79 communicate events via a callback mechanism.
80
81 You register interest in certain events by registering so-called I<event
82 watchers>, which are relatively small C structures you initialise with the
83 details of the event, and then hand it over to libev by I<starting> the
84 watcher.
85
86 =head2 FEATURES
87
88 Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
89 BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
90 for file descriptor events (C<ev_io>), the Linux C<inotify> interface
91 (for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
92 with customised rescheduling (C<ev_periodic>), synchronous signals
93 (C<ev_signal>), process status change events (C<ev_child>), and event
94 watchers dealing with the event loop mechanism itself (C<ev_idle>,
95 C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
96 file watchers (C<ev_stat>) and even limited support for fork events
97 (C<ev_fork>).
98
99 It also is quite fast (see this
100 L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
101 for example).
102
103 =head2 CONVENTIONS
104
105 Libev is very configurable. In this manual the default (and most common)
106 configuration will be described, which supports multiple event loops. For
107 more info about various configuration options please have a look at
108 B<EMBED> section in this manual. If libev was configured without support
109 for multiple event loops, then all functions taking an initial argument of
110 name C<loop> (which is always of type C<ev_loop *>) will not have
111 this argument.
112
113 =head2 TIME REPRESENTATION
114
115 Libev represents time as a single floating point number, representing the
116 (fractional) number of seconds since the (POSIX) epoch (somewhere near
117 the beginning of 1970, details are complicated, don't ask). This type is
118 called C<ev_tstamp>, which is what you should use too. It usually aliases
119 to the C<double> type in C, and when you need to do any calculations on
120 it, you should treat it as some floating point value. Unlike the name
121 component C<stamp> might indicate, it is also used for time differences
122 throughout libev.
123
124 =head1 ERROR HANDLING
125
126 Libev knows three classes of errors: operating system errors, usage errors
127 and internal errors (bugs).
128
129 When libev catches an operating system error it cannot handle (for example
130 a system call indicating a condition libev cannot fix), it calls the callback
131 set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
132 abort. The default is to print a diagnostic message and to call C<abort
133 ()>.
134
135 When libev detects a usage error such as a negative timer interval, then
136 it will print a diagnostic message and abort (via the C<assert> mechanism,
137 so C<NDEBUG> will disable this checking): these are programming errors in
138 the libev caller and need to be fixed there.
139
140 Libev also has a few internal error-checking C<assert>ions, and also has
141 extensive consistency checking code. These do not trigger under normal
142 circumstances, as they indicate either a bug in libev or worse.
143
144
145 =head1 GLOBAL FUNCTIONS
146
147 These functions can be called anytime, even before initialising the
148 library in any way.
149
150 =over 4
151
152 =item ev_tstamp ev_time ()
153
154 Returns the current time as libev would use it. Please note that the
155 C<ev_now> function is usually faster and also often returns the timestamp
156 you actually want to know.
157
158 =item ev_sleep (ev_tstamp interval)
159
160 Sleep for the given interval: The current thread will be blocked until
161 either it is interrupted or the given time interval has passed. Basically
162 this is a sub-second-resolution C<sleep ()>.
163
164 =item int ev_version_major ()
165
166 =item int ev_version_minor ()
167
168 You can find out the major and minor ABI version numbers of the library
169 you linked against by calling the functions C<ev_version_major> and
170 C<ev_version_minor>. If you want, you can compare against the global
171 symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
172 version of the library your program was compiled against.
173
174 These version numbers refer to the ABI version of the library, not the
175 release version.
176
177 Usually, it's a good idea to terminate if the major versions mismatch,
178 as this indicates an incompatible change. Minor versions are usually
179 compatible to older versions, so a larger minor version alone is usually
180 not a problem.
181
182 Example: Make sure we haven't accidentally been linked against the wrong
183 version.
184
185 assert (("libev version mismatch",
186 ev_version_major () == EV_VERSION_MAJOR
187 && ev_version_minor () >= EV_VERSION_MINOR));
188
189 =item unsigned int ev_supported_backends ()
190
191 Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
192 value) compiled into this binary of libev (independent of their
193 availability on the system you are running on). See C<ev_default_loop> for
194 a description of the set values.
195
196 Example: make sure we have the epoll method, because yeah this is cool and
197 a must have and can we have a torrent of it please!!!11
198
199 assert (("sorry, no epoll, no sex",
200 ev_supported_backends () & EVBACKEND_EPOLL));
201
202 =item unsigned int ev_recommended_backends ()
203
204 Return the set of all backends compiled into this binary of libev and also
205 recommended for this platform. This set is often smaller than the one
206 returned by C<ev_supported_backends>, as for example kqueue is broken on
207 most BSDs and will not be auto-detected unless you explicitly request it
208 (assuming you know what you are doing). This is the set of backends that
209 libev will probe for if you specify no backends explicitly.
210
211 =item unsigned int ev_embeddable_backends ()
212
213 Returns the set of backends that are embeddable in other event loops. This
214 is the theoretical, all-platform, value. To find which backends
215 might be supported on the current system, you would need to look at
216 C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
217 recommended ones.
218
219 See the description of C<ev_embed> watchers for more info.
220
221 =item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
222
223 Sets the allocation function to use (the prototype is similar - the
224 semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
225 used to allocate and free memory (no surprises here). If it returns zero
226 when memory needs to be allocated (C<size != 0>), the library might abort
227 or take some potentially destructive action.
228
229 Since some systems (at least OpenBSD and Darwin) fail to implement
230 correct C<realloc> semantics, libev will use a wrapper around the system
231 C<realloc> and C<free> functions by default.
232
233 You could override this function in high-availability programs to, say,
234 free some memory if it cannot allocate memory, to use a special allocator,
235 or even to sleep a while and retry until some memory is available.
236
237 Example: Replace the libev allocator with one that waits a bit and then
238 retries (example requires a standards-compliant C<realloc>).
239
240 static void *
241 persistent_realloc (void *ptr, size_t size)
242 {
243 for (;;)
244 {
245 void *newptr = realloc (ptr, size);
246
247 if (newptr)
248 return newptr;
249
250 sleep (60);
251 }
252 }
253
254 ...
255 ev_set_allocator (persistent_realloc);
256
257 =item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
258
259 Set the callback function to call on a retryable system call error (such
260 as failed select, poll, epoll_wait). The message is a printable string
261 indicating the system call or subsystem causing the problem. If this
262 callback is set, then libev will expect it to remedy the situation, no
263 matter what, when it returns. That is, libev will generally retry the
264 requested operation, or, if the condition doesn't go away, do bad stuff
265 (such as abort).
266
267 Example: This is basically the same thing that libev does internally, too.
268
269 static void
270 fatal_error (const char *msg)
271 {
272 perror (msg);
273 abort ();
274 }
275
276 ...
277 ev_set_syserr_cb (fatal_error);
278
279 =back
280
281 =head1 FUNCTIONS CONTROLLING THE EVENT LOOP
282
283 An event loop is described by a C<struct ev_loop *> (the C<struct>
284 is I<not> optional in this case, as there is also an C<ev_loop>
285 I<function>).
286
287 The library knows two types of such loops, the I<default> loop, which
288 supports signals and child events, and dynamically created loops which do
289 not.
290
291 =over 4
292
293 =item struct ev_loop *ev_default_loop (unsigned int flags)
294
295 This will initialise the default event loop if it hasn't been initialised
296 yet and return it. If the default loop could not be initialised, returns
297 false. If it already was initialised it simply returns it (and ignores the
298 flags. If that is troubling you, check C<ev_backend ()> afterwards).
299
300 If you don't know what event loop to use, use the one returned from this
301 function.
302
303 Note that this function is I<not> thread-safe, so if you want to use it
304 from multiple threads, you have to lock (note also that this is unlikely,
305 as loops cannot be shared easily between threads anyway).
306
307 The default loop is the only loop that can handle C<ev_signal> and
308 C<ev_child> watchers, and to do this, it always registers a handler
309 for C<SIGCHLD>. If this is a problem for your application you can either
310 create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
311 can simply overwrite the C<SIGCHLD> signal handler I<after> calling
312 C<ev_default_init>.
313
314 The flags argument can be used to specify special behaviour or specific
315 backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
316
317 The following flags are supported:
318
319 =over 4
320
321 =item C<EVFLAG_AUTO>
322
323 The default flags value. Use this if you have no clue (it's the right
324 thing, believe me).
325
326 =item C<EVFLAG_NOENV>
327
328 If this flag bit is or'ed into the flag value (or the program runs setuid
329 or setgid) then libev will I<not> look at the environment variable
330 C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
331 override the flags completely if it is found in the environment. This is
332 useful to try out specific backends to test their performance, or to work
333 around bugs.
334
335 =item C<EVFLAG_FORKCHECK>
336
337 Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
338 a fork, you can also make libev check for a fork in each iteration by
339 enabling this flag.
340
341 This works by calling C<getpid ()> on every iteration of the loop,
342 and thus this might slow down your event loop if you do a lot of loop
343 iterations and little real work, but is usually not noticeable (on my
344 GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
345 without a system call and thus I<very> fast, but my GNU/Linux system also has
346 C<pthread_atfork> which is even faster).
347
348 The big advantage of this flag is that you can forget about fork (and
349 forget about forgetting to tell libev about forking) when you use this
350 flag.
351
352 This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
353 environment variable.
354
355 =item C<EVBACKEND_SELECT> (value 1, portable select backend)
356
357 This is your standard select(2) backend. Not I<completely> standard, as
358 libev tries to roll its own fd_set with no limits on the number of fds,
359 but if that fails, expect a fairly low limit on the number of fds when
360 using this backend. It doesn't scale too well (O(highest_fd)), but its
361 usually the fastest backend for a low number of (low-numbered :) fds.
362
363 To get good performance out of this backend you need a high amount of
364 parallelism (most of the file descriptors should be busy). If you are
365 writing a server, you should C<accept ()> in a loop to accept as many
366 connections as possible during one iteration. You might also want to have
367 a look at C<ev_set_io_collect_interval ()> to increase the amount of
368 readiness notifications you get per iteration.
369
370 This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
371 C<writefds> set (and to work around Microsoft Windows bugs, also onto the
372 C<exceptfds> set on that platform).
373
374 =item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
375
376 And this is your standard poll(2) backend. It's more complicated
377 than select, but handles sparse fds better and has no artificial
378 limit on the number of fds you can use (except it will slow down
379 considerably with a lot of inactive fds). It scales similarly to select,
380 i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
381 performance tips.
382
383 This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
384 C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
385
386 =item C<EVBACKEND_EPOLL> (value 4, Linux)
387
388 For few fds, this backend is a bit little slower than poll and select,
389 but it scales phenomenally better. While poll and select usually scale
390 like O(total_fds) where n is the total number of fds (or the highest fd),
391 epoll scales either O(1) or O(active_fds).
392
393 The epoll mechanism deserves honorable mention as the most misdesigned
394 of the more advanced event mechanisms: mere annoyances include silently
395 dropping file descriptors, requiring a system call per change per file
396 descriptor (and unnecessary guessing of parameters), problems with dup and
397 so on. The biggest issue is fork races, however - if a program forks then
398 I<both> parent and child process have to recreate the epoll set, which can
399 take considerable time (one syscall per file descriptor) and is of course
400 hard to detect.
401
402 Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
403 of course I<doesn't>, and epoll just loves to report events for totally
404 I<different> file descriptors (even already closed ones, so one cannot
405 even remove them from the set) than registered in the set (especially
406 on SMP systems). Libev tries to counter these spurious notifications by
407 employing an additional generation counter and comparing that against the
408 events to filter out spurious ones, recreating the set when required.
409
410 While stopping, setting and starting an I/O watcher in the same iteration
411 will result in some caching, there is still a system call per such
412 incident (because the same I<file descriptor> could point to a different
413 I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
414 file descriptors might not work very well if you register events for both
415 file descriptors.
416
417 Best performance from this backend is achieved by not unregistering all
418 watchers for a file descriptor until it has been closed, if possible,
419 i.e. keep at least one watcher active per fd at all times. Stopping and
420 starting a watcher (without re-setting it) also usually doesn't cause
421 extra overhead. A fork can both result in spurious notifications as well
422 as in libev having to destroy and recreate the epoll object, which can
423 take considerable time and thus should be avoided.
424
425 All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
426 faster than epoll for maybe up to a hundred file descriptors, depending on
427 the usage. So sad.
428
429 While nominally embeddable in other event loops, this feature is broken in
430 all kernel versions tested so far.
431
432 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
433 C<EVBACKEND_POLL>.
434
435 =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
436
437 Kqueue deserves special mention, as at the time of this writing, it
438 was broken on all BSDs except NetBSD (usually it doesn't work reliably
439 with anything but sockets and pipes, except on Darwin, where of course
440 it's completely useless). Unlike epoll, however, whose brokenness
441 is by design, these kqueue bugs can (and eventually will) be fixed
442 without API changes to existing programs. For this reason it's not being
443 "auto-detected" unless you explicitly specify it in the flags (i.e. using
444 C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
445 system like NetBSD.
446
447 You still can embed kqueue into a normal poll or select backend and use it
448 only for sockets (after having made sure that sockets work with kqueue on
449 the target platform). See C<ev_embed> watchers for more info.
450
451 It scales in the same way as the epoll backend, but the interface to the
452 kernel is more efficient (which says nothing about its actual speed, of
453 course). While stopping, setting and starting an I/O watcher does never
454 cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
455 two event changes per incident. Support for C<fork ()> is very bad (but
456 sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
457 cases
458
459 This backend usually performs well under most conditions.
460
461 While nominally embeddable in other event loops, this doesn't work
462 everywhere, so you might need to test for this. And since it is broken
463 almost everywhere, you should only use it when you have a lot of sockets
464 (for which it usually works), by embedding it into another event loop
465 (e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
466 also broken on OS X)) and, did I mention it, using it only for sockets.
467
468 This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
469 C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
470 C<NOTE_EOF>.
471
472 =item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
473
474 This is not implemented yet (and might never be, unless you send me an
475 implementation). According to reports, C</dev/poll> only supports sockets
476 and is not embeddable, which would limit the usefulness of this backend
477 immensely.
478
479 =item C<EVBACKEND_PORT> (value 32, Solaris 10)
480
481 This uses the Solaris 10 event port mechanism. As with everything on Solaris,
482 it's really slow, but it still scales very well (O(active_fds)).
483
484 Please note that Solaris event ports can deliver a lot of spurious
485 notifications, so you need to use non-blocking I/O or other means to avoid
486 blocking when no data (or space) is available.
487
488 While this backend scales well, it requires one system call per active
489 file descriptor per loop iteration. For small and medium numbers of file
490 descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
491 might perform better.
492
493 On the positive side, with the exception of the spurious readiness
494 notifications, this backend actually performed fully to specification
495 in all tests and is fully embeddable, which is a rare feat among the
496 OS-specific backends (I vastly prefer correctness over speed hacks).
497
498 This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
499 C<EVBACKEND_POLL>.
500
501 =item C<EVBACKEND_ALL>
502
503 Try all backends (even potentially broken ones that wouldn't be tried
504 with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
505 C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
506
507 It is definitely not recommended to use this flag.
508
509 =back
510
511 If one or more of these are or'ed into the flags value, then only these
512 backends will be tried (in the reverse order as listed here). If none are
513 specified, all backends in C<ev_recommended_backends ()> will be tried.
514
515 Example: This is the most typical usage.
516
517 if (!ev_default_loop (0))
518 fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
519
520 Example: Restrict libev to the select and poll backends, and do not allow
521 environment settings to be taken into account:
522
523 ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
524
525 Example: Use whatever libev has to offer, but make sure that kqueue is
526 used if available (warning, breaks stuff, best use only with your own
527 private event loop and only if you know the OS supports your types of
528 fds):
529
530 ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
531
532 =item struct ev_loop *ev_loop_new (unsigned int flags)
533
534 Similar to C<ev_default_loop>, but always creates a new event loop that is
535 always distinct from the default loop. Unlike the default loop, it cannot
536 handle signal and child watchers, and attempts to do so will be greeted by
537 undefined behaviour (or a failed assertion if assertions are enabled).
538
539 Note that this function I<is> thread-safe, and the recommended way to use
540 libev with threads is indeed to create one loop per thread, and using the
541 default loop in the "main" or "initial" thread.
542
543 Example: Try to create a event loop that uses epoll and nothing else.
544
545 struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
546 if (!epoller)
547 fatal ("no epoll found here, maybe it hides under your chair");
548
549 =item ev_default_destroy ()
550
551 Destroys the default loop again (frees all memory and kernel state
552 etc.). None of the active event watchers will be stopped in the normal
553 sense, so e.g. C<ev_is_active> might still return true. It is your
554 responsibility to either stop all watchers cleanly yourself I<before>
555 calling this function, or cope with the fact afterwards (which is usually
556 the easiest thing, you can just ignore the watchers and/or C<free ()> them
557 for example).
558
559 Note that certain global state, such as signal state (and installed signal
560 handlers), will not be freed by this function, and related watchers (such
561 as signal and child watchers) would need to be stopped manually.
562
563 In general it is not advisable to call this function except in the
564 rare occasion where you really need to free e.g. the signal handling
565 pipe fds. If you need dynamically allocated loops it is better to use
566 C<ev_loop_new> and C<ev_loop_destroy>).
567
568 =item ev_loop_destroy (loop)
569
570 Like C<ev_default_destroy>, but destroys an event loop created by an
571 earlier call to C<ev_loop_new>.
572
573 =item ev_default_fork ()
574
575 This function sets a flag that causes subsequent C<ev_loop> iterations
576 to reinitialise the kernel state for backends that have one. Despite the
577 name, you can call it anytime, but it makes most sense after forking, in
578 the child process (or both child and parent, but that again makes little
579 sense). You I<must> call it in the child before using any of the libev
580 functions, and it will only take effect at the next C<ev_loop> iteration.
581
582 On the other hand, you only need to call this function in the child
583 process if and only if you want to use the event library in the child. If
584 you just fork+exec, you don't have to call it at all.
585
586 The function itself is quite fast and it's usually not a problem to call
587 it just in case after a fork. To make this easy, the function will fit in
588 quite nicely into a call to C<pthread_atfork>:
589
590 pthread_atfork (0, 0, ev_default_fork);
591
592 =item ev_loop_fork (loop)
593
594 Like C<ev_default_fork>, but acts on an event loop created by
595 C<ev_loop_new>. Yes, you have to call this on every allocated event loop
596 after fork that you want to re-use in the child, and how you do this is
597 entirely your own problem.
