ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.337 by root, Sun Oct 31 20:20:20 2010 UTC vs.
Revision 1.434 by root, Tue May 6 13:24:39 2014 UTC

		1	=encoding utf-8
		2
1	=head1 NAME	3	=head1 NAME
2		4
3	libev - a high performance full-featured event loop written in C	5	libev - a high performance full-featured event loop written in C
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
…		…
58	ev_timer_start (loop, &timeout_watcher);	60	ev_timer_start (loop, &timeout_watcher);
59		61
60	// now wait for events to arrive	62	// now wait for events to arrive
61	ev_run (loop, 0);	63	ev_run (loop, 0);
62		64
63	// unloop was called, so exit	65	// break was called, so exit
64	return 0;	66	return 0;
65	}	67	}
66		68
67	=head1 ABOUT THIS DOCUMENT	69	=head1 ABOUT THIS DOCUMENT
68		70
…		…
82		84
83	=head1 WHAT TO READ WHEN IN A HURRY	85	=head1 WHAT TO READ WHEN IN A HURRY
84		86
85	This manual tries to be very detailed, but unfortunately, this also makes	87	This manual tries to be very detailed, but unfortunately, this also makes
86	it very long. If you just want to know the basics of libev, I suggest	88	it very long. If you just want to know the basics of libev, I suggest
87	reading L<ANATOMY OF A WATCHER>, then the L<EXAMPLE PROGRAM> above and	89	reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
88	look up the missing functions in L<GLOBAL FUNCTIONS> and the C<ev_io> and	90	look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
89	C<ev_timer> sections in L<WATCHER TYPES>.	91	C<ev_timer> sections in L</WATCHER TYPES>.
90		92
91	=head1 ABOUT LIBEV	93	=head1 ABOUT LIBEV
92		94
93	Libev is an event loop: you register interest in certain events (such as a	95	Libev is an event loop: you register interest in certain events (such as a
94	file descriptor being readable or a timeout occurring), and it will manage	96	file descriptor being readable or a timeout occurring), and it will manage
…		…
174	=item ev_tstamp ev_time ()	176	=item ev_tstamp ev_time ()
175		177
176	Returns the current time as libev would use it. Please note that the	178	Returns the current time as libev would use it. Please note that the
177	C<ev_now> function is usually faster and also often returns the timestamp	179	C<ev_now> function is usually faster and also often returns the timestamp
178	you actually want to know. Also interesting is the combination of	180	you actually want to know. Also interesting is the combination of
179	C<ev_update_now> and C<ev_now>.	181	C<ev_now_update> and C<ev_now>.
180		182
181	=item ev_sleep (ev_tstamp interval)	183	=item ev_sleep (ev_tstamp interval)
182		184
183	Sleep for the given interval: The current thread will be blocked until	185	Sleep for the given interval: The current thread will be blocked
184	either it is interrupted or the given time interval has passed. Basically	186	until either it is interrupted or the given time interval has
		187	passed (approximately - it might return a bit earlier even if not
		188	interrupted). Returns immediately if C<< interval <= 0 >>.
		189
185	this is a sub-second-resolution C<sleep ()>.	190	Basically this is a sub-second-resolution C<sleep ()>.
		191
		192	The range of the C<interval> is limited - libev only guarantees to work
		193	with sleep times of up to one day (C<< interval <= 86400 >>).
186		194
187	=item int ev_version_major ()	195	=item int ev_version_major ()
188		196
189	=item int ev_version_minor ()	197	=item int ev_version_minor ()
190		198
…		…
241	the current system, you would need to look at C<ev_embeddable_backends ()	249	the current system, you would need to look at C<ev_embeddable_backends ()
242	& ev_supported_backends ()>, likewise for recommended ones.	250	& ev_supported_backends ()>, likewise for recommended ones.
243		251
244	See the description of C<ev_embed> watchers for more info.	252	See the description of C<ev_embed> watchers for more info.
245		253
246	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]	254	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
247		255
248	Sets the allocation function to use (the prototype is similar - the	256	Sets the allocation function to use (the prototype is similar - the
249	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is	257	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
250	used to allocate and free memory (no surprises here). If it returns zero	258	used to allocate and free memory (no surprises here). If it returns zero
251	when memory needs to be allocated (C<size != 0>), the library might abort	259	when memory needs to be allocated (C<size != 0>), the library might abort
…		…
277	}	285	}
278		286
279	...	287	...
280	ev_set_allocator (persistent_realloc);	288	ev_set_allocator (persistent_realloc);
281		289
282	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]	290	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
283		291
284	Set the callback function to call on a retryable system call error (such	292	Set the callback function to call on a retryable system call error (such
285	as failed select, poll, epoll_wait). The message is a printable string	293	as failed select, poll, epoll_wait). The message is a printable string
286	indicating the system call or subsystem causing the problem. If this	294	indicating the system call or subsystem causing the problem. If this
287	callback is set, then libev will expect it to remedy the situation, no	295	callback is set, then libev will expect it to remedy the situation, no
…		…
299	}	307	}
300		308
301	...	309	...
302	ev_set_syserr_cb (fatal_error);	310	ev_set_syserr_cb (fatal_error);
303		311
		312	=item ev_feed_signal (int signum)
		313
		314	This function can be used to "simulate" a signal receive. It is completely
		315	safe to call this function at any time, from any context, including signal
		316	handlers or random threads.
		317
		318	Its main use is to customise signal handling in your process, especially
		319	in the presence of threads. For example, you could block signals
		320	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		321	creating any loops), and in one thread, use C<sigwait> or any other
		322	mechanism to wait for signals, then "deliver" them to libev by calling
		323	C<ev_feed_signal>.
		324
304	=back	325	=back
305		326
306	=head1 FUNCTIONS CONTROLLING EVENT LOOPS	327	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
307		328
308	An event loop is described by a C<struct ev_loop *> (the C<struct> is	329	An event loop is described by a C<struct ev_loop *> (the C<struct> is
…		…
355	=item struct ev_loop *ev_loop_new (unsigned int flags)	376	=item struct ev_loop *ev_loop_new (unsigned int flags)
356		377
357	This will create and initialise a new event loop object. If the loop	378	This will create and initialise a new event loop object. If the loop
358	could not be initialised, returns false.	379	could not be initialised, returns false.
359		380
360	Note that this function I<is> thread-safe, and one common way to use	381	This function is thread-safe, and one common way to use libev with
361	libev with threads is indeed to create one loop per thread, and using the	382	threads is indeed to create one loop per thread, and using the default
362	default loop in the "main" or "initial" thread.	383	loop in the "main" or "initial" thread.
363		384
364	The flags argument can be used to specify special behaviour or specific	385	The flags argument can be used to specify special behaviour or specific
365	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).	386	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
366		387
367	The following flags are supported:	388	The following flags are supported:
…		…
377		398
378	If this flag bit is or'ed into the flag value (or the program runs setuid	399	If this flag bit is or'ed into the flag value (or the program runs setuid
379	or setgid) then libev will I<not> look at the environment variable	400	or setgid) then libev will I<not> look at the environment variable
380	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	401	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
381	override the flags completely if it is found in the environment. This is	402	override the flags completely if it is found in the environment. This is
382	useful to try out specific backends to test their performance, or to work	403	useful to try out specific backends to test their performance, to work
383	around bugs.	404	around bugs, or to make libev threadsafe (accessing environment variables
		405	cannot be done in a threadsafe way, but usually it works if no other
		406	thread modifies them).
384		407
385	=item C<EVFLAG_FORKCHECK>	408	=item C<EVFLAG_FORKCHECK>
386		409
387	Instead of calling C<ev_loop_fork> manually after a fork, you can also	410	Instead of calling C<ev_loop_fork> manually after a fork, you can also
388	make libev check for a fork in each iteration by enabling this flag.	411	make libev check for a fork in each iteration by enabling this flag.
…		…
402	environment variable.	425	environment variable.
403		426
404	=item C<EVFLAG_NOINOTIFY>	427	=item C<EVFLAG_NOINOTIFY>
405		428
406	When this flag is specified, then libev will not attempt to use the	429	When this flag is specified, then libev will not attempt to use the
407	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and	430	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
408	testing, this flag can be useful to conserve inotify file descriptors, as	431	testing, this flag can be useful to conserve inotify file descriptors, as
409	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.	432	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
410		433
411	=item C<EVFLAG_SIGNALFD>	434	=item C<EVFLAG_SIGNALFD>
412		435
413	When this flag is specified, then libev will attempt to use the	436	When this flag is specified, then libev will attempt to use the
414	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API	437	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
415	delivers signals synchronously, which makes it both faster and might make	438	delivers signals synchronously, which makes it both faster and might make
416	it possible to get the queued signal data. It can also simplify signal	439	it possible to get the queued signal data. It can also simplify signal
417	handling with threads, as long as you properly block signals in your	440	handling with threads, as long as you properly block signals in your
418	threads that are not interested in handling them.	441	threads that are not interested in handling them.
419		442
420	Signalfd will not be used by default as this changes your signal mask, and	443	Signalfd will not be used by default as this changes your signal mask, and
421	there are a lot of shoddy libraries and programs (glib's threadpool for	444	there are a lot of shoddy libraries and programs (glib's threadpool for
422	example) that can't properly initialise their signal masks.	445	example) that can't properly initialise their signal masks.
		446
		447	=item C<EVFLAG_NOSIGMASK>
		448
		449	When this flag is specified, then libev will avoid to modify the signal
		450	mask. Specifically, this means you have to make sure signals are unblocked
		451	when you want to receive them.
		452
		453	This behaviour is useful when you want to do your own signal handling, or
		454	want to handle signals only in specific threads and want to avoid libev
		455	unblocking the signals.
		456
		457	It's also required by POSIX in a threaded program, as libev calls
		458	C<sigprocmask>, whose behaviour is officially unspecified.
		459
		460	This flag's behaviour will become the default in future versions of libev.
423		461
424	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	462	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
425		463
426	This is your standard select(2) backend. Not I<completely> standard, as	464	This is your standard select(2) backend. Not I<completely> standard, as
427	libev tries to roll its own fd_set with no limits on the number of fds,	465	libev tries to roll its own fd_set with no limits on the number of fds,
…		…
455	=item C<EVBACKEND_EPOLL> (value 4, Linux)	493	=item C<EVBACKEND_EPOLL> (value 4, Linux)
456		494
457	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9	495	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
458	kernels).	496	kernels).
459		497
460	For few fds, this backend is a bit little slower than poll and select,	498	For few fds, this backend is a bit little slower than poll and select, but
461	but it scales phenomenally better. While poll and select usually scale	499	it scales phenomenally better. While poll and select usually scale like
462	like O(total_fds) where n is the total number of fds (or the highest fd),	500	O(total_fds) where total_fds is the total number of fds (or the highest
463	epoll scales either O(1) or O(active_fds).	501	fd), epoll scales either O(1) or O(active_fds).
464		502
465	The epoll mechanism deserves honorable mention as the most misdesigned	503	The epoll mechanism deserves honorable mention as the most misdesigned
466	of the more advanced event mechanisms: mere annoyances include silently	504	of the more advanced event mechanisms: mere annoyances include silently
467	dropping file descriptors, requiring a system call per change per file	505	dropping file descriptors, requiring a system call per change per file
468	descriptor (and unnecessary guessing of parameters), problems with dup,	506	descriptor (and unnecessary guessing of parameters), problems with dup,
469	returning before the timeout value requiring additional iterations and so	507	returning before the timeout value, resulting in additional iterations
		508	(and only giving 5ms accuracy while select on the same platform gives
470	on. The biggest issue is fork races, however - if a program forks then	509	0.1ms) and so on. The biggest issue is fork races, however - if a program
471	I<both> parent and child process have to recreate the epoll set, which can	510	forks then I<both> parent and child process have to recreate the epoll
472	take considerable time (one syscall per file descriptor) and is of course	511	set, which can take considerable time (one syscall per file descriptor)
473	hard to detect.	512	and is of course hard to detect.
474		513
475	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but	514	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
476	of course I<doesn't>, and epoll just loves to report events for totally	515	but of course I<doesn't>, and epoll just loves to report events for
477	I<different> file descriptors (even already closed ones, so one cannot	516	totally I<different> file descriptors (even already closed ones, so
478	even remove them from the set) than registered in the set (especially	517	one cannot even remove them from the set) than registered in the set
479	on SMP systems). Libev tries to counter these spurious notifications by	518	(especially on SMP systems). Libev tries to counter these spurious
480	employing an additional generation counter and comparing that against the	519	notifications by employing an additional generation counter and comparing
481	events to filter out spurious ones, recreating the set when required. Last	520	that against the events to filter out spurious ones, recreating the set
		521	when required. Epoll also erroneously rounds down timeouts, but gives you
		522	no way to know when and by how much, so sometimes you have to busy-wait
		523	because epoll returns immediately despite a nonzero timeout. And last
482	not least, it also refuses to work with some file descriptors which work	524	not least, it also refuses to work with some file descriptors which work
483	perfectly fine with C<select> (files, many character devices...).	525	perfectly fine with C<select> (files, many character devices...).
		526
		527	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		528	cobbled together in a hurry, no thought to design or interaction with
		529	others. Oh, the pain, will it ever stop...
484		530
485	While stopping, setting and starting an I/O watcher in the same iteration	531	While stopping, setting and starting an I/O watcher in the same iteration
486	will result in some caching, there is still a system call per such	532	will result in some caching, there is still a system call per such
487	incident (because the same I<file descriptor> could point to a different	533	incident (because the same I<file descriptor> could point to a different
488	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed	534	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
…		…
525		571
526	It scales in the same way as the epoll backend, but the interface to the	572	It scales in the same way as the epoll backend, but the interface to the
527	kernel is more efficient (which says nothing about its actual speed, of	573	kernel is more efficient (which says nothing about its actual speed, of
528	course). While stopping, setting and starting an I/O watcher does never	574	course). While stopping, setting and starting an I/O watcher does never
529	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	575	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
530	two event changes per incident. Support for C<fork ()> is very bad (but	576	two event changes per incident. Support for C<fork ()> is very bad (you
531	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect	577	might have to leak fd's on fork, but it's more sane than epoll) and it
532	cases	578	drops fds silently in similarly hard-to-detect cases.
533		579
534	This backend usually performs well under most conditions.	580	This backend usually performs well under most conditions.
535		581
536	While nominally embeddable in other event loops, this doesn't work	582	While nominally embeddable in other event loops, this doesn't work
537	everywhere, so you might need to test for this. And since it is broken	583	everywhere, so you might need to test for this. And since it is broken
…		…
554	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	600	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
555		601
556	This uses the Solaris 10 event port mechanism. As with everything on Solaris,	602	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
557	it's really slow, but it still scales very well (O(active_fds)).	603	it's really slow, but it still scales very well (O(active_fds)).
558		604
559	Please note that Solaris event ports can deliver a lot of spurious
560	notifications, so you need to use non-blocking I/O or other means to avoid
561	blocking when no data (or space) is available.
562
563	While this backend scales well, it requires one system call per active	605	While this backend scales well, it requires one system call per active
564	file descriptor per loop iteration. For small and medium numbers of file	606	file descriptor per loop iteration. For small and medium numbers of file
565	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	607	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
566	might perform better.	608	might perform better.
567		609
568	On the positive side, with the exception of the spurious readiness	610	On the positive side, this backend actually performed fully to
569	notifications, this backend actually performed fully to specification
570	in all tests and is fully embeddable, which is a rare feat among the	611	specification in all tests and is fully embeddable, which is a rare feat
571	OS-specific backends (I vastly prefer correctness over speed hacks).	612	among the OS-specific backends (I vastly prefer correctness over speed
		613	hacks).
		614
		615	On the negative side, the interface is I<bizarre> - so bizarre that
		616	even sun itself gets it wrong in their code examples: The event polling
		617	function sometimes returns events to the caller even though an error
		618	occurred, but with no indication whether it has done so or not (yes, it's
		619	even documented that way) - deadly for edge-triggered interfaces where you
		620	absolutely have to know whether an event occurred or not because you have
		621	to re-arm the watcher.
		622
		623	Fortunately libev seems to be able to work around these idiocies.
572		624
573	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	625	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
574	C<EVBACKEND_POLL>.	626	C<EVBACKEND_POLL>.
575		627
576	=item C<EVBACKEND_ALL>	628	=item C<EVBACKEND_ALL>
577		629
578	Try all backends (even potentially broken ones that wouldn't be tried	630	Try all backends (even potentially broken ones that wouldn't be tried
579	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	631	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
580	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	632	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
581		633
582	It is definitely not recommended to use this flag.	634	It is definitely not recommended to use this flag, use whatever
		635	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		636	at all.