598
599 =item int ev_is_default_loop (loop)
600
601 Returns true when the given loop is, in fact, the default loop, and false
602 otherwise.
603
604 =item unsigned int ev_loop_count (loop)
605
606 Returns the count of loop iterations for the loop, which is identical to
607 the number of times libev did poll for new events. It starts at C<0> and
608 happily wraps around with enough iterations.
609
610 This value can sometimes be useful as a generation counter of sorts (it
611 "ticks" the number of loop iterations), as it roughly corresponds with
612 C<ev_prepare> and C<ev_check> calls.
613
614 =item unsigned int ev_backend (loop)
615
616 Returns one of the C<EVBACKEND_*> flags indicating the event backend in
617 use.
618
619 =item ev_tstamp ev_now (loop)
620
621 Returns the current "event loop time", which is the time the event loop
622 received events and started processing them. This timestamp does not
623 change as long as callbacks are being processed, and this is also the base
624 time used for relative timers. You can treat it as the timestamp of the
625 event occurring (or more correctly, libev finding out about it).
626
627 =item ev_now_update (loop)
628
629 Establishes the current time by querying the kernel, updating the time
630 returned by C<ev_now ()> in the progress. This is a costly operation and
631 is usually done automatically within C<ev_loop ()>.
632
633 This function is rarely useful, but when some event callback runs for a
634 very long time without entering the event loop, updating libev's idea of
635 the current time is a good idea.
636
637 See also "The special problem of time updates" in the C<ev_timer> section.
638
639 =item ev_suspend (loop)
640
641 =item ev_resume (loop)
642
643 These two functions suspend and resume a loop, for use when the loop is
644 not used for a while and timeouts should not be processed.
645
646 A typical use case would be an interactive program such as a game: When
647 the user presses C<^Z> to suspend the game and resumes it an hour later it
648 would be best to handle timeouts as if no time had actually passed while
649 the program was suspended. This can be achieved by calling C<ev_suspend>
650 in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
651 C<ev_resume> directly afterwards to resume timer processing.
652
653 Effectively, all C<ev_timer> watchers will be delayed by the time spend
654 between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
655 will be rescheduled (that is, they will lose any events that would have
656 occured while suspended).
657
658 After calling C<ev_suspend> you B<must not> call I<any> function on the
659 given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
660 without a previous call to C<ev_suspend>.
661
662 Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
663 event loop time (see C<ev_now_update>).
664
665 =item ev_loop (loop, int flags)
666
667 Finally, this is it, the event handler. This function usually is called
668 after you initialised all your watchers and you want to start handling
669 events.
670
671 If the flags argument is specified as C<0>, it will not return until
672 either no event watchers are active anymore or C<ev_unloop> was called.
673
674 Please note that an explicit C<ev_unloop> is usually better than
675 relying on all watchers to be stopped when deciding when a program has
676 finished (especially in interactive programs), but having a program
677 that automatically loops as long as it has to and no longer by virtue
678 of relying on its watchers stopping correctly, that is truly a thing of
679 beauty.
680
681 A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
682 those events and any already outstanding ones, but will not block your
683 process in case there are no events and will return after one iteration of
684 the loop.
685
686 A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
687 necessary) and will handle those and any already outstanding ones. It
688 will block your process until at least one new event arrives (which could
689 be an event internal to libev itself, so there is no guarantee that a
690 user-registered callback will be called), and will return after one
691 iteration of the loop.
692
693 This is useful if you are waiting for some external event in conjunction
694 with something not expressible using other libev watchers (i.e. "roll your
695 own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
696 usually a better approach for this kind of thing.
697
698 Here are the gory details of what C<ev_loop> does:
699
700 - Before the first iteration, call any pending watchers.
701 * If EVFLAG_FORKCHECK was used, check for a fork.
702 - If a fork was detected (by any means), queue and call all fork watchers.
703 - Queue and call all prepare watchers.
704 - If we have been forked, detach and recreate the kernel state
705 as to not disturb the other process.
706 - Update the kernel state with all outstanding changes.
707 - Update the "event loop time" (ev_now ()).
708 - Calculate for how long to sleep or block, if at all
709 (active idle watchers, EVLOOP_NONBLOCK or not having
710 any active watchers at all will result in not sleeping).
711 - Sleep if the I/O and timer collect interval say so.
712 - Block the process, waiting for any events.
713 - Queue all outstanding I/O (fd) events.
714 - Update the "event loop time" (ev_now ()), and do time jump adjustments.
715 - Queue all expired timers.
716 - Queue all expired periodics.
717 - Unless any events are pending now, queue all idle watchers.
718 - Queue all check watchers.
719 - Call all queued watchers in reverse order (i.e. check watchers first).
720 Signals and child watchers are implemented as I/O watchers, and will
721 be handled here by queueing them when their watcher gets executed.
722 - If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
723 were used, or there are no active watchers, return, otherwise
724 continue with step *.
725
726 Example: Queue some jobs and then loop until no events are outstanding
727 anymore.
728
729 ... queue jobs here, make sure they register event watchers as long
730 ... as they still have work to do (even an idle watcher will do..)
731 ev_loop (my_loop, 0);
732 ... jobs done or somebody called unloop. yeah!
733
734 =item ev_unloop (loop, how)
735
736 Can be used to make a call to C<ev_loop> return early (but only after it
737 has processed all outstanding events). The C<how> argument must be either
738 C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
739 C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
740
741 This "unloop state" will be cleared when entering C<ev_loop> again.
742
743 It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
744
745 =item ev_ref (loop)
746
747 =item ev_unref (loop)
748
749 Ref/unref can be used to add or remove a reference count on the event
750 loop: Every watcher keeps one reference, and as long as the reference
751 count is nonzero, C<ev_loop> will not return on its own.
752
753 If you have a watcher you never unregister that should not keep C<ev_loop>
754 from returning, call ev_unref() after starting, and ev_ref() before
755 stopping it.
756
757 As an example, libev itself uses this for its internal signal pipe: It
758 is not visible to the libev user and should not keep C<ev_loop> from
759 exiting if no event watchers registered by it are active. It is also an
760 excellent way to do this for generic recurring timers or from within
761 third-party libraries. Just remember to I<unref after start> and I<ref
762 before stop> (but only if the watcher wasn't active before, or was active
763 before, respectively. Note also that libev might stop watchers itself
764 (e.g. non-repeating timers) in which case you have to C<ev_ref>
765 in the callback).
766
767 Example: Create a signal watcher, but keep it from keeping C<ev_loop>
768 running when nothing else is active.
769
770 ev_signal exitsig;
771 ev_signal_init (&exitsig, sig_cb, SIGINT);
772 ev_signal_start (loop, &exitsig);
773 evf_unref (loop);
774
775 Example: For some weird reason, unregister the above signal handler again.
776
777 ev_ref (loop);
778 ev_signal_stop (loop, &exitsig);
779
780 =item ev_set_io_collect_interval (loop, ev_tstamp interval)
781
782 =item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
783
784 These advanced functions influence the time that libev will spend waiting
785 for events. Both time intervals are by default C<0>, meaning that libev
786 will try to invoke timer/periodic callbacks and I/O callbacks with minimum
787 latency.
788
789 Setting these to a higher value (the C<interval> I<must> be >= C<0>)
790 allows libev to delay invocation of I/O and timer/periodic callbacks
791 to increase efficiency of loop iterations (or to increase power-saving
792 opportunities).
793
794 The idea is that sometimes your program runs just fast enough to handle
795 one (or very few) event(s) per loop iteration. While this makes the
796 program responsive, it also wastes a lot of CPU time to poll for new
797 events, especially with backends like C<select ()> which have a high
798 overhead for the actual polling but can deliver many events at once.
799
800 By setting a higher I<io collect interval> you allow libev to spend more
801 time collecting I/O events, so you can handle more events per iteration,
802 at the cost of increasing latency. Timeouts (both C<ev_periodic> and
803 C<ev_timer>) will be not affected. Setting this to a non-null value will
804 introduce an additional C<ev_sleep ()> call into most loop iterations.
805
806 Likewise, by setting a higher I<timeout collect interval> you allow libev
807 to spend more time collecting timeouts, at the expense of increased
808 latency/jitter/inexactness (the watcher callback will be called
809 later). C<ev_io> watchers will not be affected. Setting this to a non-null
810 value will not introduce any overhead in libev.
811
812 Many (busy) programs can usually benefit by setting the I/O collect
813 interval to a value near C<0.1> or so, which is often enough for
814 interactive servers (of course not for games), likewise for timeouts. It
815 usually doesn't make much sense to set it to a lower value than C<0.01>,
816 as this approaches the timing granularity of most systems.
817
818 Setting the I<timeout collect interval> can improve the opportunity for
819 saving power, as the program will "bundle" timer callback invocations that
820 are "near" in time together, by delaying some, thus reducing the number of
821 times the process sleeps and wakes up again. Another useful technique to
822 reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
823 they fire on, say, one-second boundaries only.
824
825 =item ev_loop_verify (loop)
826
827 This function only does something when C<EV_VERIFY> support has been
828 compiled in, which is the default for non-minimal builds. It tries to go
829 through all internal structures and checks them for validity. If anything
830 is found to be inconsistent, it will print an error message to standard
831 error and call C<abort ()>.
832
833 This can be used to catch bugs inside libev itself: under normal
834 circumstances, this function will never abort as of course libev keeps its
835 data structures consistent.
836
837 =back
838
839
840 =head1 ANATOMY OF A WATCHER
841
842 In the following description, uppercase C<TYPE> in names stands for the
843 watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
844 watchers and C<ev_io_start> for I/O watchers.
845
846 A watcher is a structure that you create and register to record your
847 interest in some event. For instance, if you want to wait for STDIN to
848 become readable, you would create an C<ev_io> watcher for that:
849
850 static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
851 {
852 ev_io_stop (w);
853 ev_unloop (loop, EVUNLOOP_ALL);
854 }
855
856 struct ev_loop *loop = ev_default_loop (0);
857
858 ev_io stdin_watcher;
859
860 ev_init (&stdin_watcher, my_cb);
861 ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
862 ev_io_start (loop, &stdin_watcher);
863
864 ev_loop (loop, 0);
865
866 As you can see, you are responsible for allocating the memory for your
867 watcher structures (and it is I<usually> a bad idea to do this on the
868 stack).
869
870 Each watcher has an associated watcher structure (called C<struct ev_TYPE>
871 or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
872
873 Each watcher structure must be initialised by a call to C<ev_init
874 (watcher *, callback)>, which expects a callback to be provided. This
875 callback gets invoked each time the event occurs (or, in the case of I/O
876 watchers, each time the event loop detects that the file descriptor given
877 is readable and/or writable).
878
879 Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
880 macro to configure it, with arguments specific to the watcher type. There
881 is also a macro to combine initialisation and setting in one call: C<<
882 ev_TYPE_init (watcher *, callback, ...) >>.
883
884 To make the watcher actually watch out for events, you have to start it
885 with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
886 *) >>), and you can stop watching for events at any time by calling the
887 corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
888
889 As long as your watcher is active (has been started but not stopped) you
890 must not touch the values stored in it. Most specifically you must never
891 reinitialise it or call its C<ev_TYPE_set> macro.
892
893 Each and every callback receives the event loop pointer as first, the
894 registered watcher structure as second, and a bitset of received events as
895 third argument.
896
897 The received events usually include a single bit per event type received
898 (you can receive multiple events at the same time). The possible bit masks
899 are:
900
901 =over 4
902
903 =item C<EV_READ>
904
905 =item C<EV_WRITE>
906
907 The file descriptor in the C<ev_io> watcher has become readable and/or
908 writable.
909
910 =item C<EV_TIMEOUT>
911
912 The C<ev_timer> watcher has timed out.
913
914 =item C<EV_PERIODIC>
915
916 The C<ev_periodic> watcher has timed out.
917
918 =item C<EV_SIGNAL>
919
920 The signal specified in the C<ev_signal> watcher has been received by a thread.
921
922 =item C<EV_CHILD>
923
924 The pid specified in the C<ev_child> watcher has received a status change.
925
926 =item C<EV_STAT>
927
928 The path specified in the C<ev_stat> watcher changed its attributes somehow.
929
930 =item C<EV_IDLE>
931
932 The C<ev_idle> watcher has determined that you have nothing better to do.
933
934 =item C<EV_PREPARE>
935
936 =item C<EV_CHECK>
937
938 All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
939 to gather new events, and all C<ev_check> watchers are invoked just after
940 C<ev_loop> has gathered them, but before it invokes any callbacks for any
941 received events. Callbacks of both watcher types can start and stop as
942 many watchers as they want, and all of them will be taken into account
943 (for example, a C<ev_prepare> watcher might start an idle watcher to keep
944 C<ev_loop> from blocking).
945
946 =item C<EV_EMBED>
947
948 The embedded event loop specified in the C<ev_embed> watcher needs attention.
949
950 =item C<EV_FORK>
951
952 The event loop has been resumed in the child process after fork (see
953 C<ev_fork>).
954
955 =item C<EV_ASYNC>
956
957 The given async watcher has been asynchronously notified (see C<ev_async>).
958
959 =item C<EV_CUSTOM>
960
961 Not ever sent (or otherwise used) by libev itself, but can be freely used
962 by libev users to signal watchers (e.g. via C<ev_feed_event>).
963
964 =item C<EV_ERROR>
965
966 An unspecified error has occurred, the watcher has been stopped. This might
967 happen because the watcher could not be properly started because libev
968 ran out of memory, a file descriptor was found to be closed or any other
969 problem. Libev considers these application bugs.
970
971 You best act on it by reporting the problem and somehow coping with the
972 watcher being stopped. Note that well-written programs should not receive
973 an error ever, so when your watcher receives it, this usually indicates a
974 bug in your program.
975
976 Libev will usually signal a few "dummy" events together with an error, for
977 example it might indicate that a fd is readable or writable, and if your
978 callbacks is well-written it can just attempt the operation and cope with
979 the error from read() or write(). This will not work in multi-threaded
980 programs, though, as the fd could already be closed and reused for another
981 thing, so beware.
982
983 =back
984
985 =head2 GENERIC WATCHER FUNCTIONS
986
987 =over 4
988
989 =item C<ev_init> (ev_TYPE *watcher, callback)
990
991 This macro initialises the generic portion of a watcher. The contents
992 of the watcher object can be arbitrary (so C<malloc> will do). Only
993 the generic parts of the watcher are initialised, you I<need> to call
994 the type-specific C<ev_TYPE_set> macro afterwards to initialise the
995 type-specific parts. For each type there is also a C<ev_TYPE_init> macro
996 which rolls both calls into one.
997
998 You can reinitialise a watcher at any time as long as it has been stopped
999 (or never started) and there are no pending events outstanding.
1000
1001 The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
1002 int revents)>.
1003
1004 Example: Initialise an C<ev_io> watcher in two steps.
1005
1006 ev_io w;
1007 ev_init (&w, my_cb);
1008 ev_io_set (&w, STDIN_FILENO, EV_READ);
1009
1010 =item C<ev_TYPE_set> (ev_TYPE *, [args])
1011
1012 This macro initialises the type-specific parts of a watcher. You need to
1013 call C<ev_init> at least once before you call this macro, but you can
1014 call C<ev_TYPE_set> any number of times. You must not, however, call this
1015 macro on a watcher that is active (it can be pending, however, which is a
1016 difference to the C<ev_init> macro).
1017
1018 Although some watcher types do not have type-specific arguments
1019 (e.g. C<ev_prepare>) you still need to call its C<set> macro.
1020
1021 See C<ev_init>, above, for an example.
1022
1023 =item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
1024
1025 This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
1026 calls into a single call. This is the most convenient method to initialise
1027 a watcher. The same limitations apply, of course.
1028
1029 Example: Initialise and set an C<ev_io> watcher in one step.
1030
1031 ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
1032
1033 =item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
1034
1035 Starts (activates) the given watcher. Only active watchers will receive
1036 events. If the watcher is already active nothing will happen.
1037
1038 Example: Start the C<ev_io> watcher that is being abused as example in this
1039 whole section.
1040
1041 ev_io_start (EV_DEFAULT_UC, &w);
1042
1043 =item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
1044
1045 Stops the given watcher if active, and clears the pending status (whether
1046 the watcher was active or not).
1047
1048 It is possible that stopped watchers are pending - for example,
1049 non-repeating timers are being stopped when they become pending - but
1050 calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
1051 pending. If you want to free or reuse the memory used by the watcher it is
1052 therefore a good idea to always call its C<ev_TYPE_stop> function.
1053
1054 =item bool ev_is_active (ev_TYPE *watcher)
1055
1056 Returns a true value iff the watcher is active (i.e. it has been started
1057 and not yet been stopped). As long as a watcher is active you must not modify
1058 it.
1059
1060 =item bool ev_is_pending (ev_TYPE *watcher)
1061
1062 Returns a true value iff the watcher is pending, (i.e. it has outstanding
1063 events but its callback has not yet been invoked). As long as a watcher
1064 is pending (but not active) you must not call an init function on it (but
1065 C<ev_TYPE_set> is safe), you must not change its priority, and you must
1066 make sure the watcher is available to libev (e.g. you cannot C<free ()>
1067 it).
1068
1069 =item callback ev_cb (ev_TYPE *watcher)
1070
1071 Returns the callback currently set on the watcher.
1072
1073 =item ev_cb_set (ev_TYPE *watcher, callback)
1074
1075 Change the callback. You can change the callback at virtually any time
1076 (modulo threads).
1077
1078 =item ev_set_priority (ev_TYPE *watcher, priority)
1079
1080 =item int ev_priority (ev_TYPE *watcher)
1081
1082 Set and query the priority of the watcher. The priority is a small
1083 integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
1084 (default: C<-2>). Pending watchers with higher priority will be invoked
1085 before watchers with lower priority, but priority will not keep watchers
1086 from being executed (except for C<ev_idle> watchers).
1087
1088 If you need to suppress invocation when higher priority events are pending
1089 you need to look at C<ev_idle> watchers, which provide this functionality.
1090
1091 You I<must not> change the priority of a watcher as long as it is active or
1092 pending.
1093
1094 Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
1095 fine, as long as you do not mind that the priority value you query might
1096 or might not have been clamped to the valid range.
1097
1098 The default priority used by watchers when no priority has been set is
1099 always C<0>, which is supposed to not be too high and not be too low :).