		637
		638	=item C<EVBACKEND_MASK>
		639
		640	Not a backend at all, but a mask to select all backend bits from a
		641	C<flags> value, in case you want to mask out any backends from a flags
		642	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
583		643
584	=back	644	=back
585		645
586	If one or more of the backend flags are or'ed into the flags value,	646	If one or more of the backend flags are or'ed into the flags value,
587	then only these backends will be tried (in the reverse order as listed	647	then only these backends will be tried (in the reverse order as listed
…		…
616	This function is normally used on loop objects allocated by	676	This function is normally used on loop objects allocated by
617	C<ev_loop_new>, but it can also be used on the default loop returned by	677	C<ev_loop_new>, but it can also be used on the default loop returned by
618	C<ev_default_loop>, in which case it is not thread-safe.	678	C<ev_default_loop>, in which case it is not thread-safe.
619		679
620	Note that it is not advisable to call this function on the default loop	680	Note that it is not advisable to call this function on the default loop
621	except in the rare occasion where you really need to free it's resources.	681	except in the rare occasion where you really need to free its resources.
622	If you need dynamically allocated loops it is better to use C<ev_loop_new>	682	If you need dynamically allocated loops it is better to use C<ev_loop_new>
623	and C<ev_loop_destroy>.	683	and C<ev_loop_destroy>.
624		684
625	=item ev_loop_fork (loop)	685	=item ev_loop_fork (loop)
626		686
627	This function sets a flag that causes subsequent C<ev_run> iterations to	687	This function sets a flag that causes subsequent C<ev_run> iterations
628	reinitialise the kernel state for backends that have one. Despite the	688	to reinitialise the kernel state for backends that have one. Despite
629	name, you can call it anytime, but it makes most sense after forking, in	689	the name, you can call it anytime you are allowed to start or stop
630	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the	690	watchers (except inside an C<ev_prepare> callback), but it makes most
		691	sense after forking, in the child process. You I<must> call it (or use
631	child before resuming or calling C<ev_run>.	692	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
632		693
633	Again, you I<have> to call it on I<any> loop that you want to re-use after	694	Again, you I<have> to call it on I<any> loop that you want to re-use after
634	a fork, I<even if you do not plan to use the loop in the parent>. This is	695	a fork, I<even if you do not plan to use the loop in the parent>. This is
635	because some kernel interfaces *cough* I<kqueue> *cough* do funny things	696	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
636	during fork.	697	during fork.
637		698
638	On the other hand, you only need to call this function in the child	699	On the other hand, you only need to call this function in the child
…		…
674	prepare and check phases.	735	prepare and check phases.
675		736
676	=item unsigned int ev_depth (loop)	737	=item unsigned int ev_depth (loop)
677		738
678	Returns the number of times C<ev_run> was entered minus the number of	739	Returns the number of times C<ev_run> was entered minus the number of
679	times C<ev_run> was exited, in other words, the recursion depth.	740	times C<ev_run> was exited normally, in other words, the recursion depth.
680		741
681	Outside C<ev_run>, this number is zero. In a callback, this number is	742	Outside C<ev_run>, this number is zero. In a callback, this number is
682	C<1>, unless C<ev_run> was invoked recursively (or from another thread),	743	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
683	in which case it is higher.	744	in which case it is higher.
684		745
685	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread	746	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
686	etc.), doesn't count as "exit" - consider this as a hint to avoid such	747	throwing an exception etc.), doesn't count as "exit" - consider this
687	ungentleman-like behaviour unless it's really convenient.	748	as a hint to avoid such ungentleman-like behaviour unless it's really
		749	convenient, in which case it is fully supported.
688		750
689	=item unsigned int ev_backend (loop)	751	=item unsigned int ev_backend (loop)
690		752
691	Returns one of the C<EVBACKEND_*> flags indicating the event backend in	753	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
692	use.	754	use.
…		…
707		769
708	This function is rarely useful, but when some event callback runs for a	770	This function is rarely useful, but when some event callback runs for a
709	very long time without entering the event loop, updating libev's idea of	771	very long time without entering the event loop, updating libev's idea of
710	the current time is a good idea.	772	the current time is a good idea.
711		773
712	See also L<The special problem of time updates> in the C<ev_timer> section.	774	See also L</The special problem of time updates> in the C<ev_timer> section.
713		775
714	=item ev_suspend (loop)	776	=item ev_suspend (loop)
715		777
716	=item ev_resume (loop)	778	=item ev_resume (loop)
717		779
…		…
735	without a previous call to C<ev_suspend>.	797	without a previous call to C<ev_suspend>.
736		798
737	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the	799	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
738	event loop time (see C<ev_now_update>).	800	event loop time (see C<ev_now_update>).
739		801
740	=item ev_run (loop, int flags)	802	=item bool ev_run (loop, int flags)
741		803
742	Finally, this is it, the event handler. This function usually is called	804	Finally, this is it, the event handler. This function usually is called
743	after you have initialised all your watchers and you want to start	805	after you have initialised all your watchers and you want to start
744	handling events. It will ask the operating system for any new events, call	806	handling events. It will ask the operating system for any new events, call
745	the watcher callbacks, an then repeat the whole process indefinitely: This	807	the watcher callbacks, and then repeat the whole process indefinitely: This
746	is why event loops are called I<loops>.	808	is why event loops are called I<loops>.
747		809
748	If the flags argument is specified as C<0>, it will keep handling events	810	If the flags argument is specified as C<0>, it will keep handling events
749	until either no event watchers are active anymore or C<ev_break> was	811	until either no event watchers are active anymore or C<ev_break> was
750	called.	812	called.
		813
		814	The return value is false if there are no more active watchers (which
		815	usually means "all jobs done" or "deadlock"), and true in all other cases
		816	(which usually means " you should call C<ev_run> again").
751		817
752	Please note that an explicit C<ev_break> is usually better than	818	Please note that an explicit C<ev_break> is usually better than
753	relying on all watchers to be stopped when deciding when a program has	819	relying on all watchers to be stopped when deciding when a program has
754	finished (especially in interactive programs), but having a program	820	finished (especially in interactive programs), but having a program
755	that automatically loops as long as it has to and no longer by virtue	821	that automatically loops as long as it has to and no longer by virtue
756	of relying on its watchers stopping correctly, that is truly a thing of	822	of relying on its watchers stopping correctly, that is truly a thing of
757	beauty.	823	beauty.
758		824
		825	This function is I<mostly> exception-safe - you can break out of a
		826	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		827	exception and so on. This does not decrement the C<ev_depth> value, nor
		828	will it clear any outstanding C<EVBREAK_ONE> breaks.
		829
759	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle	830	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
760	those events and any already outstanding ones, but will not wait and	831	those events and any already outstanding ones, but will not wait and
761	block your process in case there are no events and will return after one	832	block your process in case there are no events and will return after one
762	iteration of the loop. This is sometimes useful to poll and handle new	833	iteration of the loop. This is sometimes useful to poll and handle new
763	events while doing lengthy calculations, to keep the program responsive.	834	events while doing lengthy calculations, to keep the program responsive.
…		…
772	This is useful if you are waiting for some external event in conjunction	843	This is useful if you are waiting for some external event in conjunction
773	with something not expressible using other libev watchers (i.e. "roll your	844	with something not expressible using other libev watchers (i.e. "roll your
774	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is	845	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
775	usually a better approach for this kind of thing.	846	usually a better approach for this kind of thing.
776		847
777	Here are the gory details of what C<ev_run> does:	848	Here are the gory details of what C<ev_run> does (this is for your
		849	understanding, not a guarantee that things will work exactly like this in
		850	future versions):
778		851
779	- Increment loop depth.	852	- Increment loop depth.
780	- Reset the ev_break status.	853	- Reset the ev_break status.
781	- Before the first iteration, call any pending watchers.	854	- Before the first iteration, call any pending watchers.
782	LOOP:	855	LOOP:
…		…
815	anymore.	888	anymore.
816		889
817	... queue jobs here, make sure they register event watchers as long	890	... queue jobs here, make sure they register event watchers as long
818	... as they still have work to do (even an idle watcher will do..)	891	... as they still have work to do (even an idle watcher will do..)
819	ev_run (my_loop, 0);	892	ev_run (my_loop, 0);
820	... jobs done or somebody called unloop. yeah!	893	... jobs done or somebody called break. yeah!