1100
1101 See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1102 priorities.
1103
1104 =item ev_invoke (loop, ev_TYPE *watcher, int revents)
1105
1106 Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1107 C<loop> nor C<revents> need to be valid as long as the watcher callback
1108 can deal with that fact, as both are simply passed through to the
1109 callback.
1110
1111 =item int ev_clear_pending (loop, ev_TYPE *watcher)
1112
1113 If the watcher is pending, this function clears its pending status and
1114 returns its C<revents> bitset (as if its callback was invoked). If the
1115 watcher isn't pending it does nothing and returns C<0>.
1116
1117 Sometimes it can be useful to "poll" a watcher instead of waiting for its
1118 callback to be invoked, which can be accomplished with this function.
1119
1120 =back
1121
1122
1123 =head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
1124
1125 Each watcher has, by default, a member C<void *data> that you can change
1126 and read at any time: libev will completely ignore it. This can be used
1127 to associate arbitrary data with your watcher. If you need more data and
1128 don't want to allocate memory and store a pointer to it in that data
1129 member, you can also "subclass" the watcher type and provide your own
1130 data:
1131
1132 struct my_io
1133 {
1134 ev_io io;
1135 int otherfd;
1136 void *somedata;
1137 struct whatever *mostinteresting;
1138 };
1139
1140 ...
1141 struct my_io w;
1142 ev_io_init (&w.io, my_cb, fd, EV_READ);
1143
1144 And since your callback will be called with a pointer to the watcher, you
1145 can cast it back to your own type:
1146
1147 static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
1148 {
1149 struct my_io *w = (struct my_io *)w_;
1150 ...
1151 }
1152
1153 More interesting and less C-conformant ways of casting your callback type
1154 instead have been omitted.
1155
1156 Another common scenario is to use some data structure with multiple
1157 embedded watchers:
1158
1159 struct my_biggy
1160 {
1161 int some_data;
1162 ev_timer t1;
1163 ev_timer t2;
1164 }
1165
1166 In this case getting the pointer to C<my_biggy> is a bit more
1167 complicated: Either you store the address of your C<my_biggy> struct
1168 in the C<data> member of the watcher (for woozies), or you need to use
1169 some pointer arithmetic using C<offsetof> inside your watchers (for real
1170 programmers):
1171
1172 #include <stddef.h>
1173
1174 static void
1175 t1_cb (EV_P_ ev_timer *w, int revents)
1176 {
1177 struct my_biggy big = (struct my_biggy *
1178 (((char *)w) - offsetof (struct my_biggy, t1));
1179 }
1180
1181 static void
1182 t2_cb (EV_P_ ev_timer *w, int revents)
1183 {
1184 struct my_biggy big = (struct my_biggy *
1185 (((char *)w) - offsetof (struct my_biggy, t2));
1186 }
1187
1188 =head2 WATCHER PRIORITY MODELS
1189
1190 Many event loops support I<watcher priorities>, which are usually small
1191 integers that influence the ordering of event callback invocation
1192 between watchers in some way, all else being equal.
1193
1194 In libev, Watcher priorities can be set using C<ev_set_priority>. See its
1195 description for the more technical details such as the actual priority
1196 range.
1197
1198 There are two common ways how these these priorities are being interpreted
1199 by event loops:
1200
1201 In the more common lock-out model, higher priorities "lock out" invocation
1202 of lower priority watchers, which means as long as higher priority
1203 watchers receive events, lower priority watchers are not being invoked.
1204
1205 The less common only-for-ordering model uses priorities solely to order
1206 callback invocation within a single event loop iteration: Higher priority
1207 watchers are invoked before lower priority ones, but they all get invoked
1208 before polling for new events.
1209
1210 Libev uses the second (only-for-ordering) model for all its watchers
1211 except for idle watchers (which use the lock-out model).
1212
1213 The rationale behind this is that implementing the lock-out model for
1214 watchers is not well supported by most kernel interfaces, and most event
1215 libraries will just poll for the same events again and again as long as
1216 their callbacks have not been executed, which is very inefficient in the
1217 common case of one high-priority watcher locking out a mass of lower
1218 priority ones.
1219
1220 Static (ordering) priorities are most useful when you have two or more
1221 watchers handling the same resource: a typical usage example is having an
1222 C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
1223 timeouts. Under load, data might be received while the program handles
1224 other jobs, but since timers normally get invoked first, the timeout
1225 handler will be executed before checking for data. In that case, giving
1226 the timer a lower priority than the I/O watcher ensures that I/O will be
1227 handled first even under adverse conditions (which is usually, but not
1228 always, what you want).
1229
1230 Since idle watchers use the "lock-out" model, meaning that idle watchers
1231 will only be executed when no same or higher priority watchers have
1232 received events, they can be used to implement the "lock-out" model when
1233 required.
1234
1235 For example, to emulate how many other event libraries handle priorities,
1236 you can associate an C<ev_idle> watcher to each such watcher, and in
1237 the normal watcher callback, you just start the idle watcher. The real
1238 processing is done in the idle watcher callback. This causes libev to
1239 continously poll and process kernel event data for the watcher, but when
1240 the lock-out case is known to be rare (which in turn is rare :), this is
1241 workable.
1242
1243 Usually, however, the lock-out model implemented that way will perform
1244 miserably under the type of load it was designed to handle. In that case,
1245 it might be preferable to stop the real watcher before starting the
1246 idle watcher, so the kernel will not have to process the event in case
1247 the actual processing will be delayed for considerable time.
1248
1249 Here is an example of an I/O watcher that should run at a strictly lower
1250 priority than the default, and which should only process data when no
1251 other events are pending:
1252
1253 ev_idle idle; // actual processing watcher
1254 ev_io io; // actual event watcher
1255
1256 static void
1257 io_cb (EV_P_ ev_io *w, int revents)
1258 {
1259 // stop the I/O watcher, we received the event, but
1260 // are not yet ready to handle it.
1261 ev_io_stop (EV_A_ w);
1262
1263 // start the idle watcher to ahndle the actual event.
1264 // it will not be executed as long as other watchers
1265 // with the default priority are receiving events.
1266 ev_idle_start (EV_A_ &idle);
1267 }
1268
1269 static void
1270 idle-cb (EV_P_ ev_idle *w, int revents)
1271 {
1272 // actual processing
1273 read (STDIN_FILENO, ...);
1274
1275 // have to start the I/O watcher again, as
1276 // we have handled the event
1277 ev_io_start (EV_P_ &io);
1278 }
1279
1280 // initialisation
1281 ev_idle_init (&idle, idle_cb);
1282 ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
1283 ev_io_start (EV_DEFAULT_ &io);
1284
1285 In the "real" world, it might also be beneficial to start a timer, so that
1286 low-priority connections can not be locked out forever under load. This
1287 enables your program to keep a lower latency for important connections
1288 during short periods of high load, while not completely locking out less
1289 important ones.
1290
1291
1292 =head1 WATCHER TYPES
1293
1294 This section describes each watcher in detail, but will not repeat
1295 information given in the last section. Any initialisation/set macros,
1296 functions and members specific to the watcher type are explained.
1297
1298 Members are additionally marked with either I<[read-only]>, meaning that,
1299 while the watcher is active, you can look at the member and expect some
1300 sensible content, but you must not modify it (you can modify it while the
1301 watcher is stopped to your hearts content), or I<[read-write]>, which
1302 means you can expect it to have some sensible content while the watcher
1303 is active, but you can also modify it. Modifying it may not do something
1304 sensible or take immediate effect (or do anything at all), but libev will
1305 not crash or malfunction in any way.
1306
1307
1308 =head2 C<ev_io> - is this file descriptor readable or writable?
1309
1310 I/O watchers check whether a file descriptor is readable or writable
1311 in each iteration of the event loop, or, more precisely, when reading
1312 would not block the process and writing would at least be able to write
1313 some data. This behaviour is called level-triggering because you keep
1314 receiving events as long as the condition persists. Remember you can stop
1315 the watcher if you don't want to act on the event and neither want to
1316 receive future events.
1317
1318 In general you can register as many read and/or write event watchers per
1319 fd as you want (as long as you don't confuse yourself). Setting all file
1320 descriptors to non-blocking mode is also usually a good idea (but not
1321 required if you know what you are doing).
1322
1323 If you cannot use non-blocking mode, then force the use of a
1324 known-to-be-good backend (at the time of this writing, this includes only
1325 C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
1326
1327 Another thing you have to watch out for is that it is quite easy to
1328 receive "spurious" readiness notifications, that is your callback might
1329 be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1330 because there is no data. Not only are some backends known to create a
1331 lot of those (for example Solaris ports), it is very easy to get into
1332 this situation even with a relatively standard program structure. Thus
1333 it is best to always use non-blocking I/O: An extra C<read>(2) returning
1334 C<EAGAIN> is far preferable to a program hanging until some data arrives.
1335
1336 If you cannot run the fd in non-blocking mode (for example you should
1337 not play around with an Xlib connection), then you have to separately
1338 re-test whether a file descriptor is really ready with a known-to-be good
1339 interface such as poll (fortunately in our Xlib example, Xlib already
1340 does this on its own, so its quite safe to use). Some people additionally
1341 use C<SIGALRM> and an interval timer, just to be sure you won't block
1342 indefinitely.
1343
1344 But really, best use non-blocking mode.
1345
1346 =head3 The special problem of disappearing file descriptors
1347
1348 Some backends (e.g. kqueue, epoll) need to be told about closing a file
1349 descriptor (either due to calling C<close> explicitly or any other means,
1350 such as C<dup2>). The reason is that you register interest in some file
1351 descriptor, but when it goes away, the operating system will silently drop
1352 this interest. If another file descriptor with the same number then is
1353 registered with libev, there is no efficient way to see that this is, in
1354 fact, a different file descriptor.
1355
1356 To avoid having to explicitly tell libev about such cases, libev follows
1357 the following policy: Each time C<ev_io_set> is being called, libev
1358 will assume that this is potentially a new file descriptor, otherwise
1359 it is assumed that the file descriptor stays the same. That means that
1360 you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
1361 descriptor even if the file descriptor number itself did not change.
1362
1363 This is how one would do it normally anyway, the important point is that
1364 the libev application should not optimise around libev but should leave
1365 optimisations to libev.
1366
1367 =head3 The special problem of dup'ed file descriptors
1368
1369 Some backends (e.g. epoll), cannot register events for file descriptors,
1370 but only events for the underlying file descriptions. That means when you
1371 have C<dup ()>'ed file descriptors or weirder constellations, and register
1372 events for them, only one file descriptor might actually receive events.
1373
1374 There is no workaround possible except not registering events
1375 for potentially C<dup ()>'ed file descriptors, or to resort to
1376 C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1377
1378 =head3 The special problem of fork
1379
1380 Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1381 useless behaviour. Libev fully supports fork, but needs to be told about
1382 it in the child.
1383
1384 To support fork in your programs, you either have to call
1385 C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
1386 enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1387 C<EVBACKEND_POLL>.
1388
1389 =head3 The special problem of SIGPIPE
1390
1391 While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1392 when writing to a pipe whose other end has been closed, your program gets
1393 sent a SIGPIPE, which, by default, aborts your program. For most programs
1394 this is sensible behaviour, for daemons, this is usually undesirable.
1395
1396 So when you encounter spurious, unexplained daemon exits, make sure you
1397 ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1398 somewhere, as that would have given you a big clue).
1399
1400
1401 =head3 Watcher-Specific Functions
1402
1403 =over 4
1404
1405 =item ev_io_init (ev_io *, callback, int fd, int events)
1406
1407 =item ev_io_set (ev_io *, int fd, int events)
1408
1409 Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1410 receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
1411 C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
1412
1413 =item int fd [read-only]
1414
1415 The file descriptor being watched.
1416
1417 =item int events [read-only]
1418
1419 The events being watched.
1420
1421 =back
1422
1423 =head3 Examples
1424
1425 Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1426 readable, but only once. Since it is likely line-buffered, you could
1427 attempt to read a whole line in the callback.
1428
1429 static void
1430 stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
1431 {
1432 ev_io_stop (loop, w);
1433 .. read from stdin here (or from w->fd) and handle any I/O errors
1434 }
1435
1436 ...
1437 struct ev_loop *loop = ev_default_init (0);
1438 ev_io stdin_readable;
1439 ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1440 ev_io_start (loop, &stdin_readable);
1441 ev_loop (loop, 0);
1442
1443
1444 =head2 C<ev_timer> - relative and optionally repeating timeouts
1445
1446 Timer watchers are simple relative timers that generate an event after a
1447 given time, and optionally repeating in regular intervals after that.
1448
1449 The timers are based on real time, that is, if you register an event that
1450 times out after an hour and you reset your system clock to January last
1451 year, it will still time out after (roughly) one hour. "Roughly" because
1452 detecting time jumps is hard, and some inaccuracies are unavoidable (the
1453 monotonic clock option helps a lot here).
1454
1455 The callback is guaranteed to be invoked only I<after> its timeout has
1456 passed. If multiple timers become ready during the same loop iteration
1457 then the ones with earlier time-out values are invoked before ones with
1458 later time-out values (but this is no longer true when a callback calls
1459 C<ev_loop> recursively).
1460
1461 =head3 Be smart about timeouts
1462
1463 Many real-world problems involve some kind of timeout, usually for error
1464 recovery. A typical example is an HTTP request - if the other side hangs,
1465 you want to raise some error after a while.
1466
1467 What follows are some ways to handle this problem, from obvious and
1468 inefficient to smart and efficient.
1469
1470 In the following, a 60 second activity timeout is assumed - a timeout that
1471 gets reset to 60 seconds each time there is activity (e.g. each time some
1472 data or other life sign was received).
1473
1474 =over 4
1475
1476 =item 1. Use a timer and stop, reinitialise and start it on activity.
1477
1478 This is the most obvious, but not the most simple way: In the beginning,
1479 start the watcher:
1480
1481 ev_timer_init (timer, callback, 60., 0.);
1482 ev_timer_start (loop, timer);
1483
1484 Then, each time there is some activity, C<ev_timer_stop> it, initialise it
1485 and start it again:
1486
1487 ev_timer_stop (loop, timer);
1488 ev_timer_set (timer, 60., 0.);
1489 ev_timer_start (loop, timer);
1490
1491 This is relatively simple to implement, but means that each time there is
1492 some activity, libev will first have to remove the timer from its internal
1493 data structure and then add it again. Libev tries to be fast, but it's
1494 still not a constant-time operation.
1495
1496 =item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
1497
1498 This is the easiest way, and involves using C<ev_timer_again> instead of
1499 C<ev_timer_start>.
1500
1501 To implement this, configure an C<ev_timer> with a C<repeat> value
1502 of C<60> and then call C<ev_timer_again> at start and each time you
1503 successfully read or write some data. If you go into an idle state where
1504 you do not expect data to travel on the socket, you can C<ev_timer_stop>
1505 the timer, and C<ev_timer_again> will automatically restart it if need be.
1506
1507 That means you can ignore both the C<ev_timer_start> function and the
1508 C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
1509 member and C<ev_timer_again>.
1510
1511 At start:
1512
1513 ev_timer_init (timer, callback);
1514 timer->repeat = 60.;
1515 ev_timer_again (loop, timer);
1516
1517 Each time there is some activity:
1518
1519 ev_timer_again (loop, timer);
1520
1521 It is even possible to change the time-out on the fly, regardless of
1522 whether the watcher is active or not:
1523
1524 timer->repeat = 30.;
1525 ev_timer_again (loop, timer);
1526
1527 This is slightly more efficient then stopping/starting the timer each time
1528 you want to modify its timeout value, as libev does not have to completely
1529 remove and re-insert the timer from/into its internal data structure.
1530
1531 It is, however, even simpler than the "obvious" way to do it.
1532
1533 =item 3. Let the timer time out, but then re-arm it as required.
1534
1535 This method is more tricky, but usually most efficient: Most timeouts are
1536 relatively long compared to the intervals between other activity - in
1537 our example, within 60 seconds, there are usually many I/O events with
1538 associated activity resets.
1539
1540 In this case, it would be more efficient to leave the C<ev_timer> alone,
1541 but remember the time of last activity, and check for a real timeout only
1542 within the callback:
1543
1544 ev_tstamp last_activity; // time of last activity
1545
1546 static void
1547 callback (EV_P_ ev_timer *w, int revents)
1548 {
1549 ev_tstamp now = ev_now (EV_A);
1550 ev_tstamp timeout = last_activity + 60.;
1551
1552 // if last_activity + 60. is older than now, we did time out
1553 if (timeout < now)
1554 {
1555 // timeout occured, take action
1556 }
1557 else
1558 {
1559 // callback was invoked, but there was some activity, re-arm
1560 // the watcher to fire in last_activity + 60, which is
1561 // guaranteed to be in the future, so "again" is positive:
1562 w->repeat = timeout - now;
1563 ev_timer_again (EV_A_ w);
1564 }
1565 }
1566
1567 To summarise the callback: first calculate the real timeout (defined
1568 as "60 seconds after the last activity"), then check if that time has
1569 been reached, which means something I<did>, in fact, time out. Otherwise
1570 the callback was invoked too early (C<timeout> is in the future), so
1571 re-schedule the timer to fire at that future time, to see if maybe we have
1572 a timeout then.
1573
1574 Note how C<ev_timer_again> is used, taking advantage of the
1575 C<ev_timer_again> optimisation when the timer is already running.
1576
1577 This scheme causes more callback invocations (about one every 60 seconds
1578 minus half the average time between activity), but virtually no calls to
1579 libev to change the timeout.
1580
1581 To start the timer, simply initialise the watcher and set C<last_activity>
1582 to the current time (meaning we just have some activity :), then call the
1583 callback, which will "do the right thing" and start the timer:
1584
1585 ev_timer_init (timer, callback);
1586 last_activity = ev_now (loop);
1587 callback (loop, timer, EV_TIMEOUT);
1588
1589 And when there is some activity, simply store the current time in
1590 C<last_activity>, no libev calls at all:
1591
1592 last_actiivty = ev_now (loop);
1593
1594 This technique is slightly more complex, but in most cases where the
1595 time-out is unlikely to be triggered, much more efficient.
1596
1597 Changing the timeout is trivial as well (if it isn't hard-coded in the
1598 callback :) - just change the timeout and invoke the callback, which will
1599 fix things for you.
1600
1601 =item 4. Wee, just use a double-linked list for your timeouts.
1602
1603 If there is not one request, but many thousands (millions...), all
1604 employing some kind of timeout with the same timeout value, then one can
1605 do even better:
1606
1607 When starting the timeout, calculate the timeout value and put the timeout
1608 at the I<end> of the list.
1609
1610 Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
1611 the list is expected to fire (for example, using the technique #3).
1612
1613 When there is some activity, remove the timer from the list, recalculate
1614 the timeout, append it to the end of the list again, and make sure to
1615 update the C<ev_timer> if it was taken from the beginning of the list.
1616
1617 This way, one can manage an unlimited number of timeouts in O(1) time for
1618 starting, stopping and updating the timers, at the expense of a major
1619 complication, and having to use a constant timeout. The constant timeout
1620 ensures that the list stays sorted.
1621
1622 =back
1623
1624 So which method the best?
1625
1626 Method #2 is a simple no-brain-required solution that is adequate in most
1627 situations. Method #3 requires a bit more thinking, but handles many cases
1628 better, and isn't very complicated either. In most case, choosing either
1629 one is fine, with #3 being better in typical situations.
1630
1631 Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1632 rather complicated, but extremely efficient, something that really pays
1633 off after the first million or so of active timers, i.e. it's usually
1634 overkill :)
1635
1636 =head3 The special problem of time updates
1637
1638 Establishing the current time is a costly operation (it usually takes at
1639 least two system calls): EV therefore updates its idea of the current
1640 time only before and after C<ev_loop> collects new events, which causes a
1641 growing difference between C<ev_now ()> and C<ev_time ()> when handling
1642 lots of events in one iteration.
1643
1644 The relative timeouts are calculated relative to the C<ev_now ()>
1645 time. This is usually the right thing as this timestamp refers to the time
1646 of the event triggering whatever timeout you are modifying/starting. If
1647 you suspect event processing to be delayed and you I<need> to base the
1648 timeout on the current time, use something like this to adjust for this:
1649
1650 ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1651
1652 If the event loop is suspended for a long time, you can also force an
1653 update of the time returned by C<ev_now ()> by calling C<ev_now_update
1654 ()>.
1655
1656 =head3 Watcher-Specific Functions and Data Members
1657
1658 =over 4
1659
1660 =item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1661
1662 =item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1663
1664 Configure the timer to trigger after C<after> seconds. If C<repeat>
1665 is C<0.>, then it will automatically be stopped once the timeout is
1666 reached. If it is positive, then the timer will automatically be
1667 configured to trigger again C<repeat> seconds later, again, and again,
1668 until stopped manually.
1669
1670 The timer itself will do a best-effort at avoiding drift, that is, if
1671 you configure a timer to trigger every 10 seconds, then it will normally
1672 trigger at exactly 10 second intervals. If, however, your program cannot
1673 keep up with the timer (because it takes longer than those 10 seconds to
1674 do stuff) the timer will not fire more than once per event loop iteration.
1675
1676 =item ev_timer_again (loop, ev_timer *)
1677
1678 This will act as if the timer timed out and restart it again if it is
1679 repeating. The exact semantics are:
1680
1681 If the timer is pending, its pending status is cleared.
1682
1683 If the timer is started but non-repeating, stop it (as if it timed out).
1684
1685 If the timer is repeating, either start it if necessary (with the
1686 C<repeat> value), or reset the running timer to the C<repeat> value.
1687
1688 This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
1689 usage example.
1690
1691 =item ev_tstamp repeat [read-write]
1692
1693 The current C<repeat> value. Will be used each time the watcher times out
1694 or C<ev_timer_again> is called, and determines the next timeout (if any),
1695 which is also when any modifications are taken into account.
1696
1697 =back
1698
1699 =head3 Examples
1700
1701 Example: Create a timer that fires after 60 seconds.
1702
1703 static void
1704 one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
1705 {
1706 .. one minute over, w is actually stopped right here
1707 }
1708
1709 ev_timer mytimer;
1710 ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1711 ev_timer_start (loop, &mytimer);
1712
1713 Example: Create a timeout timer that times out after 10 seconds of
1714 inactivity.
1715
1716 static void
1717 timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
1718 {
1719 .. ten seconds without any activity
1720 }
1721
1722 ev_timer mytimer;
1723 ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1724 ev_timer_again (&mytimer); /* start timer */
1725 ev_loop (loop, 0);
1726
1727 // and in some piece of code that gets executed on any "activity":
1728 // reset the timeout to start ticking again at 10 seconds
1729 ev_timer_again (&mytimer);
1730
1731
1732 =head2 C<ev_periodic> - to cron or not to cron?
1733
1734 Periodic watchers are also timers of a kind, but they are very versatile
1735 (and unfortunately a bit complex).
1736
1737 Unlike C<ev_timer>, periodic watchers are not based on real time (or
1738 relative time, the physical time that passes) but on wall clock time
1739 (absolute time, the thing you can read on your calender or clock). The
1740 difference is that wall clock time can run faster or slower than real
1741 time, and time jumps are not uncommon (e.g. when you adjust your
1742 wrist-watch).
1743
1744 You can tell a periodic watcher to trigger after some specific point
1745 in time: for example, if you tell a periodic watcher to trigger "in 10
1746 seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
1747 not a delay) and then reset your system clock to January of the previous
1748 year, then it will take a year or more to trigger the event (unlike an
1749 C<ev_timer>, which would still trigger roughly 10 seconds after starting
1750 it, as it uses a relative timeout).
1751
1752 C<ev_periodic> watchers can also be used to implement vastly more complex
1753 timers, such as triggering an event on each "midnight, local time", or
1754 other complicated rules. This cannot be done with C<ev_timer> watchers, as
1755 those cannot react to time jumps.
1756
1757 As with timers, the callback is guaranteed to be invoked only when the
1758 point in time where it is supposed to trigger has passed. If multiple
1759 timers become ready during the same loop iteration then the ones with
1760 earlier time-out values are invoked before ones with later time-out values
1761 (but this is no longer true when a callback calls C<ev_loop> recursively).
1762
1763 =head3 Watcher-Specific Functions and Data Members
1764
1765 =over 4
1766
1767 =item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1768
1769 =item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1770
1771 Lots of arguments, let's sort it out... There are basically three modes of
1772 operation, and we will explain them from simplest to most complex:
1773
1774 =over 4
1775
1776 =item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
1777
1778 In this configuration the watcher triggers an event after the wall clock
1779 time C<offset> has passed. It will not repeat and will not adjust when a
1780 time jump occurs, that is, if it is to be run at January 1st 2011 then it
1781 will be stopped and invoked when the system clock reaches or surpasses
1782 this point in time.
1783
1784 =item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
1785
1786 In this mode the watcher will always be scheduled to time out at the next
1787 C<offset + N * interval> time (for some integer N, which can also be
1788 negative) and then repeat, regardless of any time jumps. The C<offset>
1789 argument is merely an offset into the C<interval> periods.
1790
1791 This can be used to create timers that do not drift with respect to the
1792 system clock, for example, here is an C<ev_periodic> that triggers each
1793 hour, on the hour (with respect to UTC):
1794
1795 ev_periodic_set (&periodic, 0., 3600., 0);
1796
1797 This doesn't mean there will always be 3600 seconds in between triggers,
1798 but only that the callback will be called when the system time shows a
1799 full hour (UTC), or more correctly, when the system time is evenly divisible
1800 by 3600.
1801
1802 Another way to think about it (for the mathematically inclined) is that
1803 C<ev_periodic> will try to run the callback in this mode at the next possible
1804 time where C<time = offset (mod interval)>, regardless of any time jumps.
1805
1806 For numerical stability it is preferable that the C<offset> value is near
1807 C<ev_now ()> (the current time), but there is no range requirement for
1808 this value, and in fact is often specified as zero.
1809
1810 Note also that there is an upper limit to how often a timer can fire (CPU
1811 speed for example), so if C<interval> is very small then timing stability
1812 will of course deteriorate. Libev itself tries to be exact to be about one
1813 millisecond (if the OS supports it and the machine is fast enough).
1814
1815 =item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
1816
1817 In this mode the values for C<interval> and C<offset> are both being
1818 ignored. Instead, each time the periodic watcher gets scheduled, the
1819 reschedule callback will be called with the watcher as first, and the
1820 current time as second argument.
1821
1822 NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
1823 or make ANY other event loop modifications whatsoever, unless explicitly
1824 allowed by documentation here>.
1825
1826 If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
1827 it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
1828 only event loop modification you are allowed to do).
1829
1830 The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
1831 *w, ev_tstamp now)>, e.g.:
1832
1833 static ev_tstamp
1834 my_rescheduler (ev_periodic *w, ev_tstamp now)
1835 {
1836 return now + 60.;
1837 }
1838
1839 It must return the next time to trigger, based on the passed time value
1840 (that is, the lowest time value larger than to the second argument). It
1841 will usually be called just before the callback will be triggered, but
1842 might be called at other times, too.
1843
1844 NOTE: I<< This callback must always return a time that is higher than or
1845 equal to the passed C<now> value >>.
1846
1847 This can be used to create very complex timers, such as a timer that
1848 triggers on "next midnight, local time". To do this, you would calculate the
1849 next midnight after C<now> and return the timestamp value for this. How
1850 you do this is, again, up to you (but it is not trivial, which is the main
1851 reason I omitted it as an example).
1852
1853 =back
1854
1855 =item ev_periodic_again (loop, ev_periodic *)
1856
1857 Simply stops and restarts the periodic watcher again. This is only useful
1858 when you changed some parameters or the reschedule callback would return
1859 a different time than the last time it was called (e.g. in a crond like
1860 program when the crontabs have changed).
1861
1862 =item ev_tstamp ev_periodic_at (ev_periodic *)
1863
1864 When active, returns the absolute time that the watcher is supposed
1865 to trigger next. This is not the same as the C<offset> argument to
1866 C<ev_periodic_set>, but indeed works even in interval and manual
1867 rescheduling modes.
1868
1869 =item ev_tstamp offset [read-write]
1870
1871 When repeating, this contains the offset value, otherwise this is the
1872 absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
1873 although libev might modify this value for better numerical stability).
1874
1875 Can be modified any time, but changes only take effect when the periodic
1876 timer fires or C<ev_periodic_again> is being called.
1877
1878 =item ev_tstamp interval [read-write]
1879
1880 The current interval value. Can be modified any time, but changes only
1881 take effect when the periodic timer fires or C<ev_periodic_again> is being
1882 called.
1883
1884 =item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
1885
1886 The current reschedule callback, or C<0>, if this functionality is
1887 switched off. Can be changed any time, but changes only take effect when
1888 the periodic timer fires or C<ev_periodic_again> is being called.
1889
1890 =back
1891
1892 =head3 Examples
1893
1894 Example: Call a callback every hour, or, more precisely, whenever the
1895 system time is divisible by 3600. The callback invocation times have
1896 potentially a lot of jitter, but good long-term stability.
1897
1898 static void
1899 clock_cb (struct ev_loop *loop, ev_io *w, int revents)
1900 {
1901 ... its now a full hour (UTC, or TAI or whatever your clock follows)
1902 }
1903
1904 ev_periodic hourly_tick;
1905 ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1906 ev_periodic_start (loop, &hourly_tick);
1907
1908 Example: The same as above, but use a reschedule callback to do it:
1909
1910 #include <math.h>
1911
1912 static ev_tstamp
1913 my_scheduler_cb (ev_periodic *w, ev_tstamp now)
1914 {
1915 return now + (3600. - fmod (now, 3600.));
1916 }
1917
1918 ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1919
1920 Example: Call a callback every hour, starting now:
1921
1922 ev_periodic hourly_tick;
1923 ev_periodic_init (&hourly_tick, clock_cb,
1924 fmod (ev_now (loop), 3600.), 3600., 0);
1925 ev_periodic_start (loop, &hourly_tick);
1926
1927
1928 =head2 C<ev_signal> - signal me when a signal gets signalled!
1929
1930 Signal watchers will trigger an event when the process receives a specific
1931 signal one or more times. Even though signals are very asynchronous, libev
1932 will try it's best to deliver signals synchronously, i.e. as part of the
1933 normal event processing, like any other event.
1934
1935 If you want signals asynchronously, just use C<sigaction> as you would
1936 do without libev and forget about sharing the signal. You can even use
1937 C<ev_async> from a signal handler to synchronously wake up an event loop.
1938
1939 You can configure as many watchers as you like per signal. Only when the
1940 first watcher gets started will libev actually register a signal handler
1941 with the kernel (thus it coexists with your own signal handlers as long as
1942 you don't register any with libev for the same signal). Similarly, when
1943 the last signal watcher for a signal is stopped, libev will reset the
1944 signal handler to SIG_DFL (regardless of what it was set to before).
1945
1946 If possible and supported, libev will install its handlers with
1947 C<SA_RESTART> behaviour enabled, so system calls should not be unduly
1948 interrupted. If you have a problem with system calls getting interrupted by
1949 signals you can block all signals in an C<ev_check> watcher and unblock
1950 them in an C<ev_prepare> watcher.
1951
1952 =head3 Watcher-Specific Functions and Data Members
1953
1954 =over 4
1955
1956 =item ev_signal_init (ev_signal *, callback, int signum)
1957
1958 =item ev_signal_set (ev_signal *, int signum)
1959
1960 Configures the watcher to trigger on the given signal number (usually one
1961 of the C<SIGxxx> constants).
1962
1963 =item int signum [read-only]
1964
1965 The signal the watcher watches out for.
1966
1967 =back
1968
1969 =head3 Examples
1970
1971 Example: Try to exit cleanly on SIGINT.
1972
1973 static void
1974 sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
1975 {
1976 ev_unloop (loop, EVUNLOOP_ALL);
1977 }
1978
1979 ev_signal signal_watcher;
1980 ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1981 ev_signal_start (loop, &signal_watcher);
1982
1983
1984 =head2 C<ev_child> - watch out for process status changes
1985
1986 Child watchers trigger when your process receives a SIGCHLD in response to
1987 some child status changes (most typically when a child of yours dies or
1988 exits). It is permissible to install a child watcher I<after> the child
1989 has been forked (which implies it might have already exited), as long
1990 as the event loop isn't entered (or is continued from a watcher), i.e.,
1991 forking and then immediately registering a watcher for the child is fine,
1992 but forking and registering a watcher a few event loop iterations later is
1993 not.
1994
1995 Only the default event loop is capable of handling signals, and therefore
1996 you can only register child watchers in the default event loop.
1997
1998 =head3 Process Interaction
1999
2000 Libev grabs C<SIGCHLD> as soon as the default event loop is
2001 initialised. This is necessary to guarantee proper behaviour even if
2002 the first child watcher is started after the child exits. The occurrence
2003 of C<SIGCHLD> is recorded asynchronously, but child reaping is done
2004 synchronously as part of the event loop processing. Libev always reaps all
2005 children, even ones not watched.
2006
2007 =head3 Overriding the Built-In Processing
2008
2009 Libev offers no special support for overriding the built-in child
2010 processing, but if your application collides with libev's default child
2011 handler, you can override it easily by installing your own handler for
2012 C<SIGCHLD> after initialising the default loop, and making sure the
2013 default loop never gets destroyed. You are encouraged, however, to use an
2014 event-based approach to child reaping and thus use libev's support for
2015 that, so other libev users can use C<ev_child> watchers freely.
2016
2017 =head3 Stopping the Child Watcher
2018
2019 Currently, the child watcher never gets stopped, even when the
2020 child terminates, so normally one needs to stop the watcher in the
2021 callback. Future versions of libev might stop the watcher automatically
2022 when a child exit is detected.
2023
2024 =head3 Watcher-Specific Functions and Data Members
2025
2026 =over 4
2027
2028 =item ev_child_init (ev_child *, callback, int pid, int trace)
2029
2030 =item ev_child_set (ev_child *, int pid, int trace)
2031
2032 Configures the watcher to wait for status changes of process C<pid> (or
2033 I<any> process if C<pid> is specified as C<0>). The callback can look
2034 at the C<rstatus> member of the C<ev_child> watcher structure to see
2035 the status word (use the macros from C<sys/wait.h> and see your systems
2036 C<waitpid> documentation). The C<rpid> member contains the pid of the
2037 process causing the status change. C<trace> must be either C<0> (only
2038 activate the watcher when the process terminates) or C<1> (additionally
2039 activate the watcher when the process is stopped or continued).
2040
2041 =item int pid [read-only]
2042
2043 The process id this watcher watches out for, or C<0>, meaning any process id.
2044
2045 =item int rpid [read-write]
2046
2047 The process id that detected a status change.
2048
2049 =item int rstatus [read-write]
2050
2051 The process exit/trace status caused by C<rpid> (see your systems
2052 C<waitpid> and C<sys/wait.h> documentation for details).
2053
2054 =back
2055
2056 =head3 Examples
2057
2058 Example: C<fork()> a new process and install a child handler to wait for
2059 its completion.
2060
2061 ev_child cw;
2062
2063 static void
2064 child_cb (EV_P_ ev_child *w, int revents)
2065 {
2066 ev_child_stop (EV_A_ w);
2067 printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
2068 }
2069
2070 pid_t pid = fork ();
2071
2072 if (pid < 0)
2073 // error
2074 else if (pid == 0)
2075 {
2076 // the forked child executes here
2077 exit (1);
2078 }
2079 else
2080 {
2081 ev_child_init (&cw, child_cb, pid, 0);
2082 ev_child_start (EV_DEFAULT_ &cw);
2083 }
2084
2085
2086 =head2 C<ev_stat> - did the file attributes just change?
2087
2088 This watches a file system path for attribute changes. That is, it calls
2089 C<stat> on that path in regular intervals (or when the OS says it changed)
2090 and sees if it changed compared to the last time, invoking the callback if
2091 it did.
2092
2093 The path does not need to exist: changing from "path exists" to "path does
2094 not exist" is a status change like any other. The condition "path does not
2095 exist" (or more correctly "path cannot be stat'ed") is signified by the
2096 C<st_nlink> field being zero (which is otherwise always forced to be at
2097 least one) and all the other fields of the stat buffer having unspecified
2098 contents.
2099
2100 The path I<must not> end in a slash or contain special components such as
2101 C<.> or C<..>. The path I<should> be absolute: If it is relative and
2102 your working directory changes, then the behaviour is undefined.