821		894
822	=item ev_break (loop, how)	895	=item ev_break (loop, how)
823		896
824	Can be used to make a call to C<ev_run> return early (but only after it	897	Can be used to make a call to C<ev_run> return early (but only after it
825	has processed all outstanding events). The C<how> argument must be either	898	has processed all outstanding events). The C<how> argument must be either
826	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or	899	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
827	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.	900	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
828		901
829	This "break state" will be cleared when entering C<ev_run> again.	902	This "break state" will be cleared on the next call to C<ev_run>.
830		903
831	It is safe to call C<ev_break> from outside any C<ev_run> calls, too.	904	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		905	which case it will have no effect.
832		906
833	=item ev_ref (loop)	907	=item ev_ref (loop)
834		908
835	=item ev_unref (loop)	909	=item ev_unref (loop)
836		910
…		…
857	running when nothing else is active.	931	running when nothing else is active.
858		932
859	ev_signal exitsig;	933	ev_signal exitsig;
860	ev_signal_init (&exitsig, sig_cb, SIGINT);	934	ev_signal_init (&exitsig, sig_cb, SIGINT);
861	ev_signal_start (loop, &exitsig);	935	ev_signal_start (loop, &exitsig);
862	evf_unref (loop);	936	ev_unref (loop);
863		937
864	Example: For some weird reason, unregister the above signal handler again.	938	Example: For some weird reason, unregister the above signal handler again.
865		939
866	ev_ref (loop);	940	ev_ref (loop);
867	ev_signal_stop (loop, &exitsig);	941	ev_signal_stop (loop, &exitsig);
…		…
887	overhead for the actual polling but can deliver many events at once.	961	overhead for the actual polling but can deliver many events at once.
888		962
889	By setting a higher I<io collect interval> you allow libev to spend more	963	By setting a higher I<io collect interval> you allow libev to spend more
890	time collecting I/O events, so you can handle more events per iteration,	964	time collecting I/O events, so you can handle more events per iteration,
891	at the cost of increasing latency. Timeouts (both C<ev_periodic> and	965	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
892	C<ev_timer>) will be not affected. Setting this to a non-null value will	966	C<ev_timer>) will not be affected. Setting this to a non-null value will
893	introduce an additional C<ev_sleep ()> call into most loop iterations. The	967	introduce an additional C<ev_sleep ()> call into most loop iterations. The
894	sleep time ensures that libev will not poll for I/O events more often then	968	sleep time ensures that libev will not poll for I/O events more often then
895	once per this interval, on average.	969	once per this interval, on average (as long as the host time resolution is
		970	good enough).
896		971
897	Likewise, by setting a higher I<timeout collect interval> you allow libev	972	Likewise, by setting a higher I<timeout collect interval> you allow libev
898	to spend more time collecting timeouts, at the expense of increased	973	to spend more time collecting timeouts, at the expense of increased
899	latency/jitter/inexactness (the watcher callback will be called	974	latency/jitter/inexactness (the watcher callback will be called
900	later). C<ev_io> watchers will not be affected. Setting this to a non-null	975	later). C<ev_io> watchers will not be affected. Setting this to a non-null
…		…
946	invoke the actual watchers inside another context (another thread etc.).	1021	invoke the actual watchers inside another context (another thread etc.).
947		1022
948	If you want to reset the callback, use C<ev_invoke_pending> as new	1023	If you want to reset the callback, use C<ev_invoke_pending> as new
949	callback.	1024	callback.
950		1025
951	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))	1026	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
952		1027
953	Sometimes you want to share the same loop between multiple threads. This	1028	Sometimes you want to share the same loop between multiple threads. This
954	can be done relatively simply by putting mutex_lock/unlock calls around	1029	can be done relatively simply by putting mutex_lock/unlock calls around
955	each call to a libev function.	1030	each call to a libev function.
956		1031
957	However, C<ev_run> can run an indefinite time, so it is not feasible	1032	However, C<ev_run> can run an indefinite time, so it is not feasible
958	to wait for it to return. One way around this is to wake up the event	1033	to wait for it to return. One way around this is to wake up the event
959	loop via C<ev_break> and C<av_async_send>, another way is to set these	1034	loop via C<ev_break> and C<ev_async_send>, another way is to set these
960	I<release> and I<acquire> callbacks on the loop.	1035	I<release> and I<acquire> callbacks on the loop.
961		1036
962	When set, then C<release> will be called just before the thread is	1037	When set, then C<release> will be called just before the thread is
963	suspended waiting for new events, and C<acquire> is called just	1038	suspended waiting for new events, and C<acquire> is called just
964	afterwards.	1039	afterwards.
…		…
979	See also the locking example in the C<THREADS> section later in this	1054	See also the locking example in the C<THREADS> section later in this
980	document.	1055	document.
981		1056
982	=item ev_set_userdata (loop, void *data)	1057	=item ev_set_userdata (loop, void *data)
983		1058
984	=item ev_userdata (loop)	1059	=item void *ev_userdata (loop)
985		1060
986	Set and retrieve a single C<void *> associated with a loop. When	1061	Set and retrieve a single C<void *> associated with a loop. When
987	C<ev_set_userdata> has never been called, then C<ev_userdata> returns	1062	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
988	C<0.>	1063	C<0>.
989		1064
990	These two functions can be used to associate arbitrary data with a loop,	1065	These two functions can be used to associate arbitrary data with a loop,
991	and are intended solely for the C<invoke_pending_cb>, C<release> and	1066	and are intended solely for the C<invoke_pending_cb>, C<release> and
992	C<acquire> callbacks described above, but of course can be (ab-)used for	1067	C<acquire> callbacks described above, but of course can be (ab-)used for
993	any other purpose as well.	1068	any other purpose as well.
…		…
1104		1179
1105	=item C<EV_PREPARE>	1180	=item C<EV_PREPARE>
1106		1181
1107	=item C<EV_CHECK>	1182	=item C<EV_CHECK>
1108		1183
1109	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts	1184	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
1110	to gather new events, and all C<ev_check> watchers are invoked just after	1185	gather new events, and all C<ev_check> watchers are queued (not invoked)
1111	C<ev_run> has gathered them, but before it invokes any callbacks for any	1186	just after C<ev_run> has gathered them, but before it queues any callbacks
		1187	for any received events. That means C<ev_prepare> watchers are the last
		1188	watchers invoked before the event loop sleeps or polls for new events, and
		1189	C<ev_check> watchers will be invoked before any other watchers of the same
		1190	or lower priority within an event loop iteration.
		1191
1112	received events. Callbacks of both watcher types can start and stop as	1192	Callbacks of both watcher types can start and stop as many watchers as
1113	many watchers as they want, and all of them will be taken into account	1193	they want, and all of them will be taken into account (for example, a
1114	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1194	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
1115	C<ev_run> from blocking).	1195	blocking).
1116		1196
1117	=item C<EV_EMBED>	1197	=item C<EV_EMBED>
1118		1198
1119	The embedded event loop specified in the C<ev_embed> watcher needs attention.	1199	The embedded event loop specified in the C<ev_embed> watcher needs attention.
1120		1200
…		…
1243		1323
1244	=item callback ev_cb (ev_TYPE *watcher)	1324	=item callback ev_cb (ev_TYPE *watcher)
1245		1325
1246	Returns the callback currently set on the watcher.	1326	Returns the callback currently set on the watcher.
1247		1327
1248	=item ev_cb_set (ev_TYPE *watcher, callback)	1328	=item ev_set_cb (ev_TYPE *watcher, callback)
1249		1329
1250	Change the callback. You can change the callback at virtually any time	1330	Change the callback. You can change the callback at virtually any time
1251	(modulo threads).	1331	(modulo threads).
1252		1332
1253	=item ev_set_priority (ev_TYPE *watcher, int priority)	1333	=item ev_set_priority (ev_TYPE *watcher, int priority)
…		…
1271	or might not have been clamped to the valid range.	1351	or might not have been clamped to the valid range.
1272		1352
1273	The default priority used by watchers when no priority has been set is	1353	The default priority used by watchers when no priority has been set is
1274	always C<0>, which is supposed to not be too high and not be too low :).	1354	always C<0>, which is supposed to not be too high and not be too low :).
1275		1355
1276	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of	1356	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1277	priorities.	1357	priorities.
1278		1358
1279	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1359	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
1280		1360
1281	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1361	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
…		…
1306	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related	1386	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1307	functions that do not need a watcher.	1387	functions that do not need a watcher.
1308		1388
1309	=back	1389	=back
1310		1390
1311	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1391	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
1312		1392	OWN COMPOSITE WATCHERS> idioms.
1313	Each watcher has, by default, a member C<void *data> that you can change
1314	and read at any time: libev will completely ignore it. This can be used
1315	to associate arbitrary data with your watcher. If you need more data and
1316	don't want to allocate memory and store a pointer to it in that data
1317	member, you can also "subclass" the watcher type and provide your own
1318	data:
1319
1320	struct my_io
1321	{
1322	ev_io io;
1323	int otherfd;
1324	void *somedata;
1325	struct whatever *mostinteresting;
1326	};
1327
1328	...
1329	struct my_io w;
1330	ev_io_init (&w.io, my_cb, fd, EV_READ);
1331
1332	And since your callback will be called with a pointer to the watcher, you
1333	can cast it back to your own type:
1334
1335	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
1336	{
1337	struct my_io *w = (struct my_io *)w_;
1338	...
1339	}
1340
1341	More interesting and less C-conformant ways of casting your callback type
1342	instead have been omitted.
1343
1344	Another common scenario is to use some data structure with multiple
1345	embedded watchers:
1346
1347	struct my_biggy
1348	{
1349	int some_data;
1350	ev_timer t1;
1351	ev_timer t2;
1352	}
1353
1354	In this case getting the pointer to C<my_biggy> is a bit more
1355	complicated: Either you store the address of your C<my_biggy> struct
1356	in the C<data> member of the watcher (for woozies), or you need to use
1357	some pointer arithmetic using C<offsetof> inside your watchers (for real
1358	programmers):
1359
1360	#include <stddef.h>
1361
1362	static void
1363	t1_cb (EV_P_ ev_timer *w, int revents)
1364	{
1365	struct my_biggy big = (struct my_biggy *)
1366	(((char *)w) - offsetof (struct my_biggy, t1));
1367	}
1368
1369	static void
1370	t2_cb (EV_P_ ev_timer *w, int revents)
1371	{
1372	struct my_biggy big = (struct my_biggy *)
1373	(((char *)w) - offsetof (struct my_biggy, t2));
1374	}
1375		1393
1376	=head2 WATCHER STATES	1394	=head2 WATCHER STATES
1377		1395
1378	There are various watcher states mentioned throughout this manual -	1396	There are various watcher states mentioned throughout this manual -
1379	active, pending and so on. In this section these states and the rules to	1397	active, pending and so on. In this section these states and the rules to
1380	transition between them will be described in more detail - and while these	1398	transition between them will be described in more detail - and while these
1381	rules might look complicated, they usually do "the right thing".	1399	rules might look complicated, they usually do "the right thing".
1382		1400
1383	=over 4	1401	=over 4
1384		1402
1385	=item initialiased	1403	=item initialised
1386		1404
1387	Before a watcher can be registered with the event looop it has to be	1405	Before a watcher can be registered with the event loop it has to be
1388	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to	1406	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1389	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.	1407	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1390		1408
1391	In this state it is simply some block of memory that is suitable for use	1409	In this state it is simply some block of memory that is suitable for
1392	in an event loop. It can be moved around, freed, reused etc. at will.	1410	use in an event loop. It can be moved around, freed, reused etc. at
		1411	will - as long as you either keep the memory contents intact, or call
		1412	C<ev_TYPE_init> again.
1393		1413
1394	=item started/running/active	1414	=item started/running/active
1395		1415
1396	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes	1416	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1397	property of the event loop, and is actively waiting for events. While in	1417	property of the event loop, and is actively waiting for events. While in
…		…
1425	latter will clear any pending state the watcher might be in, regardless	1445	latter will clear any pending state the watcher might be in, regardless
1426	of whether it was active or not, so stopping a watcher explicitly before	1446	of whether it was active or not, so stopping a watcher explicitly before
1427	freeing it is often a good idea.	1447	freeing it is often a good idea.
1428		1448
1429	While stopped (and not pending) the watcher is essentially in the	1449	While stopped (and not pending) the watcher is essentially in the
1430	initialised state, that is it can be reused, moved, modified in any way	1450	initialised state, that is, it can be reused, moved, modified in any way
1431	you wish.	1451	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1452	it again).
1432		1453
1433	=back	1454	=back
1434		1455
1435	=head2 WATCHER PRIORITY MODELS	1456	=head2 WATCHER PRIORITY MODELS
1436		1457
…		…
1565	In general you can register as many read and/or write event watchers per	1586	In general you can register as many read and/or write event watchers per
1566	fd as you want (as long as you don't confuse yourself). Setting all file	1587	fd as you want (as long as you don't confuse yourself). Setting all file
1567	descriptors to non-blocking mode is also usually a good idea (but not	1588	descriptors to non-blocking mode is also usually a good idea (but not
1568	required if you know what you are doing).	1589	required if you know what you are doing).
1569		1590
1570	If you cannot use non-blocking mode, then force the use of a
1571	known-to-be-good backend (at the time of this writing, this includes only
1572	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
1573	descriptors for which non-blocking operation makes no sense (such as
1574	files) - libev doesn't guarantee any specific behaviour in that case.
1575
1576	Another thing you have to watch out for is that it is quite easy to	1591	Another thing you have to watch out for is that it is quite easy to
1577	receive "spurious" readiness notifications, that is your callback might	1592	receive "spurious" readiness notifications, that is, your callback might
1578	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1593	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1579	because there is no data. Not only are some backends known to create a	1594	because there is no data. It is very easy to get into this situation even
1580	lot of those (for example Solaris ports), it is very easy to get into	1595	with a relatively standard program structure. Thus it is best to always
1581	this situation even with a relatively standard program structure. Thus	1596	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1582	it is best to always use non-blocking I/O: An extra C<read>(2) returning
1583	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1597	preferable to a program hanging until some data arrives.
1584		1598
1585	If you cannot run the fd in non-blocking mode (for example you should	1599	If you cannot run the fd in non-blocking mode (for example you should
1586	not play around with an Xlib connection), then you have to separately	1600	not play around with an Xlib connection), then you have to separately
1587	re-test whether a file descriptor is really ready with a known-to-be good	1601	re-test whether a file descriptor is really ready with a known-to-be good
1588	interface such as poll (fortunately in our Xlib example, Xlib already	1602	interface such as poll (fortunately in the case of Xlib, it already does
1589	does this on its own, so its quite safe to use). Some people additionally	1603	this on its own, so its quite safe to use). Some people additionally
1590	use C<SIGALRM> and an interval timer, just to be sure you won't block	1604	use C<SIGALRM> and an interval timer, just to be sure you won't block
1591	indefinitely.	1605	indefinitely.
1592		1606
1593	But really, best use non-blocking mode.	1607	But really, best use non-blocking mode.
1594		1608
…		…
1622		1636
1623	There is no workaround possible except not registering events	1637	There is no workaround possible except not registering events
1624	for potentially C<dup ()>'ed file descriptors, or to resort to	1638	for potentially C<dup ()>'ed file descriptors, or to resort to
1625	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1639	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1626		1640
		1641	=head3 The special problem of files
		1642
		1643	Many people try to use C<select> (or libev) on file descriptors
		1644	representing files, and expect it to become ready when their program
		1645	doesn't block on disk accesses (which can take a long time on their own).
		1646
		1647	However, this cannot ever work in the "expected" way - you get a readiness
		1648	notification as soon as the kernel knows whether and how much data is
		1649	there, and in the case of open files, that's always the case, so you
		1650	always get a readiness notification instantly, and your read (or possibly
		1651	write) will still block on the disk I/O.
		1652
		1653	Another way to view it is that in the case of sockets, pipes, character
		1654	devices and so on, there is another party (the sender) that delivers data
		1655	on its own, but in the case of files, there is no such thing: the disk
		1656	will not send data on its own, simply because it doesn't know what you
		1657	wish to read - you would first have to request some data.
		1658
		1659	Since files are typically not-so-well supported by advanced notification
		1660	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1661	to files, even though you should not use it. The reason for this is
		1662	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1663	usually a tty, often a pipe, but also sometimes files or special devices
		1664	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1665	F</dev/urandom>), and even though the file might better be served with
		1666	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1667	it "just works" instead of freezing.