2103
2104 Since there is no portable change notification interface available, the
2105 portable implementation simply calls C<stat(2)> regularly on the path
2106 to see if it changed somehow. You can specify a recommended polling
2107 interval for this case. If you specify a polling interval of C<0> (highly
2108 recommended!) then a I<suitable, unspecified default> value will be used
2109 (which you can expect to be around five seconds, although this might
2110 change dynamically). Libev will also impose a minimum interval which is
2111 currently around C<0.1>, but that's usually overkill.
2112
2113 This watcher type is not meant for massive numbers of stat watchers,
2114 as even with OS-supported change notifications, this can be
2115 resource-intensive.
2116
2117 At the time of this writing, the only OS-specific interface implemented
2118 is the Linux inotify interface (implementing kqueue support is left as an
2119 exercise for the reader. Note, however, that the author sees no way of
2120 implementing C<ev_stat> semantics with kqueue, except as a hint).
2121
2122 =head3 ABI Issues (Largefile Support)
2123
2124 Libev by default (unless the user overrides this) uses the default
2125 compilation environment, which means that on systems with large file
2126 support disabled by default, you get the 32 bit version of the stat
2127 structure. When using the library from programs that change the ABI to
2128 use 64 bit file offsets the programs will fail. In that case you have to
2129 compile libev with the same flags to get binary compatibility. This is
2130 obviously the case with any flags that change the ABI, but the problem is
2131 most noticeably displayed with ev_stat and large file support.
2132
2133 The solution for this is to lobby your distribution maker to make large
2134 file interfaces available by default (as e.g. FreeBSD does) and not
2135 optional. Libev cannot simply switch on large file support because it has
2136 to exchange stat structures with application programs compiled using the
2137 default compilation environment.
2138
2139 =head3 Inotify and Kqueue
2140
2141 When C<inotify (7)> support has been compiled into libev and present at
2142 runtime, it will be used to speed up change detection where possible. The
2143 inotify descriptor will be created lazily when the first C<ev_stat>
2144 watcher is being started.
2145
2146 Inotify presence does not change the semantics of C<ev_stat> watchers
2147 except that changes might be detected earlier, and in some cases, to avoid
2148 making regular C<stat> calls. Even in the presence of inotify support
2149 there are many cases where libev has to resort to regular C<stat> polling,
2150 but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
2151 many bugs), the path exists (i.e. stat succeeds), and the path resides on
2152 a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
2153 xfs are fully working) libev usually gets away without polling.
2154
2155 There is no support for kqueue, as apparently it cannot be used to
2156 implement this functionality, due to the requirement of having a file
2157 descriptor open on the object at all times, and detecting renames, unlinks
2158 etc. is difficult.
2159
2160 =head3 C<stat ()> is a synchronous operation
2161
2162 Libev doesn't normally do any kind of I/O itself, and so is not blocking
2163 the process. The exception are C<ev_stat> watchers - those call C<stat
2164 ()>, which is a synchronous operation.
2165
2166 For local paths, this usually doesn't matter: unless the system is very
2167 busy or the intervals between stat's are large, a stat call will be fast,
2168 as the path data is usually in memory already (except when starting the
2169 watcher).
2170
2171 For networked file systems, calling C<stat ()> can block an indefinite
2172 time due to network issues, and even under good conditions, a stat call
2173 often takes multiple milliseconds.
2174
2175 Therefore, it is best to avoid using C<ev_stat> watchers on networked
2176 paths, although this is fully supported by libev.
2177
2178 =head3 The special problem of stat time resolution
2179
2180 The C<stat ()> system call only supports full-second resolution portably,
2181 and even on systems where the resolution is higher, most file systems
2182 still only support whole seconds.
2183
2184 That means that, if the time is the only thing that changes, you can
2185 easily miss updates: on the first update, C<ev_stat> detects a change and
2186 calls your callback, which does something. When there is another update
2187 within the same second, C<ev_stat> will be unable to detect unless the
2188 stat data does change in other ways (e.g. file size).
2189
2190 The solution to this is to delay acting on a change for slightly more
2191 than a second (or till slightly after the next full second boundary), using
2192 a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
2193 ev_timer_again (loop, w)>).
2194
2195 The C<.02> offset is added to work around small timing inconsistencies
2196 of some operating systems (where the second counter of the current time
2197 might be be delayed. One such system is the Linux kernel, where a call to
2198 C<gettimeofday> might return a timestamp with a full second later than
2199 a subsequent C<time> call - if the equivalent of C<time ()> is used to
2200 update file times then there will be a small window where the kernel uses
2201 the previous second to update file times but libev might already execute
2202 the timer callback).
2203
2204 =head3 Watcher-Specific Functions and Data Members
2205
2206 =over 4
2207
2208 =item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
2209
2210 =item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
2211
2212 Configures the watcher to wait for status changes of the given
2213 C<path>. The C<interval> is a hint on how quickly a change is expected to
2214 be detected and should normally be specified as C<0> to let libev choose
2215 a suitable value. The memory pointed to by C<path> must point to the same
2216 path for as long as the watcher is active.
2217
2218 The callback will receive an C<EV_STAT> event when a change was detected,
2219 relative to the attributes at the time the watcher was started (or the
2220 last change was detected).
2221
2222 =item ev_stat_stat (loop, ev_stat *)
2223
2224 Updates the stat buffer immediately with new values. If you change the
2225 watched path in your callback, you could call this function to avoid
2226 detecting this change (while introducing a race condition if you are not
2227 the only one changing the path). Can also be useful simply to find out the
2228 new values.
2229
2230 =item ev_statdata attr [read-only]
2231
2232 The most-recently detected attributes of the file. Although the type is
2233 C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
2234 suitable for your system, but you can only rely on the POSIX-standardised
2235 members to be present. If the C<st_nlink> member is C<0>, then there was
2236 some error while C<stat>ing the file.
2237
2238 =item ev_statdata prev [read-only]
2239
2240 The previous attributes of the file. The callback gets invoked whenever
2241 C<prev> != C<attr>, or, more precisely, one or more of these members
2242 differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
2243 C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
2244
2245 =item ev_tstamp interval [read-only]
2246
2247 The specified interval.
2248
2249 =item const char *path [read-only]
2250
2251 The file system path that is being watched.
2252
2253 =back
2254
2255 =head3 Examples
2256
2257 Example: Watch C</etc/passwd> for attribute changes.
2258
2259 static void
2260 passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
2261 {
2262 /* /etc/passwd changed in some way */
2263 if (w->attr.st_nlink)
2264 {
2265 printf ("passwd current size %ld\n", (long)w->attr.st_size);
2266 printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
2267 printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
2268 }
2269 else
2270 /* you shalt not abuse printf for puts */
2271 puts ("wow, /etc/passwd is not there, expect problems. "
2272 "if this is windows, they already arrived\n");
2273 }
2274
2275 ...
2276 ev_stat passwd;
2277
2278 ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
2279 ev_stat_start (loop, &passwd);
2280
2281 Example: Like above, but additionally use a one-second delay so we do not
2282 miss updates (however, frequent updates will delay processing, too, so
2283 one might do the work both on C<ev_stat> callback invocation I<and> on
2284 C<ev_timer> callback invocation).
2285
2286 static ev_stat passwd;
2287 static ev_timer timer;
2288
2289 static void
2290 timer_cb (EV_P_ ev_timer *w, int revents)
2291 {
2292 ev_timer_stop (EV_A_ w);
2293
2294 /* now it's one second after the most recent passwd change */
2295 }
2296
2297 static void
2298 stat_cb (EV_P_ ev_stat *w, int revents)
2299 {
2300 /* reset the one-second timer */
2301 ev_timer_again (EV_A_ &timer);
2302 }
2303
2304 ...
2305 ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
2306 ev_stat_start (loop, &passwd);
2307 ev_timer_init (&timer, timer_cb, 0., 1.02);
2308
2309
2310 =head2 C<ev_idle> - when you've got nothing better to do...
2311
2312 Idle watchers trigger events when no other events of the same or higher
2313 priority are pending (prepare, check and other idle watchers do not count
2314 as receiving "events").
2315
2316 That is, as long as your process is busy handling sockets or timeouts
2317 (or even signals, imagine) of the same or higher priority it will not be
2318 triggered. But when your process is idle (or only lower-priority watchers
2319 are pending), the idle watchers are being called once per event loop
2320 iteration - until stopped, that is, or your process receives more events
2321 and becomes busy again with higher priority stuff.
2322
2323 The most noteworthy effect is that as long as any idle watchers are
2324 active, the process will not block when waiting for new events.
2325
2326 Apart from keeping your process non-blocking (which is a useful
2327 effect on its own sometimes), idle watchers are a good place to do
2328 "pseudo-background processing", or delay processing stuff to after the
2329 event loop has handled all outstanding events.
2330
2331 =head3 Watcher-Specific Functions and Data Members
2332
2333 =over 4
2334
2335 =item ev_idle_init (ev_idle *, callback)
2336
2337 Initialises and configures the idle watcher - it has no parameters of any
2338 kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
2339 believe me.
2340
2341 =back
2342
2343 =head3 Examples
2344
2345 Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
2346 callback, free it. Also, use no error checking, as usual.
2347
2348 static void
2349 idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
2350 {
2351 free (w);
2352 // now do something you wanted to do when the program has
2353 // no longer anything immediate to do.
2354 }
2355
2356 ev_idle *idle_watcher = malloc (sizeof (ev_idle));
2357 ev_idle_init (idle_watcher, idle_cb);
2358 ev_idle_start (loop, idle_cb);
2359
2360
2361 =head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2362
2363 Prepare and check watchers are usually (but not always) used in pairs:
2364 prepare watchers get invoked before the process blocks and check watchers
2365 afterwards.
2366
2367 You I<must not> call C<ev_loop> or similar functions that enter
2368 the current event loop from either C<ev_prepare> or C<ev_check>
2369 watchers. Other loops than the current one are fine, however. The
2370 rationale behind this is that you do not need to check for recursion in
2371 those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
2372 C<ev_check> so if you have one watcher of each kind they will always be
2373 called in pairs bracketing the blocking call.
2374
2375 Their main purpose is to integrate other event mechanisms into libev and
2376 their use is somewhat advanced. They could be used, for example, to track
2377 variable changes, implement your own watchers, integrate net-snmp or a
2378 coroutine library and lots more. They are also occasionally useful if
2379 you cache some data and want to flush it before blocking (for example,
2380 in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
2381 watcher).
2382
2383 This is done by examining in each prepare call which file descriptors
2384 need to be watched by the other library, registering C<ev_io> watchers
2385 for them and starting an C<ev_timer> watcher for any timeouts (many
2386 libraries provide exactly this functionality). Then, in the check watcher,
2387 you check for any events that occurred (by checking the pending status
2388 of all watchers and stopping them) and call back into the library. The
2389 I/O and timer callbacks will never actually be called (but must be valid
2390 nevertheless, because you never know, you know?).
2391
2392 As another example, the Perl Coro module uses these hooks to integrate
2393 coroutines into libev programs, by yielding to other active coroutines
2394 during each prepare and only letting the process block if no coroutines
2395 are ready to run (it's actually more complicated: it only runs coroutines
2396 with priority higher than or equal to the event loop and one coroutine
2397 of lower priority, but only once, using idle watchers to keep the event
2398 loop from blocking if lower-priority coroutines are active, thus mapping
2399 low-priority coroutines to idle/background tasks).
2400
2401 It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
2402 priority, to ensure that they are being run before any other watchers
2403 after the poll (this doesn't matter for C<ev_prepare> watchers).
2404
2405 Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
2406 activate ("feed") events into libev. While libev fully supports this, they
2407 might get executed before other C<ev_check> watchers did their job. As
2408 C<ev_check> watchers are often used to embed other (non-libev) event
2409 loops those other event loops might be in an unusable state until their
2410 C<ev_check> watcher ran (always remind yourself to coexist peacefully with
2411 others).
2412
2413 =head3 Watcher-Specific Functions and Data Members
2414
2415 =over 4
2416
2417 =item ev_prepare_init (ev_prepare *, callback)
2418
2419 =item ev_check_init (ev_check *, callback)
2420
2421 Initialises and configures the prepare or check watcher - they have no
2422 parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
2423 macros, but using them is utterly, utterly, utterly and completely
2424 pointless.
2425
2426 =back
2427
2428 =head3 Examples
2429
2430 There are a number of principal ways to embed other event loops or modules
2431 into libev. Here are some ideas on how to include libadns into libev
2432 (there is a Perl module named C<EV::ADNS> that does this, which you could
2433 use as a working example. Another Perl module named C<EV::Glib> embeds a
2434 Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
2435 Glib event loop).
2436
2437 Method 1: Add IO watchers and a timeout watcher in a prepare handler,
2438 and in a check watcher, destroy them and call into libadns. What follows
2439 is pseudo-code only of course. This requires you to either use a low
2440 priority for the check watcher or use C<ev_clear_pending> explicitly, as
2441 the callbacks for the IO/timeout watchers might not have been called yet.
2442
2443 static ev_io iow [nfd];
2444 static ev_timer tw;
2445
2446 static void
2447 io_cb (struct ev_loop *loop, ev_io *w, int revents)
2448 {
2449 }
2450
2451 // create io watchers for each fd and a timer before blocking
2452 static void
2453 adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
2454 {
2455 int timeout = 3600000;
2456 struct pollfd fds [nfd];
2457 // actual code will need to loop here and realloc etc.
2458 adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2459
2460 /* the callback is illegal, but won't be called as we stop during check */
2461 ev_timer_init (&tw, 0, timeout * 1e-3);
2462 ev_timer_start (loop, &tw);
2463
2464 // create one ev_io per pollfd
2465 for (int i = 0; i < nfd; ++i)
2466 {
2467 ev_io_init (iow + i, io_cb, fds [i].fd,
2468 ((fds [i].events & POLLIN ? EV_READ : 0)
2469 | (fds [i].events & POLLOUT ? EV_WRITE : 0)));
2470
2471 fds [i].revents = 0;
2472 ev_io_start (loop, iow + i);
2473 }
2474 }
2475
2476 // stop all watchers after blocking
2477 static void
2478 adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
2479 {
2480 ev_timer_stop (loop, &tw);
2481
2482 for (int i = 0; i < nfd; ++i)
2483 {
2484 // set the relevant poll flags
2485 // could also call adns_processreadable etc. here
2486 struct pollfd *fd = fds + i;
2487 int revents = ev_clear_pending (iow + i);
2488 if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
2489 if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
2490
2491 // now stop the watcher
2492 ev_io_stop (loop, iow + i);
2493 }
2494
2495 adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
2496 }
2497
2498 Method 2: This would be just like method 1, but you run C<adns_afterpoll>
2499 in the prepare watcher and would dispose of the check watcher.
2500
2501 Method 3: If the module to be embedded supports explicit event
2502 notification (libadns does), you can also make use of the actual watcher
2503 callbacks, and only destroy/create the watchers in the prepare watcher.
2504
2505 static void
2506 timer_cb (EV_P_ ev_timer *w, int revents)
2507 {
2508 adns_state ads = (adns_state)w->data;
2509 update_now (EV_A);
2510
2511 adns_processtimeouts (ads, &tv_now);
2512 }
2513
2514 static void
2515 io_cb (EV_P_ ev_io *w, int revents)
2516 {
2517 adns_state ads = (adns_state)w->data;
2518 update_now (EV_A);
2519
2520 if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
2521 if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
2522 }
2523
2524 // do not ever call adns_afterpoll
2525
2526 Method 4: Do not use a prepare or check watcher because the module you
2527 want to embed is not flexible enough to support it. Instead, you can
2528 override their poll function. The drawback with this solution is that the
2529 main loop is now no longer controllable by EV. The C<Glib::EV> module uses
2530 this approach, effectively embedding EV as a client into the horrible
2531 libglib event loop.
2532
2533 static gint
2534 event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2535 {
2536 int got_events = 0;
2537
2538 for (n = 0; n < nfds; ++n)
2539 // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
2540
2541 if (timeout >= 0)
2542 // create/start timer
2543
2544 // poll
2545 ev_loop (EV_A_ 0);
2546
2547 // stop timer again
2548 if (timeout >= 0)
2549 ev_timer_stop (EV_A_ &to);
2550
2551 // stop io watchers again - their callbacks should have set
2552 for (n = 0; n < nfds; ++n)
2553 ev_io_stop (EV_A_ iow [n]);
2554
2555 return got_events;
2556 }
2557
2558
2559 =head2 C<ev_embed> - when one backend isn't enough...
2560
2561 This is a rather advanced watcher type that lets you embed one event loop
2562 into another (currently only C<ev_io> events are supported in the embedded
2563 loop, other types of watchers might be handled in a delayed or incorrect
2564 fashion and must not be used).
2565
2566 There are primarily two reasons you would want that: work around bugs and
2567 prioritise I/O.
2568
2569 As an example for a bug workaround, the kqueue backend might only support
2570 sockets on some platform, so it is unusable as generic backend, but you
2571 still want to make use of it because you have many sockets and it scales
2572 so nicely. In this case, you would create a kqueue-based loop and embed
2573 it into your default loop (which might use e.g. poll). Overall operation
2574 will be a bit slower because first libev has to call C<poll> and then
2575 C<kevent>, but at least you can use both mechanisms for what they are
2576 best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
2577
2578 As for prioritising I/O: under rare circumstances you have the case where
2579 some fds have to be watched and handled very quickly (with low latency),
2580 and even priorities and idle watchers might have too much overhead. In
2581 this case you would put all the high priority stuff in one loop and all
2582 the rest in a second one, and embed the second one in the first.
2583
2584 As long as the watcher is active, the callback will be invoked every
2585 time there might be events pending in the embedded loop. The callback
2586 must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
2587 sweep and invoke their callbacks (the callback doesn't need to invoke the
2588 C<ev_embed_sweep> function directly, it could also start an idle watcher
2589 to give the embedded loop strictly lower priority for example).
2590
2591 You can also set the callback to C<0>, in which case the embed watcher
2592 will automatically execute the embedded loop sweep whenever necessary.
2593
2594 Fork detection will be handled transparently while the C<ev_embed> watcher
2595 is active, i.e., the embedded loop will automatically be forked when the
2596 embedding loop forks. In other cases, the user is responsible for calling
2597 C<ev_loop_fork> on the embedded loop.
2598
2599 Unfortunately, not all backends are embeddable: only the ones returned by
2600 C<ev_embeddable_backends> are, which, unfortunately, does not include any
2601 portable one.
2602
2603 So when you want to use this feature you will always have to be prepared
2604 that you cannot get an embeddable loop. The recommended way to get around
2605 this is to have a separate variables for your embeddable loop, try to
2606 create it, and if that fails, use the normal loop for everything.