		1668
		1669	So avoid file descriptors pointing to files when you know it (e.g. use
		1670	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1671	when you rarely read from a file instead of from a socket, and want to
		1672	reuse the same code path.
		1673
1627	=head3 The special problem of fork	1674	=head3 The special problem of fork
1628		1675
1629	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1676	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1630	useless behaviour. Libev fully supports fork, but needs to be told about	1677	useless behaviour. Libev fully supports fork, but needs to be told about
1631	it in the child.	1678	it in the child if you want to continue to use it in the child.
1632		1679
1633	To support fork in your programs, you either have to call	1680	To support fork in your child processes, you have to call C<ev_loop_fork
1634	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,	1681	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1635	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1682	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1636	C<EVBACKEND_POLL>.
1637		1683
1638	=head3 The special problem of SIGPIPE	1684	=head3 The special problem of SIGPIPE
1639		1685
1640	While not really specific to libev, it is easy to forget about C<SIGPIPE>:	1686	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1641	when writing to a pipe whose other end has been closed, your program gets	1687	when writing to a pipe whose other end has been closed, your program gets
…		…
1739	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1785	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1740	monotonic clock option helps a lot here).	1786	monotonic clock option helps a lot here).
1741		1787
1742	The callback is guaranteed to be invoked only I<after> its timeout has	1788	The callback is guaranteed to be invoked only I<after> its timeout has
1743	passed (not I<at>, so on systems with very low-resolution clocks this	1789	passed (not I<at>, so on systems with very low-resolution clocks this
1744	might introduce a small delay). If multiple timers become ready during the	1790	might introduce a small delay, see "the special problem of being too
		1791	early", below). If multiple timers become ready during the same loop
1745	same loop iteration then the ones with earlier time-out values are invoked	1792	iteration then the ones with earlier time-out values are invoked before
1746	before ones of the same priority with later time-out values (but this is	1793	ones of the same priority with later time-out values (but this is no
1747	no longer true when a callback calls C<ev_run> recursively).	1794	longer true when a callback calls C<ev_run> recursively).
1748		1795
1749	=head3 Be smart about timeouts	1796	=head3 Be smart about timeouts
1750		1797
1751	Many real-world problems involve some kind of timeout, usually for error	1798	Many real-world problems involve some kind of timeout, usually for error
1752	recovery. A typical example is an HTTP request - if the other side hangs,	1799	recovery. A typical example is an HTTP request - if the other side hangs,
…		…
1827		1874
1828	In this case, it would be more efficient to leave the C<ev_timer> alone,	1875	In this case, it would be more efficient to leave the C<ev_timer> alone,
1829	but remember the time of last activity, and check for a real timeout only	1876	but remember the time of last activity, and check for a real timeout only
1830	within the callback:	1877	within the callback:
1831		1878
		1879	ev_tstamp timeout = 60.;
1832	ev_tstamp last_activity; // time of last activity	1880	ev_tstamp last_activity; // time of last activity
		1881	ev_timer timer;
1833		1882
1834	static void	1883	static void
1835	callback (EV_P_ ev_timer *w, int revents)	1884	callback (EV_P_ ev_timer *w, int revents)
1836	{	1885	{
1837	ev_tstamp now = ev_now (EV_A);	1886	// calculate when the timeout would happen
1838	ev_tstamp timeout = last_activity + 60.;	1887	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
1839		1888
1840	// if last_activity + 60. is older than now, we did time out	1889	// if negative, it means we the timeout already occurred
1841	if (timeout < now)	1890	if (after < 0.)
1842	{	1891	{
1843	// timeout occurred, take action	1892	// timeout occurred, take action
1844	}	1893	}
1845	else	1894	else
1846	{	1895	{
1847	// callback was invoked, but there was some activity, re-arm	1896	// callback was invoked, but there was some recent
1848	// the watcher to fire in last_activity + 60, which is	1897	// activity. simply restart the timer to time out
1849	// guaranteed to be in the future, so "again" is positive:	1898	// after "after" seconds, which is the earliest time
1850	w->repeat = timeout - now;	1899	// the timeout can occur.
		1900	ev_timer_set (w, after, 0.);
1851	ev_timer_again (EV_A_ w);	1901	ev_timer_start (EV_A_ w);
1852	}	1902	}
1853	}	1903	}
1854		1904
1855	To summarise the callback: first calculate the real timeout (defined	1905	To summarise the callback: first calculate in how many seconds the
1856	as "60 seconds after the last activity"), then check if that time has	1906	timeout will occur (by calculating the absolute time when it would occur,
1857	been reached, which means something I<did>, in fact, time out. Otherwise	1907	C<last_activity + timeout>, and subtracting the current time, C<ev_now
1858	the callback was invoked too early (C<timeout> is in the future), so	1908	(EV_A)> from that).
1859	re-schedule the timer to fire at that future time, to see if maybe we have
1860	a timeout then.
1861		1909
1862	Note how C<ev_timer_again> is used, taking advantage of the	1910	If this value is negative, then we are already past the timeout, i.e. we
1863	C<ev_timer_again> optimisation when the timer is already running.	1911	timed out, and need to do whatever is needed in this case.
		1912
		1913	Otherwise, we now the earliest time at which the timeout would trigger,
		1914	and simply start the timer with this timeout value.
		1915
		1916	In other words, each time the callback is invoked it will check whether
		1917	the timeout occurred. If not, it will simply reschedule itself to check
		1918	again at the earliest time it could time out. Rinse. Repeat.
1864		1919
1865	This scheme causes more callback invocations (about one every 60 seconds	1920	This scheme causes more callback invocations (about one every 60 seconds
1866	minus half the average time between activity), but virtually no calls to	1921	minus half the average time between activity), but virtually no calls to
1867	libev to change the timeout.	1922	libev to change the timeout.
1868		1923
1869	To start the timer, simply initialise the watcher and set C<last_activity>	1924	To start the machinery, simply initialise the watcher and set
1870	to the current time (meaning we just have some activity :), then call the	1925	C<last_activity> to the current time (meaning there was some activity just
1871	callback, which will "do the right thing" and start the timer:	1926	now), then call the callback, which will "do the right thing" and start
		1927	the timer:
1872		1928
		1929	last_activity = ev_now (EV_A);
1873	ev_init (timer, callback);	1930	ev_init (&timer, callback);
1874	last_activity = ev_now (loop);	1931	callback (EV_A_ &timer, 0);
1875	callback (loop, timer, EV_TIMER);
1876		1932
1877	And when there is some activity, simply store the current time in	1933	When there is some activity, simply store the current time in
1878	C<last_activity>, no libev calls at all:	1934	C<last_activity>, no libev calls at all:
1879		1935
		1936	if (activity detected)
1880	last_activity = ev_now (loop);	1937	last_activity = ev_now (EV_A);
		1938
		1939	When your timeout value changes, then the timeout can be changed by simply
		1940	providing a new value, stopping the timer and calling the callback, which
		1941	will again do the right thing (for example, time out immediately :).
		1942
		1943	timeout = new_value;
		1944	ev_timer_stop (EV_A_ &timer);
		1945	callback (EV_A_ &timer, 0);
1881		1946
1882	This technique is slightly more complex, but in most cases where the	1947	This technique is slightly more complex, but in most cases where the
1883	time-out is unlikely to be triggered, much more efficient.	1948	time-out is unlikely to be triggered, much more efficient.
1884
1885	Changing the timeout is trivial as well (if it isn't hard-coded in the
1886	callback :) - just change the timeout and invoke the callback, which will
1887	fix things for you.
1888		1949
1889	=item 4. Wee, just use a double-linked list for your timeouts.	1950	=item 4. Wee, just use a double-linked list for your timeouts.
1890		1951
1891	If there is not one request, but many thousands (millions...), all	1952	If there is not one request, but many thousands (millions...), all
1892	employing some kind of timeout with the same timeout value, then one can	1953	employing some kind of timeout with the same timeout value, then one can
…		…
1919	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is	1980	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1920	rather complicated, but extremely efficient, something that really pays	1981	rather complicated, but extremely efficient, something that really pays
1921	off after the first million or so of active timers, i.e. it's usually	1982	off after the first million or so of active timers, i.e. it's usually
1922	overkill :)	1983	overkill :)
1923		1984
		1985	=head3 The special problem of being too early
		1986
		1987	If you ask a timer to call your callback after three seconds, then
		1988	you expect it to be invoked after three seconds - but of course, this
		1989	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		1990	guaranteed to any precision by libev - imagine somebody suspending the
		1991	process with a STOP signal for a few hours for example.
		1992
		1993	So, libev tries to invoke your callback as soon as possible I<after> the
		1994	delay has occurred, but cannot guarantee this.
		1995
		1996	A less obvious failure mode is calling your callback too early: many event
		1997	loops compare timestamps with a "elapsed delay >= requested delay", but
		1998	this can cause your callback to be invoked much earlier than you would
		1999	expect.
		2000
		2001	To see why, imagine a system with a clock that only offers full second
		2002	resolution (think windows if you can't come up with a broken enough OS
		2003	yourself). If you schedule a one-second timer at the time 500.9, then the
		2004	event loop will schedule your timeout to elapse at a system time of 500
		2005	(500.9 truncated to the resolution) + 1, or 501.
		2006
		2007	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2008	501" and invoke the callback 0.1s after it was started, even though a
		2009	one-second delay was requested - this is being "too early", despite best
		2010	intentions.
		2011
		2012	This is the reason why libev will never invoke the callback if the elapsed
		2013	delay equals the requested delay, but only when the elapsed delay is
		2014	larger than the requested delay. In the example above, libev would only invoke
		2015	the callback at system time 502, or 1.1s after the timer was started.
		2016
		2017	So, while libev cannot guarantee that your callback will be invoked
		2018	exactly when requested, it I<can> and I<does> guarantee that the requested
		2019	delay has actually elapsed, or in other words, it always errs on the "too
		2020	late" side of things.
		2021
1924	=head3 The special problem of time updates	2022	=head3 The special problem of time updates
1925		2023
1926	Establishing the current time is a costly operation (it usually takes at	2024	Establishing the current time is a costly operation (it usually takes
1927	least two system calls): EV therefore updates its idea of the current	2025	at least one system call): EV therefore updates its idea of the current
1928	time only before and after C<ev_run> collects new events, which causes a	2026	time only before and after C<ev_run> collects new events, which causes a
1929	growing difference between C<ev_now ()> and C<ev_time ()> when handling	2027	growing difference between C<ev_now ()> and C<ev_time ()> when handling
1930	lots of events in one iteration.	2028	lots of events in one iteration.
1931		2029
1932	The relative timeouts are calculated relative to the C<ev_now ()>	2030	The relative timeouts are calculated relative to the C<ev_now ()>
1933	time. This is usually the right thing as this timestamp refers to the time	2031	time. This is usually the right thing as this timestamp refers to the time
1934	of the event triggering whatever timeout you are modifying/starting. If	2032	of the event triggering whatever timeout you are modifying/starting. If
1935	you suspect event processing to be delayed and you I<need> to base the	2033	you suspect event processing to be delayed and you I<need> to base the
1936	timeout on the current time, use something like this to adjust for this:	2034	timeout on the current time, use something like the following to adjust
		2035	for it:
1937		2036
1938	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2037	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
1939		2038
1940	If the event loop is suspended for a long time, you can also force an	2039	If the event loop is suspended for a long time, you can also force an
1941	update of the time returned by C<ev_now ()> by calling C<ev_now_update	2040	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1942	()>.	2041	()>, although that will push the event time of all outstanding events
		2042	further into the future.
		2043
		2044	=head3 The special problem of unsynchronised clocks
		2045
		2046	Modern systems have a variety of clocks - libev itself uses the normal
		2047	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2048	jumps).
		2049
		2050	Neither of these clocks is synchronised with each other or any other clock
		2051	on the system, so C<ev_time ()> might return a considerably different time
		2052	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2053	a call to C<gettimeofday> might return a second count that is one higher
		2054	than a directly following call to C<time>.
		2055
		2056	The moral of this is to only compare libev-related timestamps with
		2057	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2058	a second or so.
		2059
		2060	One more problem arises due to this lack of synchronisation: if libev uses
		2061	the system monotonic clock and you compare timestamps from C<ev_time>
		2062	or C<ev_now> from when you started your timer and when your callback is
		2063	invoked, you will find that sometimes the callback is a bit "early".
		2064
		2065	This is because C<ev_timer>s work in real time, not wall clock time, so
		2066	libev makes sure your callback is not invoked before the delay happened,
		2067	I<measured according to the real time>, not the system clock.
		2068
		2069	If your timeouts are based on a physical timescale (e.g. "time out this
		2070	connection after 100 seconds") then this shouldn't bother you as it is
		2071	exactly the right behaviour.
		2072
		2073	If you want to compare wall clock/system timestamps to your timers, then
		2074	you need to use C<ev_periodic>s, as these are based on the wall clock
		2075	time, where your comparisons will always generate correct results.
1943		2076
1944	=head3 The special problems of suspended animation	2077	=head3 The special problems of suspended animation
1945		2078
1946	When you leave the server world it is quite customary to hit machines that	2079	When you leave the server world it is quite customary to hit machines that
1947	can suspend/hibernate - what happens to the clocks during such a suspend?	2080	can suspend/hibernate - what happens to the clocks during such a suspend?
…		…
1991	keep up with the timer (because it takes longer than those 10 seconds to	2124	keep up with the timer (because it takes longer than those 10 seconds to
1992	do stuff) the timer will not fire more than once per event loop iteration.	2125	do stuff) the timer will not fire more than once per event loop iteration.
1993		2126
1994	=item ev_timer_again (loop, ev_timer *)	2127	=item ev_timer_again (loop, ev_timer *)
1995		2128
1996	This will act as if the timer timed out and restart it again if it is	2129	This will act as if the timer timed out, and restarts it again if it is
1997	repeating. The exact semantics are:	2130	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2131	timeout to the C<repeat> value and calling C<ev_timer_start>.
1998		2132
		2133	The exact semantics are as in the following rules, all of which will be
		2134	applied to the watcher:
		2135
		2136	=over 4
		2137
1999	If the timer is pending, its pending status is cleared.	2138	=item If the timer is pending, the pending status is always cleared.
2000		2139
2001	If the timer is started but non-repeating, stop it (as if it timed out).	2140	=item If the timer is started but non-repeating, stop it (as if it timed
		2141	out, without invoking it).
2002		2142
2003	If the timer is repeating, either start it if necessary (with the	2143	=item If the timer is repeating, make the C<repeat> value the new timeout
2004	C<repeat> value), or reset the running timer to the C<repeat> value.	2144	and start the timer, if necessary.
2005		2145
		2146	=back
		2147
2006	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a	2148	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
2007	usage example.	2149	usage example.
2008		2150
2009	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)	2151	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
2010		2152
2011	Returns the remaining time until a timer fires. If the timer is active,	2153	Returns the remaining time until a timer fires. If the timer is active,
…		…
2131		2273
2132	Another way to think about it (for the mathematically inclined) is that	2274	Another way to think about it (for the mathematically inclined) is that
2133	C<ev_periodic> will try to run the callback in this mode at the next possible	2275	C<ev_periodic> will try to run the callback in this mode at the next possible
2134	time where C<time = offset (mod interval)>, regardless of any time jumps.	2276	time where C<time = offset (mod interval)>, regardless of any time jumps.
2135		2277
2136	For numerical stability it is preferable that the C<offset> value is near	2278	The C<interval> I<MUST> be positive, and for numerical stability, the
2137	C<ev_now ()> (the current time), but there is no range requirement for	2279	interval value should be higher than C<1/8192> (which is around 100
2138	this value, and in fact is often specified as zero.	2280	microseconds) and C<offset> should be higher than C<0> and should have
		2281	at most a similar magnitude as the current time (say, within a factor of
		2282	ten). Typical values for offset are, in fact, C<0> or something between
		2283	C<0> and C<interval>, which is also the recommended range.
2139		2284
2140	Note also that there is an upper limit to how often a timer can fire (CPU	2285	Note also that there is an upper limit to how often a timer can fire (CPU
2141	speed for example), so if C<interval> is very small then timing stability	2286	speed for example), so if C<interval> is very small then timing stability
2142	will of course deteriorate. Libev itself tries to be exact to be about one	2287	will of course deteriorate. Libev itself tries to be exact to be about one
2143	millisecond (if the OS supports it and the machine is fast enough).	2288	millisecond (if the OS supports it and the machine is fast enough).