2607
2608 =head3 C<ev_embed> and fork
2609
2610 While the C<ev_embed> watcher is running, forks in the embedding loop will
2611 automatically be applied to the embedded loop as well, so no special
2612 fork handling is required in that case. When the watcher is not running,
2613 however, it is still the task of the libev user to call C<ev_loop_fork ()>
2614 as applicable.
2615
2616 =head3 Watcher-Specific Functions and Data Members
2617
2618 =over 4
2619
2620 =item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
2621
2622 =item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
2623
2624 Configures the watcher to embed the given loop, which must be
2625 embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
2626 invoked automatically, otherwise it is the responsibility of the callback
2627 to invoke it (it will continue to be called until the sweep has been done,
2628 if you do not want that, you need to temporarily stop the embed watcher).
2629
2630 =item ev_embed_sweep (loop, ev_embed *)
2631
2632 Make a single, non-blocking sweep over the embedded loop. This works
2633 similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
2634 appropriate way for embedded loops.
2635
2636 =item struct ev_loop *other [read-only]
2637
2638 The embedded event loop.
2639
2640 =back
2641
2642 =head3 Examples
2643
2644 Example: Try to get an embeddable event loop and embed it into the default
2645 event loop. If that is not possible, use the default loop. The default
2646 loop is stored in C<loop_hi>, while the embeddable loop is stored in
2647 C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
2648 used).
2649
2650 struct ev_loop *loop_hi = ev_default_init (0);
2651 struct ev_loop *loop_lo = 0;
2652 ev_embed embed;
2653
2654 // see if there is a chance of getting one that works
2655 // (remember that a flags value of 0 means autodetection)
2656 loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2657 ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
2658 : 0;
2659
2660 // if we got one, then embed it, otherwise default to loop_hi
2661 if (loop_lo)
2662 {
2663 ev_embed_init (&embed, 0, loop_lo);
2664 ev_embed_start (loop_hi, &embed);
2665 }
2666 else
2667 loop_lo = loop_hi;
2668
2669 Example: Check if kqueue is available but not recommended and create
2670 a kqueue backend for use with sockets (which usually work with any
2671 kqueue implementation). Store the kqueue/socket-only event loop in
2672 C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2673
2674 struct ev_loop *loop = ev_default_init (0);
2675 struct ev_loop *loop_socket = 0;
2676 ev_embed embed;
2677
2678 if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2679 if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2680 {
2681 ev_embed_init (&embed, 0, loop_socket);
2682 ev_embed_start (loop, &embed);
2683 }
2684
2685 if (!loop_socket)
2686 loop_socket = loop;
2687
2688 // now use loop_socket for all sockets, and loop for everything else
2689
2690
2691 =head2 C<ev_fork> - the audacity to resume the event loop after a fork
2692
2693 Fork watchers are called when a C<fork ()> was detected (usually because
2694 whoever is a good citizen cared to tell libev about it by calling
2695 C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
2696 event loop blocks next and before C<ev_check> watchers are being called,
2697 and only in the child after the fork. If whoever good citizen calling
2698 C<ev_default_fork> cheats and calls it in the wrong process, the fork
2699 handlers will be invoked, too, of course.
2700
2701 =head3 Watcher-Specific Functions and Data Members
2702
2703 =over 4
2704
2705 =item ev_fork_init (ev_signal *, callback)
2706
2707 Initialises and configures the fork watcher - it has no parameters of any
2708 kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2709 believe me.
2710
2711 =back
2712
2713
2714 =head2 C<ev_async> - how to wake up another event loop
2715
2716 In general, you cannot use an C<ev_loop> from multiple threads or other
2717 asynchronous sources such as signal handlers (as opposed to multiple event
2718 loops - those are of course safe to use in different threads).
2719
2720 Sometimes, however, you need to wake up another event loop you do not
2721 control, for example because it belongs to another thread. This is what
2722 C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
2723 can signal it by calling C<ev_async_send>, which is thread- and signal
2724 safe.
2725
2726 This functionality is very similar to C<ev_signal> watchers, as signals,
2727 too, are asynchronous in nature, and signals, too, will be compressed
2728 (i.e. the number of callback invocations may be less than the number of
2729 C<ev_async_sent> calls).
2730
2731 Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2732 just the default loop.
2733
2734 =head3 Queueing
2735
2736 C<ev_async> does not support queueing of data in any way. The reason
2737 is that the author does not know of a simple (or any) algorithm for a
2738 multiple-writer-single-reader queue that works in all cases and doesn't
2739 need elaborate support such as pthreads.
2740
2741 That means that if you want to queue data, you have to provide your own
2742 queue. But at least I can tell you how to implement locking around your
2743 queue:
2744
2745 =over 4
2746
2747 =item queueing from a signal handler context
2748
2749 To implement race-free queueing, you simply add to the queue in the signal
2750 handler but you block the signal handler in the watcher callback. Here is
2751 an example that does that for some fictitious SIGUSR1 handler:
2752
2753 static ev_async mysig;
2754
2755 static void
2756 sigusr1_handler (void)
2757 {
2758 sometype data;
2759
2760 // no locking etc.
2761 queue_put (data);
2762 ev_async_send (EV_DEFAULT_ &mysig);
2763 }
2764
2765 static void
2766 mysig_cb (EV_P_ ev_async *w, int revents)
2767 {
2768 sometype data;
2769 sigset_t block, prev;
2770
2771 sigemptyset (&block);
2772 sigaddset (&block, SIGUSR1);
2773 sigprocmask (SIG_BLOCK, &block, &prev);
2774
2775 while (queue_get (&data))
2776 process (data);
2777
2778 if (sigismember (&prev, SIGUSR1)
2779 sigprocmask (SIG_UNBLOCK, &block, 0);
2780 }
2781
2782 (Note: pthreads in theory requires you to use C<pthread_setmask>
2783 instead of C<sigprocmask> when you use threads, but libev doesn't do it
2784 either...).
2785
2786 =item queueing from a thread context
2787
2788 The strategy for threads is different, as you cannot (easily) block
2789 threads but you can easily preempt them, so to queue safely you need to
2790 employ a traditional mutex lock, such as in this pthread example:
2791
2792 static ev_async mysig;
2793 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2794
2795 static void
2796 otherthread (void)
2797 {
2798 // only need to lock the actual queueing operation
2799 pthread_mutex_lock (&mymutex);
2800 queue_put (data);
2801 pthread_mutex_unlock (&mymutex);
2802
2803 ev_async_send (EV_DEFAULT_ &mysig);
2804 }
2805
2806 static void
2807 mysig_cb (EV_P_ ev_async *w, int revents)
2808 {
2809 pthread_mutex_lock (&mymutex);
2810
2811 while (queue_get (&data))
2812 process (data);
2813
2814 pthread_mutex_unlock (&mymutex);
2815 }
2816
2817 =back
2818
2819
2820 =head3 Watcher-Specific Functions and Data Members
2821
2822 =over 4
2823
2824 =item ev_async_init (ev_async *, callback)
2825
2826 Initialises and configures the async watcher - it has no parameters of any
2827 kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
2828 trust me.
2829
2830 =item ev_async_send (loop, ev_async *)
2831
2832 Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2833 an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2834 C<ev_feed_event>, this call is safe to do from other threads, signal or
2835 similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
2836 section below on what exactly this means).
2837
2838 Note that, as with other watchers in libev, multiple events might get
2839 compressed into a single callback invocation (another way to look at this
2840 is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
2841 reset when the event loop detects that).
2842
2843 This call incurs the overhead of a system call only once per event loop
2844 iteration, so while the overhead might be noticeable, it doesn't apply to
2845 repeated calls to C<ev_async_send> for the same event loop.
2846
2847 =item bool = ev_async_pending (ev_async *)
2848
2849 Returns a non-zero value when C<ev_async_send> has been called on the
2850 watcher but the event has not yet been processed (or even noted) by the
2851 event loop.
2852
2853 C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2854 the loop iterates next and checks for the watcher to have become active,
2855 it will reset the flag again. C<ev_async_pending> can be used to very
2856 quickly check whether invoking the loop might be a good idea.
2857
2858 Not that this does I<not> check whether the watcher itself is pending,
2859 only whether it has been requested to make this watcher pending: there
2860 is a time window between the event loop checking and resetting the async
2861 notification, and the callback being invoked.
2862
2863 =back
2864
2865
2866 =head1 OTHER FUNCTIONS
2867
2868 There are some other functions of possible interest. Described. Here. Now.
2869
2870 =over 4
2871
2872 =item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2873
2874 This function combines a simple timer and an I/O watcher, calls your
2875 callback on whichever event happens first and automatically stops both
2876 watchers. This is useful if you want to wait for a single event on an fd
2877 or timeout without having to allocate/configure/start/stop/free one or
2878 more watchers yourself.
2879
2880 If C<fd> is less than 0, then no I/O watcher will be started and the
2881 C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
2882 the given C<fd> and C<events> set will be created and started.
2883
2884 If C<timeout> is less than 0, then no timeout watcher will be
2885 started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2886 repeat = 0) will be started. C<0> is a valid timeout.
2887
2888 The callback has the type C<void (*cb)(int revents, void *arg)> and gets
2889 passed an C<revents> set like normal event callbacks (a combination of
2890 C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2891 value passed to C<ev_once>. Note that it is possible to receive I<both>
2892 a timeout and an io event at the same time - you probably should give io
2893 events precedence.
2894
2895 Example: wait up to ten seconds for data to appear on STDIN_FILENO.
2896
2897 static void stdin_ready (int revents, void *arg)
2898 {
2899 if (revents & EV_READ)
2900 /* stdin might have data for us, joy! */;
2901 else if (revents & EV_TIMEOUT)
2902 /* doh, nothing entered */;
2903 }
2904
2905 ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2906
2907 =item ev_feed_event (struct ev_loop *, watcher *, int revents)
2908
2909 Feeds the given event set into the event loop, as if the specified event
2910 had happened for the specified watcher (which must be a pointer to an
2911 initialised but not necessarily started event watcher).
2912
2913 =item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
2914
2915 Feed an event on the given fd, as if a file descriptor backend detected
2916 the given events it.
2917
2918 =item ev_feed_signal_event (struct ev_loop *loop, int signum)
2919
2920 Feed an event as if the given signal occurred (C<loop> must be the default
2921 loop!).
2922
2923 =back
2924
2925
2926 =head1 LIBEVENT EMULATION
2927
2928 Libev offers a compatibility emulation layer for libevent. It cannot
2929 emulate the internals of libevent, so here are some usage hints:
2930
2931 =over 4
2932
2933 =item * Use it by including <event.h>, as usual.
2934
2935 =item * The following members are fully supported: ev_base, ev_callback,
2936 ev_arg, ev_fd, ev_res, ev_events.
2937
2938 =item * Avoid using ev_flags and the EVLIST_*-macros, while it is
2939 maintained by libev, it does not work exactly the same way as in libevent (consider
2940 it a private API).
2941
2942 =item * Priorities are not currently supported. Initialising priorities
2943 will fail and all watchers will have the same priority, even though there
2944 is an ev_pri field.
2945
2946 =item * In libevent, the last base created gets the signals, in libev, the
2947 first base created (== the default loop) gets the signals.
2948
2949 =item * Other members are not supported.
2950
2951 =item * The libev emulation is I<not> ABI compatible to libevent, you need
2952 to use the libev header file and library.
2953
2954 =back
2955
2956 =head1 C++ SUPPORT
2957
2958 Libev comes with some simplistic wrapper classes for C++ that mainly allow
2959 you to use some convenience methods to start/stop watchers and also change
2960 the callback model to a model using method callbacks on objects.
2961
2962 To use it,
2963
2964 #include <ev++.h>
2965
2966 This automatically includes F<ev.h> and puts all of its definitions (many
2967 of them macros) into the global namespace. All C++ specific things are
2968 put into the C<ev> namespace. It should support all the same embedding
2969 options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
2970
2971 Care has been taken to keep the overhead low. The only data member the C++
2972 classes add (compared to plain C-style watchers) is the event loop pointer
2973 that the watcher is associated with (or no additional members at all if
2974 you disable C<EV_MULTIPLICITY> when embedding libev).
2975
2976 Currently, functions, and static and non-static member functions can be
2977 used as callbacks. Other types should be easy to add as long as they only
2978 need one additional pointer for context. If you need support for other
2979 types of functors please contact the author (preferably after implementing
2980 it).
2981
2982 Here is a list of things available in the C<ev> namespace:
2983
2984 =over 4
2985
2986 =item C<ev::READ>, C<ev::WRITE> etc.
2987
2988 These are just enum values with the same values as the C<EV_READ> etc.
2989 macros from F<ev.h>.
2990
2991 =item C<ev::tstamp>, C<ev::now>
2992
2993 Aliases to the same types/functions as with the C<ev_> prefix.
2994
2995 =item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
2996
2997 For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
2998 the same name in the C<ev> namespace, with the exception of C<ev_signal>
2999 which is called C<ev::sig> to avoid clashes with the C<signal> macro
3000 defines by many implementations.
3001
3002 All of those classes have these methods:
3003
3004 =over 4
3005
3006 =item ev::TYPE::TYPE ()
3007
3008 =item ev::TYPE::TYPE (struct ev_loop *)
3009
3010 =item ev::TYPE::~TYPE
3011
3012 The constructor (optionally) takes an event loop to associate the watcher
3013 with. If it is omitted, it will use C<EV_DEFAULT>.
3014
3015 The constructor calls C<ev_init> for you, which means you have to call the
3016 C<set> method before starting it.
3017
3018 It will not set a callback, however: You have to call the templated C<set>
3019 method to set a callback before you can start the watcher.
3020
3021 (The reason why you have to use a method is a limitation in C++ which does
3022 not allow explicit template arguments for constructors).
3023
3024 The destructor automatically stops the watcher if it is active.
3025
3026 =item w->set<class, &class::method> (object *)
3027
3028 This method sets the callback method to call. The method has to have a
3029 signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
3030 first argument and the C<revents> as second. The object must be given as
3031 parameter and is stored in the C<data> member of the watcher.
3032
3033 This method synthesizes efficient thunking code to call your method from
3034 the C callback that libev requires. If your compiler can inline your
3035 callback (i.e. it is visible to it at the place of the C<set> call and
3036 your compiler is good :), then the method will be fully inlined into the
3037 thunking function, making it as fast as a direct C callback.
3038
3039 Example: simple class declaration and watcher initialisation
3040
3041 struct myclass
3042 {
3043 void io_cb (ev::io &w, int revents) { }
3044 }
3045
3046 myclass obj;
3047 ev::io iow;
3048 iow.set <myclass, &myclass::io_cb> (&obj);
3049
3050 =item w->set (object *)
3051
3052 This is an B<experimental> feature that might go away in a future version.
3053
3054 This is a variation of a method callback - leaving out the method to call
3055 will default the method to C<operator ()>, which makes it possible to use
3056 functor objects without having to manually specify the C<operator ()> all
3057 the time. Incidentally, you can then also leave out the template argument
3058 list.
3059
3060 The C<operator ()> method prototype must be C<void operator ()(watcher &w,
3061 int revents)>.
3062
3063 See the method-C<set> above for more details.
3064
3065 Example: use a functor object as callback.
3066
3067 struct myfunctor
3068 {
3069 void operator() (ev::io &w, int revents)
3070 {
3071 ...
3072 }
3073 }
3074
3075 myfunctor f;
3076
3077 ev::io w;
3078 w.set (&f);
3079
3080 =item w->set<function> (void *data = 0)
3081
3082 Also sets a callback, but uses a static method or plain function as
3083 callback. The optional C<data> argument will be stored in the watcher's
3084 C<data> member and is free for you to use.
3085
3086 The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
3087
3088 See the method-C<set> above for more details.
3089
3090 Example: Use a plain function as callback.
3091
3092 static void io_cb (ev::io &w, int revents) { }
3093 iow.set <io_cb> ();
3094
3095 =item w->set (struct ev_loop *)
3096
3097 Associates a different C<struct ev_loop> with this watcher. You can only
3098 do this when the watcher is inactive (and not pending either).
3099
3100 =item w->set ([arguments])
3101
3102 Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
3103 called at least once. Unlike the C counterpart, an active watcher gets
3104 automatically stopped and restarted when reconfiguring it with this
3105 method.
3106
3107 =item w->start ()
3108
3109 Starts the watcher. Note that there is no C<loop> argument, as the
3110 constructor already stores the event loop.
3111
3112 =item w->stop ()
3113
3114 Stops the watcher if it is active. Again, no C<loop> argument.
3115
3116 =item w->again () (C<ev::timer>, C<ev::periodic> only)
3117
3118 For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
3119 C<ev_TYPE_again> function.
3120
3121 =item w->sweep () (C<ev::embed> only)
3122
3123 Invokes C<ev_embed_sweep>.
3124
3125 =item w->update () (C<ev::stat> only)
3126
3127 Invokes C<ev_stat_stat>.
3128
3129 =back
3130
3131 =back
3132
3133 Example: Define a class with an IO and idle watcher, start one of them in
3134 the constructor.
3135
3136 class myclass
3137 {
3138 ev::io io ; void io_cb (ev::io &w, int revents);
3139 ev::idle idle; void idle_cb (ev::idle &w, int revents);
3140
3141 myclass (int fd)
3142 {
3143 io .set <myclass, &myclass::io_cb > (this);
3144 idle.set <myclass, &myclass::idle_cb> (this);
3145
3146 io.start (fd, ev::READ);
3147 }
3148 };
3149
3150
3151 =head1 OTHER LANGUAGE BINDINGS
3152
3153 Libev does not offer other language bindings itself, but bindings for a
3154 number of languages exist in the form of third-party packages. If you know
3155 any interesting language binding in addition to the ones listed here, drop
3156 me a note.
3157
3158 =over 4
3159
3160 =item Perl
3161
3162 The EV module implements the full libev API and is actually used to test
3163 libev. EV is developed together with libev. Apart from the EV core module,
3164 there are additional modules that implement libev-compatible interfaces
3165 to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
3166 C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
3167 and C<EV::Glib>).
3168
3169 It can be found and installed via CPAN, its homepage is at
3170 L<http://software.schmorp.de/pkg/EV>.
3171
3172 =item Python
3173
3174 Python bindings can be found at L<http://code.google.com/p/pyev/>. It
3175 seems to be quite complete and well-documented.
3176
3177 =item Ruby
3178
3179 Tony Arcieri has written a ruby extension that offers access to a subset
3180 of the libev API and adds file handle abstractions, asynchronous DNS and
3181 more on top of it. It can be found via gem servers. Its homepage is at
3182 L<http://rev.rubyforge.org/>.
3183
3184 Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
3185 makes rev work even on mingw.
3186
3187 =item Haskell
3188
3189 A haskell binding to libev is available at
3190 L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
3191
3192 =item D
3193
3194 Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3195 be found at L<http://proj.llucax.com.ar/wiki/evd>.
3196
3197 =item Ocaml
3198
3199 Erkki Seppala has written Ocaml bindings for libev, to be found at
3200 L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
3201
3202 =back
3203
3204
3205 =head1 MACRO MAGIC
3206
3207 Libev can be compiled with a variety of options, the most fundamental
3208 of which is C<EV_MULTIPLICITY>. This option determines whether (most)
3209 functions and callbacks have an initial C<struct ev_loop *> argument.
3210
3211 To make it easier to write programs that cope with either variant, the
3212 following macros are defined:
3213
3214 =over 4
3215
3216 =item C<EV_A>, C<EV_A_>
3217
3218 This provides the loop I<argument> for functions, if one is required ("ev
3219 loop argument"). The C<EV_A> form is used when this is the sole argument,
3220 C<EV_A_> is used when other arguments are following. Example:
3221
3222 ev_unref (EV_A);
3223 ev_timer_add (EV_A_ watcher);
3224 ev_loop (EV_A_ 0);
3225
3226 It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
3227 which is often provided by the following macro.
3228
3229 =item C<EV_P>, C<EV_P_>
3230
3231 This provides the loop I<parameter> for functions, if one is required ("ev
3232 loop parameter"). The C<EV_P> form is used when this is the sole parameter,
3233 C<EV_P_> is used when other parameters are following. Example:
3234
3235 // this is how ev_unref is being declared
3236 static void ev_unref (EV_P);
3237
3238 // this is how you can declare your typical callback
3239 static void cb (EV_P_ ev_timer *w, int revents)
3240
3241 It declares a parameter C<loop> of type C<struct ev_loop *>, quite
3242 suitable for use with C<EV_A>.
3243
3244 =item C<EV_DEFAULT>, C<EV_DEFAULT_>
3245
3246 Similar to the other two macros, this gives you the value of the default
3247 loop, if multiple loops are supported ("ev loop default").
3248
3249 =item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3250
3251 Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3252 default loop has been initialised (C<UC> == unchecked). Their behaviour
3253 is undefined when the default loop has not been initialised by a previous
3254 execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
3255
3256 It is often prudent to use C<EV_DEFAULT> when initialising the first
3257 watcher in a function but use C<EV_DEFAULT_UC> afterwards.
3258
3259 =back
3260
3261 Example: Declare and initialise a check watcher, utilising the above
3262 macros so it will work regardless of whether multiple loops are supported
3263 or not.
3264
3265 static void
3266 check_cb (EV_P_ ev_timer *w, int revents)
3267 {
3268 ev_check_stop (EV_A_ w);
3269 }
3270
3271 ev_check check;
3272 ev_check_init (&check, check_cb);
3273 ev_check_start (EV_DEFAULT_ &check);
3274 ev_loop (EV_DEFAULT_ 0);
3275
3276 =head1 EMBEDDING
3277
3278 Libev can (and often is) directly embedded into host
3279 applications. Examples of applications that embed it include the Deliantra
3280 Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
3281 and rxvt-unicode.
3282
3283 The goal is to enable you to just copy the necessary files into your
3284 source directory without having to change even a single line in them, so
3285 you can easily upgrade by simply copying (or having a checked-out copy of
3286 libev somewhere in your source tree).
3287
3288 =head2 FILESETS
3289
3290 Depending on what features you need you need to include one or more sets of files
3291 in your application.
3292
3293 =head3 CORE EVENT LOOP
3294
3295 To include only the libev core (all the C<ev_*> functions), with manual
3296 configuration (no autoconf):
3297
3298 #define EV_STANDALONE 1
3299 #include "ev.c"
3300
3301 This will automatically include F<ev.h>, too, and should be done in a
3302 single C source file only to provide the function implementations. To use
3303 it, do the same for F<ev.h> in all files wishing to use this API (best
3304 done by writing a wrapper around F<ev.h> that you can include instead and
3305 where you can put other configuration options):
3306
3307 #define EV_STANDALONE 1
3308 #include "ev.h"
3309
3310 Both header files and implementation files can be compiled with a C++
3311 compiler (at least, that's a stated goal, and breakage will be treated
3312 as a bug).
3313
3314 You need the following files in your source tree, or in a directory
3315 in your include path (e.g. in libev/ when using -Ilibev):
3316
3317 ev.h
3318 ev.c
3319 ev_vars.h
3320 ev_wrap.h
3321
3322 ev_win32.c required on win32 platforms only
3323
3324 ev_select.c only when select backend is enabled (which is enabled by default)
3325 ev_poll.c only when poll backend is enabled (disabled by default)
3326 ev_epoll.c only when the epoll backend is enabled (disabled by default)
3327 ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
3328 ev_port.c only when the solaris port backend is enabled (disabled by default)
3329
3330 F<ev.c> includes the backend files directly when enabled, so you only need
3331 to compile this single file.
3332
3333 =head3 LIBEVENT COMPATIBILITY API
3334
3335 To include the libevent compatibility API, also include:
3336
3337 #include "event.c"
3338
3339 in the file including F<ev.c>, and:
3340
3341 #include "event.h"
3342
3343 in the files that want to use the libevent API. This also includes F<ev.h>.
3344
3345 You need the following additional files for this:
3346
3347 event.h
3348 event.c
3349
3350 =head3 AUTOCONF SUPPORT
3351
3352 Instead of using C<EV_STANDALONE=1> and providing your configuration in
3353 whatever way you want, you can also C<m4_include([libev.m4])> in your
3354 F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
3355 include F<config.h> and configure itself accordingly.
3356
3357 For this of course you need the m4 file:
3358
3359 libev.m4
3360
3361 =head2 PREPROCESSOR SYMBOLS/MACROS
3362
3363 Libev can be configured via a variety of preprocessor symbols you have to
3364 define before including any of its files. The default in the absence of
3365 autoconf is documented for every option.
3366
3367 =over 4
3368
3369 =item EV_STANDALONE
3370
3371 Must always be C<1> if you do not use autoconf configuration, which
3372 keeps libev from including F<config.h>, and it also defines dummy
3373 implementations for some libevent functions (such as logging, which is not
3374 supported). It will also not define any of the structs usually found in
3375 F<event.h> that are not directly supported by the libev core alone.
3376
3377 In stanbdalone mode, libev will still try to automatically deduce the
3378 configuration, but has to be more conservative.
3379
3380 =item EV_USE_MONOTONIC
3381
3382 If defined to be C<1>, libev will try to detect the availability of the
3383 monotonic clock option at both compile time and runtime. Otherwise no
3384 use of the monotonic clock option will be attempted. If you enable this,
3385 you usually have to link against librt or something similar. Enabling it
3386 when the functionality isn't available is safe, though, although you have
3387 to make sure you link against any libraries where the C<clock_gettime>
3388 function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
3389
3390 =item EV_USE_REALTIME
3391
3392 If defined to be C<1>, libev will try to detect the availability of the
3393 real-time clock option at compile time (and assume its availability
3394 at runtime if successful). Otherwise no use of the real-time clock
3395 option will be attempted. This effectively replaces C<gettimeofday>
3396 by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
3397 correctness. See the note about libraries in the description of
3398 C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
3399 C<EV_USE_CLOCK_SYSCALL>.
3400
3401 =item EV_USE_CLOCK_SYSCALL
3402
3403 If defined to be C<1>, libev will try to use a direct syscall instead
3404 of calling the system-provided C<clock_gettime> function. This option
3405 exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
3406 unconditionally pulls in C<libpthread>, slowing down single-threaded
3407 programs needlessly. Using a direct syscall is slightly slower (in
3408 theory), because no optimised vdso implementation can be used, but avoids
3409 the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
3410 higher, as it simplifies linking (no need for C<-lrt>).
3411
3412 =item EV_USE_NANOSLEEP
3413
3414 If defined to be C<1>, libev will assume that C<nanosleep ()> is available
3415 and will use it for delays. Otherwise it will use C<select ()>.
3416
3417 =item EV_USE_EVENTFD
3418
3419 If defined to be C<1>, then libev will assume that C<eventfd ()> is
3420 available and will probe for kernel support at runtime. This will improve
3421 C<ev_signal> and C<ev_async> performance and reduce resource consumption.
3422 If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
3423 2.7 or newer, otherwise disabled.
3424
3425 =item EV_USE_SELECT
3426
3427 If undefined or defined to be C<1>, libev will compile in support for the
3428 C<select>(2) backend. No attempt at auto-detection will be done: if no
3429 other method takes over, select will be it. Otherwise the select backend
3430 will not be compiled in.
3431
3432 =item EV_SELECT_USE_FD_SET
3433
3434 If defined to C<1>, then the select backend will use the system C<fd_set>
3435 structure. This is useful if libev doesn't compile due to a missing
3436 C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
3437 on exotic systems. This usually limits the range of file descriptors to
3438 some low limit such as 1024 or might have other limitations (winsocket
3439 only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
3440 configures the maximum size of the C<fd_set>.
3441
3442 =item EV_SELECT_IS_WINSOCKET
3443
3444 When defined to C<1>, the select backend will assume that
3445 select/socket/connect etc. don't understand file descriptors but
3446 wants osf handles on win32 (this is the case when the select to
3447 be used is the winsock select). This means that it will call
3448 C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
3449 it is assumed that all these functions actually work on fds, even
3450 on win32. Should not be defined on non-win32 platforms.
3451
3452 =item EV_FD_TO_WIN32_HANDLE
3453
3454 If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
3455 file descriptors to socket handles. When not defining this symbol (the
3456 default), then libev will call C<_get_osfhandle>, which is usually
3457 correct. In some cases, programs use their own file descriptor management,
3458 in which case they can provide this function to map fds to socket handles.
3459
3460 =item EV_USE_POLL
3461
3462 If defined to be C<1>, libev will compile in support for the C<poll>(2)
3463 backend. Otherwise it will be enabled on non-win32 platforms. It
3464 takes precedence over select.
3465
3466 =item EV_USE_EPOLL
3467
3468 If defined to be C<1>, libev will compile in support for the Linux
3469 C<epoll>(7) backend. Its availability will be detected at runtime,
3470 otherwise another method will be used as fallback. This is the preferred
3471 backend for GNU/Linux systems. If undefined, it will be enabled if the
3472 headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3473
3474 =item EV_USE_KQUEUE
3475
3476 If defined to be C<1>, libev will compile in support for the BSD style
3477 C<kqueue>(2) backend. Its actual availability will be detected at runtime,
3478 otherwise another method will be used as fallback. This is the preferred
3479 backend for BSD and BSD-like systems, although on most BSDs kqueue only
3480 supports some types of fds correctly (the only platform we found that
3481 supports ptys for example was NetBSD), so kqueue might be compiled in, but
3482 not be used unless explicitly requested. The best way to use it is to find
3483 out whether kqueue supports your type of fd properly and use an embedded
3484 kqueue loop.
3485
3486 =item EV_USE_PORT
3487
3488 If defined to be C<1>, libev will compile in support for the Solaris
3489 10 port style backend. Its availability will be detected at runtime,
3490 otherwise another method will be used as fallback. This is the preferred
3491 backend for Solaris 10 systems.
3492
3493 =item EV_USE_DEVPOLL
3494
3495 Reserved for future expansion, works like the USE symbols above.
3496
3497 =item EV_USE_INOTIFY
3498
3499 If defined to be C<1>, libev will compile in support for the Linux inotify
3500 interface to speed up C<ev_stat> watchers. Its actual availability will
3501 be detected at runtime. If undefined, it will be enabled if the headers
3502 indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3503
3504 =item EV_ATOMIC_T
3505
3506 Libev requires an integer type (suitable for storing C<0> or C<1>) whose
3507 access is atomic with respect to other threads or signal contexts. No such
3508 type is easily found in the C language, so you can provide your own type
3509 that you know is safe for your purposes. It is used both for signal handler "locking"
3510 as well as for signal and thread safety in C<ev_async> watchers.
3511
3512 In the absence of this define, libev will use C<sig_atomic_t volatile>
3513 (from F<signal.h>), which is usually good enough on most platforms.
3514
3515 =item EV_H
3516
3517 The name of the F<ev.h> header file used to include it. The default if
3518 undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
3519 used to virtually rename the F<ev.h> header file in case of conflicts.
3520
3521 =item EV_CONFIG_H
3522
3523 If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
3524 F<ev.c>'s idea of where to find the F<config.h> file, similarly to
3525 C<EV_H>, above.
3526
3527 =item EV_EVENT_H
3528
3529 Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
3530 of how the F<event.h> header can be found, the default is C<"event.h">.
3531
3532 =item EV_PROTOTYPES
3533
3534 If defined to be C<0>, then F<ev.h> will not define any function
3535 prototypes, but still define all the structs and other symbols. This is
3536 occasionally useful if you want to provide your own wrapper functions
3537 around libev functions.
3538
3539 =item EV_MULTIPLICITY
3540
3541 If undefined or defined to C<1>, then all event-loop-specific functions
3542 will have the C<struct ev_loop *> as first argument, and you can create
3543 additional independent event loops. Otherwise there will be no support
3544 for multiple event loops and there is no first event loop pointer
3545 argument. Instead, all functions act on the single default loop.
3546
3547 =item EV_MINPRI
3548
3549 =item EV_MAXPRI
3550
3551 The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
3552 C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
3553 provide for more priorities by overriding those symbols (usually defined
3554 to be C<-2> and C<2>, respectively).
3555
3556 When doing priority-based operations, libev usually has to linearly search
3557 all the priorities, so having many of them (hundreds) uses a lot of space
3558 and time, so using the defaults of five priorities (-2 .. +2) is usually
3559 fine.
3560
3561 If your embedding application does not need any priorities, defining these
3562 both to C<0> will save some memory and CPU.
3563
3564 =item EV_PERIODIC_ENABLE
3565
3566 If undefined or defined to be C<1>, then periodic timers are supported. If
3567 defined to be C<0>, then they are not. Disabling them saves a few kB of
3568 code.
3569
3570 =item EV_IDLE_ENABLE
3571
3572 If undefined or defined to be C<1>, then idle watchers are supported. If
3573 defined to be C<0>, then they are not. Disabling them saves a few kB of
3574 code.
3575
3576 =item EV_EMBED_ENABLE
3577
3578 If undefined or defined to be C<1>, then embed watchers are supported. If
3579 defined to be C<0>, then they are not. Embed watchers rely on most other
3580 watcher types, which therefore must not be disabled.
3581
3582 =item EV_STAT_ENABLE
3583
3584 If undefined or defined to be C<1>, then stat watchers are supported. If
3585 defined to be C<0>, then they are not.
3586
3587 =item EV_FORK_ENABLE
3588
3589 If undefined or defined to be C<1>, then fork watchers are supported. If
3590 defined to be C<0>, then they are not.
3591
3592 =item EV_ASYNC_ENABLE
3593
3594 If undefined or defined to be C<1>, then async watchers are supported. If
3595 defined to be C<0>, then they are not.
3596
3597 =item EV_MINIMAL
3598
3599 If you need to shave off some kilobytes of code at the expense of some
3600 speed, define this symbol to C<1>. Currently this is used to override some
3601 inlining decisions, saves roughly 30% code size on amd64. It also selects a
3602 much smaller 2-heap for timer management over the default 4-heap.
3603
3604 =item EV_PID_HASHSIZE
3605
3606 C<ev_child> watchers use a small hash table to distribute workload by
3607 pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
3608 than enough. If you need to manage thousands of children you might want to
3609 increase this value (I<must> be a power of two).
3610
3611 =item EV_INOTIFY_HASHSIZE
3612
3613 C<ev_stat> watchers use a small hash table to distribute workload by
3614 inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
3615 usually more than enough. If you need to manage thousands of C<ev_stat>
3616 watchers you might want to increase this value (I<must> be a power of
3617 two).
3618
3619 =item EV_USE_4HEAP
3620
3621 Heaps are not very cache-efficient. To improve the cache-efficiency of the
3622 timer and periodics heaps, libev uses a 4-heap when this symbol is defined
3623 to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
3624 faster performance with many (thousands) of watchers.
3625
3626 The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3627 (disabled).
3628
3629 =item EV_HEAP_CACHE_AT
3630
3631 Heaps are not very cache-efficient. To improve the cache-efficiency of the
3632 timer and periodics heaps, libev can cache the timestamp (I<at>) within
3633 the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
3634 which uses 8-12 bytes more per watcher and a few hundred bytes more code,
3635 but avoids random read accesses on heap changes. This improves performance
3636 noticeably with many (hundreds) of watchers.
3637
3638 The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3639 (disabled).
3640
3641 =item EV_VERIFY
3642
3643 Controls how much internal verification (see C<ev_loop_verify ()>) will
3644 be done: If set to C<0>, no internal verification code will be compiled
3645 in. If set to C<1>, then verification code will be compiled in, but not
3646 called. If set to C<2>, then the internal verification code will be
3647 called once per loop, which can slow down libev. If set to C<3>, then the
3648 verification code will be called very frequently, which will slow down
3649 libev considerably.
3650
3651 The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
3652 C<0>.
3653
3654 =item EV_COMMON
3655
3656 By default, all watchers have a C<void *data> member. By redefining
3657 this macro to a something else you can include more and other types of
3658 members. You have to define it each time you include one of the files,
3659 though, and it must be identical each time.
3660
3661 For example, the perl EV module uses something like this:
3662
3663 #define EV_COMMON \
3664 SV *self; /* contains this struct */ \
3665 SV *cb_sv, *fh /* note no trailing ";" */
3666
3667 =item EV_CB_DECLARE (type)
3668
3669 =item EV_CB_INVOKE (watcher, revents)
3670
3671 =item ev_set_cb (ev, cb)
3672
3673 Can be used to change the callback member declaration in each watcher,
3674 and the way callbacks are invoked and set. Must expand to a struct member
3675 definition and a statement, respectively. See the F<ev.h> header file for
3676 their default definitions. One possible use for overriding these is to
3677 avoid the C<struct ev_loop *> as first argument in all cases, or to use
3678 method calls instead of plain function calls in C++.
3679
3680 =back
3681
3682 =head2 EXPORTED API SYMBOLS
3683
3684 If you need to re-export the API (e.g. via a DLL) and you need a list of
3685 exported symbols, you can use the provided F<Symbol.*> files which list
3686 all public symbols, one per line:
3687
3688 Symbols.ev for libev proper
3689 Symbols.event for the libevent emulation
3690
3691 This can also be used to rename all public symbols to avoid clashes with
3692 multiple versions of libev linked together (which is obviously bad in
3693 itself, but sometimes it is inconvenient to avoid this).
3694
3695 A sed command like this will create wrapper C<#define>'s that you need to
3696 include before including F<ev.h>:
3697
3698 <Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
3699
3700 This would create a file F<wrap.h> which essentially looks like this:
3701
3702 #define ev_backend myprefix_ev_backend
3703 #define ev_check_start myprefix_ev_check_start
3704 #define ev_check_stop myprefix_ev_check_stop
3705 ...
3706
3707 =head2 EXAMPLES
3708
3709 For a real-world example of a program the includes libev
3710 verbatim, you can have a look at the EV perl module
3711 (L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
3712 the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
3713 interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
3714 will be compiled. It is pretty complex because it provides its own header
3715 file.
3716
3717 The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
3718 that everybody includes and which overrides some configure choices:
3719
3720 #define EV_MINIMAL 1
3721 #define EV_USE_POLL 0
3722 #define EV_MULTIPLICITY 0
3723 #define EV_PERIODIC_ENABLE 0
3724 #define EV_STAT_ENABLE 0
3725 #define EV_FORK_ENABLE 0
3726 #define EV_CONFIG_H <config.h>
3727 #define EV_MINPRI 0
3728 #define EV_MAXPRI 0
3729
3730 #include "ev++.h"
3731
3732 And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3733
3734 #include "ev_cpp.h"
3735 #include "ev.c"
3736
3737 =head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
3738
3739 =head2 THREADS AND COROUTINES
3740
3741 =head3 THREADS
3742
3743 All libev functions are reentrant and thread-safe unless explicitly
3744 documented otherwise, but libev implements no locking itself. This means
3745 that you can use as many loops as you want in parallel, as long as there
3746 are no concurrent calls into any libev function with the same loop
3747 parameter (C<ev_default_*> calls have an implicit default loop parameter,
3748 of course): libev guarantees that different event loops share no data
3749 structures that need any locking.
3750
3751 Or to put it differently: calls with different loop parameters can be done
3752 concurrently from multiple threads, calls with the same loop parameter
3753 must be done serially (but can be done from different threads, as long as
3754 only one thread ever is inside a call at any point in time, e.g. by using
3755 a mutex per loop).
3756
3757 Specifically to support threads (and signal handlers), libev implements
3758 so-called C<ev_async> watchers, which allow some limited form of
3759 concurrency on the same event loop, namely waking it up "from the
3760 outside".
3761
3762 If you want to know which design (one loop, locking, or multiple loops
3763 without or something else still) is best for your problem, then I cannot
3764 help you, but here is some generic advice:
3765
3766 =over 4
3767
3768 =item * most applications have a main thread: use the default libev loop
3769 in that thread, or create a separate thread running only the default loop.
3770
3771 This helps integrating other libraries or software modules that use libev
3772 themselves and don't care/know about threading.
3773
3774 =item * one loop per thread is usually a good model.
3775
3776 Doing this is almost never wrong, sometimes a better-performance model
3777 exists, but it is always a good start.
3778
3779 =item * other models exist, such as the leader/follower pattern, where one
3780 loop is handed through multiple threads in a kind of round-robin fashion.
3781
3782 Choosing a model is hard - look around, learn, know that usually you can do
3783 better than you currently do :-)
3784
3785 =item * often you need to talk to some other thread which blocks in the
3786 event loop.
3787
3788 C<ev_async> watchers can be used to wake them up from other threads safely
3789 (or from signal contexts...).
3790
3791 An example use would be to communicate signals or other events that only
3792 work in the default loop by registering the signal watcher with the
3793 default loop and triggering an C<ev_async> watcher from the default loop
3794 watcher callback into the event loop interested in the signal.
3795
3796 =back
3797
3798 =head3 COROUTINES
3799
3800 Libev is very accommodating to coroutines ("cooperative threads"):
3801 libev fully supports nesting calls to its functions from different
3802 coroutines (e.g. you can call C<ev_loop> on the same loop from two
3803 different coroutines, and switch freely between both coroutines running the
3804 loop, as long as you don't confuse yourself). The only exception is that
3805 you must not do this from C<ev_periodic> reschedule callbacks.
3806
3807 Care has been taken to ensure that libev does not keep local state inside
3808 C<ev_loop>, and other calls do not usually allow for coroutine switches as
3809 they do not call any callbacks.
3810
3811 =head2 COMPILER WARNINGS
3812
3813 Depending on your compiler and compiler settings, you might get no or a
3814 lot of warnings when compiling libev code. Some people are apparently
3815 scared by this.
3816
3817 However, these are unavoidable for many reasons. For one, each compiler
3818 has different warnings, and each user has different tastes regarding
3819 warning options. "Warn-free" code therefore cannot be a goal except when
3820 targeting a specific compiler and compiler-version.
3821
3822 Another reason is that some compiler warnings require elaborate
3823 workarounds, or other changes to the code that make it less clear and less
3824 maintainable.
3825
3826 And of course, some compiler warnings are just plain stupid, or simply
3827 wrong (because they don't actually warn about the condition their message
3828 seems to warn about). For example, certain older gcc versions had some
3829 warnings that resulted an extreme number of false positives. These have
3830 been fixed, but some people still insist on making code warn-free with
3831 such buggy versions.
3832
3833 While libev is written to generate as few warnings as possible,
3834 "warn-free" code is not a goal, and it is recommended not to build libev
3835 with any compiler warnings enabled unless you are prepared to cope with
3836 them (e.g. by ignoring them). Remember that warnings are just that:
3837 warnings, not errors, or proof of bugs.
3838
3839
3840 =head2 VALGRIND
3841
3842 Valgrind has a special section here because it is a popular tool that is
3843 highly useful. Unfortunately, valgrind reports are very hard to interpret.
3844
3845 If you think you found a bug (memory leak, uninitialised data access etc.)
3846 in libev, then check twice: If valgrind reports something like:
3847
3848 ==2274== definitely lost: 0 bytes in 0 blocks.
3849 ==2274== possibly lost: 0 bytes in 0 blocks.
3850 ==2274== still reachable: 256 bytes in 1 blocks.
3851
3852 Then there is no memory leak, just as memory accounted to global variables
3853 is not a memleak - the memory is still being referenced, and didn't leak.
3854
3855 Similarly, under some circumstances, valgrind might report kernel bugs
3856 as if it were a bug in libev (e.g. in realloc or in the poll backend,
3857 although an acceptable workaround has been found here), or it might be
3858 confused.
3859
3860 Keep in mind that valgrind is a very good tool, but only a tool. Don't
3861 make it into some kind of religion.
3862
3863 If you are unsure about something, feel free to contact the mailing list
3864 with the full valgrind report and an explanation on why you think this
3865 is a bug in libev (best check the archives, too :). However, don't be
3866 annoyed when you get a brisk "this is no bug" answer and take the chance
3867 of learning how to interpret valgrind properly.
3868
3869 If you need, for some reason, empty reports from valgrind for your project
3870 I suggest using suppression lists.
3871
3872
3873 =head1 PORTABILITY NOTES
3874
3875 =head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
3876
3877 Win32 doesn't support any of the standards (e.g. POSIX) that libev
3878 requires, and its I/O model is fundamentally incompatible with the POSIX
3879 model. Libev still offers limited functionality on this platform in
3880 the form of the C<EVBACKEND_SELECT> backend, and only supports socket
3881 descriptors. This only applies when using Win32 natively, not when using
3882 e.g. cygwin.
3883
3884 Lifting these limitations would basically require the full
3885 re-implementation of the I/O system. If you are into these kinds of
3886 things, then note that glib does exactly that for you in a very portable
3887 way (note also that glib is the slowest event library known to man).
3888
3889 There is no supported compilation method available on windows except
3890 embedding it into other applications.
3891
3892 Not a libev limitation but worth mentioning: windows apparently doesn't
3893 accept large writes: instead of resulting in a partial write, windows will
3894 either accept everything or return C<ENOBUFS> if the buffer is too large,
3895 so make sure you only write small amounts into your sockets (less than a
3896 megabyte seems safe, but this apparently depends on the amount of memory
3897 available).
3898
3899 Due to the many, low, and arbitrary limits on the win32 platform and
3900 the abysmal performance of winsockets, using a large number of sockets
3901 is not recommended (and not reasonable). If your program needs to use
3902 more than a hundred or so sockets, then likely it needs to use a totally
3903 different implementation for windows, as libev offers the POSIX readiness
3904 notification model, which cannot be implemented efficiently on windows
3905 (Microsoft monopoly games).
3906
3907 A typical way to use libev under windows is to embed it (see the embedding
3908 section for details) and use the following F<evwrap.h> header file instead
3909 of F<ev.h>:
3910
3911 #define EV_STANDALONE /* keeps ev from requiring config.h */
3912 #define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3913
3914 #include "ev.h"
3915
3916 And compile the following F<evwrap.c> file into your project (make sure
3917 you do I<not> compile the F<ev.c> or any other embedded source files!):
3918
3919 #include "evwrap.h"
3920 #include "ev.c"
3921
3922 =over 4
3923
3924 =item The winsocket select function
3925
3926 The winsocket C<select> function doesn't follow POSIX in that it
3927 requires socket I<handles> and not socket I<file descriptors> (it is
3928 also extremely buggy). This makes select very inefficient, and also
3929 requires a mapping from file descriptors to socket handles (the Microsoft
3930 C runtime provides the function C<_open_osfhandle> for this). See the
3931 discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
3932 C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
3933
3934 The configuration for a "naked" win32 using the Microsoft runtime
3935 libraries and raw winsocket select is:
3936
3937 #define EV_USE_SELECT 1
3938 #define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3939
3940 Note that winsockets handling of fd sets is O(n), so you can easily get a
3941 complexity in the O(n²) range when using win32.
3942
3943 =item Limited number of file descriptors
3944
3945 Windows has numerous arbitrary (and low) limits on things.
3946
3947 Early versions of winsocket's select only supported waiting for a maximum
3948 of C<64> handles (probably owning to the fact that all windows kernels
3949 can only wait for C<64> things at the same time internally; Microsoft
3950 recommends spawning a chain of threads and wait for 63 handles and the
3951 previous thread in each. Great).
3952
3953 Newer versions support more handles, but you need to define C<FD_SETSIZE>
3954 to some high number (e.g. C<2048>) before compiling the winsocket select
3955 call (which might be in libev or elsewhere, for example, perl does its own
3956 select emulation on windows).
3957
3958 Another limit is the number of file descriptors in the Microsoft runtime
3959 libraries, which by default is C<64> (there must be a hidden I<64> fetish
3960 or something like this inside Microsoft). You can increase this by calling
3961 C<_setmaxstdio>, which can increase this limit to C<2048> (another
3962 arbitrary limit), but is broken in many versions of the Microsoft runtime
3963 libraries.
3964
3965 This might get you to about C<512> or C<2048> sockets (depending on
3966 windows version and/or the phase of the moon). To get more, you need to
3967 wrap all I/O functions and provide your own fd management, but the cost of
3968 calling select (O(n²)) will likely make this unworkable.
3969
3970 =back
3971
3972 =head2 PORTABILITY REQUIREMENTS
3973
3974 In addition to a working ISO-C implementation and of course the
3975 backend-specific APIs, libev relies on a few additional extensions:
3976
3977 =over 4
3978
3979 =item C<void (*)(ev_watcher_type *, int revents)> must have compatible
3980 calling conventions regardless of C<ev_watcher_type *>.
3981
3982 Libev assumes not only that all watcher pointers have the same internal
3983 structure (guaranteed by POSIX but not by ISO C for example), but it also
3984 assumes that the same (machine) code can be used to call any watcher
3985 callback: The watcher callbacks have different type signatures, but libev
3986 calls them using an C<ev_watcher *> internally.
3987
3988 =item C<sig_atomic_t volatile> must be thread-atomic as well
3989
3990 The type C<sig_atomic_t volatile> (or whatever is defined as
3991 C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
3992 threads. This is not part of the specification for C<sig_atomic_t>, but is
3993 believed to be sufficiently portable.
3994
3995 =item C<sigprocmask> must work in a threaded environment
3996
3997 Libev uses C<sigprocmask> to temporarily block signals. This is not
3998 allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
3999 pthread implementations will either allow C<sigprocmask> in the "main
4000 thread" or will block signals process-wide, both behaviours would
4001 be compatible with libev. Interaction between C<sigprocmask> and
4002 C<pthread_sigmask> could complicate things, however.
4003
4004 The most portable way to handle signals is to block signals in all threads
4005 except the initial one, and run the default loop in the initial thread as
4006 well.
4007
4008 =item C<long> must be large enough for common memory allocation sizes
4009
4010 To improve portability and simplify its API, libev uses C<long> internally
4011 instead of C<size_t> when allocating its data structures. On non-POSIX
4012 systems (Microsoft...) this might be unexpectedly low, but is still at
4013 least 31 bits everywhere, which is enough for hundreds of millions of
4014 watchers.
4015
4016 =item C<double> must hold a time value in seconds with enough accuracy
4017
4018 The type C<double> is used to represent timestamps. It is required to
4019 have at least 51 bits of mantissa (and 9 bits of exponent), which is good
4020 enough for at least into the year 4000. This requirement is fulfilled by
4021 implementations implementing IEEE 754 (basically all existing ones).
4022
4023 =back
4024
4025 If you know of other additional requirements drop me a note.
4026
4027
4028 =head1 ALGORITHMIC COMPLEXITIES
4029
4030 In this section the complexities of (many of) the algorithms used inside
4031 libev will be documented. For complexity discussions about backends see
4032 the documentation for C<ev_default_init>.
4033
4034 All of the following are about amortised time: If an array needs to be
4035 extended, libev needs to realloc and move the whole array, but this
4036 happens asymptotically rarer with higher number of elements, so O(1) might
4037 mean that libev does a lengthy realloc operation in rare cases, but on
4038 average it is much faster and asymptotically approaches constant time.
4039
4040 =over 4
4041
4042 =item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
4043
4044 This means that, when you have a watcher that triggers in one hour and
4045 there are 100 watchers that would trigger before that, then inserting will
4046 have to skip roughly seven (C<ld 100>) of these watchers.
4047
4048 =item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
4049
4050 That means that changing a timer costs less than removing/adding them,
4051 as only the relative motion in the event queue has to be paid for.
4052
4053 =item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
4054
4055 These just add the watcher into an array or at the head of a list.
4056
4057 =item Stopping check/prepare/idle/fork/async watchers: O(1)
4058
4059 =item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
4060
4061 These watchers are stored in lists, so they need to be walked to find the
4062 correct watcher to remove. The lists are usually short (you don't usually
4063 have many watchers waiting for the same fd or signal: one is typical, two
4064 is rare).
4065
4066 =item Finding the next timer in each loop iteration: O(1)
4067
4068 By virtue of using a binary or 4-heap, the next timer is always found at a
4069 fixed position in the storage array.
4070
4071 =item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
4072
4073 A change means an I/O watcher gets started or stopped, which requires
4074 libev to recalculate its status (and possibly tell the kernel, depending
4075 on backend and whether C<ev_io_set> was used).
4076
4077 =item Activating one watcher (putting it into the pending state): O(1)
4078
4079 =item Priority handling: O(number_of_priorities)
4080
4081 Priorities are implemented by allocating some space for each
4082 priority. When doing priority-based operations, libev usually has to
4083 linearly search all the priorities, but starting/stopping and activating
4084 watchers becomes O(1) with respect to priority handling.
4085
4086 =item Sending an ev_async: O(1)
4087
4088 =item Processing ev_async_send: O(number_of_async_watchers)
4089
4090 =item Processing signals: O(max_signal_number)
4091
4092 Sending involves a system call I<iff> there were no other C<ev_async_send>
4093 calls in the current loop iteration. Checking for async and signal events
4094 involves iterating over all running async watchers or all signal numbers.
4095
4096 =back
4097
4098
4099 =head1 GLOSSARY
4100
4101 =over 4
4102
4103 =item active
4104
4105 A watcher is active as long as it has been started (has been attached to
4106 an event loop) but not yet stopped (disassociated from the event loop).
4107
4108 =item application
4109
4110 In this document, an application is whatever is using libev.
4111
4112 =item callback
4113
4114 The address of a function that is called when some event has been
4115 detected. Callbacks are being passed the event loop, the watcher that
4116 received the event, and the actual event bitset.
4117
4118 =item callback invocation
4119
4120 The act of calling the callback associated with a watcher.
4121
4122 =item event
4123
4124 A change of state of some external event, such as data now being available
4125 for reading on a file descriptor, time having passed or simply not having
4126 any other events happening anymore.
4127
4128 In libev, events are represented as single bits (such as C<EV_READ> or
4129 C<EV_TIMEOUT>).
4130
4131 =item event library
4132
4133 A software package implementing an event model and loop.
4134
4135 =item event loop
4136
4137 An entity that handles and processes external events and converts them
4138 into callback invocations.
4139
4140 =item event model
4141
4142 The model used to describe how an event loop handles and processes
4143 watchers and events.
4144
4145 =item pending
4146
4147 A watcher is pending as soon as the corresponding event has been detected,
4148 and stops being pending as soon as the watcher will be invoked or its
4149 pending status is explicitly cleared by the application.
4150
4151 A watcher can be pending, but not active. Stopping a watcher also clears
4152 its pending status.
4153
4154 =item real time
4155
4156 The physical time that is observed. It is apparently strictly monotonic :)
4157
4158 =item wall-clock time
4159
4160 The time and date as shown on clocks. Unlike real time, it can actually
4161 be wrong and jump forwards and backwards, e.g. when the you adjust your
4162 clock.
4163
4164 =item watcher
4165
4166 A data structure that describes interest in certain events. Watchers need
4167 to be started (attached to an event loop) before they can receive events.
4168
4169 =item watcher invocation
4170
4171 The act of calling the callback associated with a watcher.
4172
4173 =back
4174
4175 =head1 AUTHOR
4176
4177 Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
4178