…		…
2251		2396
2252	ev_periodic hourly_tick;	2397	ev_periodic hourly_tick;
2253	ev_periodic_init (&hourly_tick, clock_cb,	2398	ev_periodic_init (&hourly_tick, clock_cb,
2254	fmod (ev_now (loop), 3600.), 3600., 0);	2399	fmod (ev_now (loop), 3600.), 3600., 0);
2255	ev_periodic_start (loop, &hourly_tick);	2400	ev_periodic_start (loop, &hourly_tick);
2256		2401
2257		2402
2258	=head2 C<ev_signal> - signal me when a signal gets signalled!	2403	=head2 C<ev_signal> - signal me when a signal gets signalled!
2259		2404
2260	Signal watchers will trigger an event when the process receives a specific	2405	Signal watchers will trigger an event when the process receives a specific
2261	signal one or more times. Even though signals are very asynchronous, libev	2406	signal one or more times. Even though signals are very asynchronous, libev
2262	will try it's best to deliver signals synchronously, i.e. as part of the	2407	will try its best to deliver signals synchronously, i.e. as part of the
2263	normal event processing, like any other event.	2408	normal event processing, like any other event.
2264		2409
2265	If you want signals to be delivered truly asynchronously, just use	2410	If you want signals to be delivered truly asynchronously, just use
2266	C<sigaction> as you would do without libev and forget about sharing	2411	C<sigaction> as you would do without libev and forget about sharing
2267	the signal. You can even use C<ev_async> from a signal handler to	2412	the signal. You can even use C<ev_async> from a signal handler to
…		…
2271	only within the same loop, i.e. you can watch for C<SIGINT> in your	2416	only within the same loop, i.e. you can watch for C<SIGINT> in your
2272	default loop and for C<SIGIO> in another loop, but you cannot watch for	2417	default loop and for C<SIGIO> in another loop, but you cannot watch for
2273	C<SIGINT> in both the default loop and another loop at the same time. At	2418	C<SIGINT> in both the default loop and another loop at the same time. At
2274	the moment, C<SIGCHLD> is permanently tied to the default loop.	2419	the moment, C<SIGCHLD> is permanently tied to the default loop.
2275		2420
2276	When the first watcher gets started will libev actually register something	2421	Only after the first watcher for a signal is started will libev actually
2277	with the kernel (thus it coexists with your own signal handlers as long as	2422	register something with the kernel. It thus coexists with your own signal
2278	you don't register any with libev for the same signal).	2423	handlers as long as you don't register any with libev for the same signal.
2279		2424
2280	If possible and supported, libev will install its handlers with	2425	If possible and supported, libev will install its handlers with
2281	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should	2426	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
2282	not be unduly interrupted. If you have a problem with system calls getting	2427	not be unduly interrupted. If you have a problem with system calls getting
2283	interrupted by signals you can block all signals in an C<ev_check> watcher	2428	interrupted by signals you can block all signals in an C<ev_check> watcher
…		…
2286	=head3 The special problem of inheritance over fork/execve/pthread_create	2431	=head3 The special problem of inheritance over fork/execve/pthread_create
2287		2432
2288	Both the signal mask (C<sigprocmask>) and the signal disposition	2433	Both the signal mask (C<sigprocmask>) and the signal disposition
2289	(C<sigaction>) are unspecified after starting a signal watcher (and after	2434	(C<sigaction>) are unspecified after starting a signal watcher (and after
2290	stopping it again), that is, libev might or might not block the signal,	2435	stopping it again), that is, libev might or might not block the signal,
2291	and might or might not set or restore the installed signal handler.	2436	and might or might not set or restore the installed signal handler (but
		2437	see C<EVFLAG_NOSIGMASK>).
2292		2438
2293	While this does not matter for the signal disposition (libev never	2439	While this does not matter for the signal disposition (libev never
2294	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on	2440	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2295	C<execve>), this matters for the signal mask: many programs do not expect	2441	C<execve>), this matters for the signal mask: many programs do not expect
2296	certain signals to be blocked.	2442	certain signals to be blocked.
…		…
2309	I<has> to modify the signal mask, at least temporarily.	2455	I<has> to modify the signal mask, at least temporarily.
2310		2456
2311	So I can't stress this enough: I<If you do not reset your signal mask when	2457	So I can't stress this enough: I<If you do not reset your signal mask when
2312	you expect it to be empty, you have a race condition in your code>. This	2458	you expect it to be empty, you have a race condition in your code>. This
2313	is not a libev-specific thing, this is true for most event libraries.	2459	is not a libev-specific thing, this is true for most event libraries.
		2460
		2461	=head3 The special problem of threads signal handling
		2462
		2463	POSIX threads has problematic signal handling semantics, specifically,
		2464	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2465	threads in a process block signals, which is hard to achieve.
		2466
		2467	When you want to use sigwait (or mix libev signal handling with your own
		2468	for the same signals), you can tackle this problem by globally blocking
		2469	all signals before creating any threads (or creating them with a fully set
		2470	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2471	loops. Then designate one thread as "signal receiver thread" which handles
		2472	these signals. You can pass on any signals that libev might be interested
		2473	in by calling C<ev_feed_signal>.
2314		2474
2315	=head3 Watcher-Specific Functions and Data Members	2475	=head3 Watcher-Specific Functions and Data Members
2316		2476
2317	=over 4	2477	=over 4
2318		2478
…		…
2453		2613
2454	=head2 C<ev_stat> - did the file attributes just change?	2614	=head2 C<ev_stat> - did the file attributes just change?
2455		2615
2456	This watches a file system path for attribute changes. That is, it calls	2616	This watches a file system path for attribute changes. That is, it calls
2457	C<stat> on that path in regular intervals (or when the OS says it changed)	2617	C<stat> on that path in regular intervals (or when the OS says it changed)
2458	and sees if it changed compared to the last time, invoking the callback if	2618	and sees if it changed compared to the last time, invoking the callback
2459	it did.	2619	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2620	happen after the watcher has been started will be reported.
2460		2621
2461	The path does not need to exist: changing from "path exists" to "path does	2622	The path does not need to exist: changing from "path exists" to "path does
2462	not exist" is a status change like any other. The condition "path does not	2623	not exist" is a status change like any other. The condition "path does not
2463	exist" (or more correctly "path cannot be stat'ed") is signified by the	2624	exist" (or more correctly "path cannot be stat'ed") is signified by the
2464	C<st_nlink> field being zero (which is otherwise always forced to be at	2625	C<st_nlink> field being zero (which is otherwise always forced to be at
…		…
2694	Apart from keeping your process non-blocking (which is a useful	2855	Apart from keeping your process non-blocking (which is a useful
2695	effect on its own sometimes), idle watchers are a good place to do	2856	effect on its own sometimes), idle watchers are a good place to do
2696	"pseudo-background processing", or delay processing stuff to after the	2857	"pseudo-background processing", or delay processing stuff to after the
2697	event loop has handled all outstanding events.	2858	event loop has handled all outstanding events.
2698		2859
		2860	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2861
		2862	As long as there is at least one active idle watcher, libev will never
		2863	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2864	For this to work, the idle watcher doesn't need to be invoked at all - the
		2865	lowest priority will do.
		2866
		2867	This mode of operation can be useful together with an C<ev_check> watcher,
		2868	to do something on each event loop iteration - for example to balance load
		2869	between different connections.
		2870
		2871	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2872	example.
		2873
2699	=head3 Watcher-Specific Functions and Data Members	2874	=head3 Watcher-Specific Functions and Data Members
2700		2875
2701	=over 4	2876	=over 4
2702		2877
2703	=item ev_idle_init (ev_idle *, callback)	2878	=item ev_idle_init (ev_idle *, callback)
…		…
2714	callback, free it. Also, use no error checking, as usual.	2889	callback, free it. Also, use no error checking, as usual.
2715		2890
2716	static void	2891	static void
2717	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)	2892	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
2718	{	2893	{
		2894	// stop the watcher
		2895	ev_idle_stop (loop, w);
		2896
		2897	// now we can free it
2719	free (w);	2898	free (w);
		2899
2720	// now do something you wanted to do when the program has	2900	// now do something you wanted to do when the program has
2721	// no longer anything immediate to do.	2901	// no longer anything immediate to do.
2722	}	2902	}
2723		2903
2724	ev_idle *idle_watcher = malloc (sizeof (ev_idle));	2904	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
…		…
2726	ev_idle_start (loop, idle_watcher);	2906	ev_idle_start (loop, idle_watcher);
2727		2907
2728		2908
2729	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	2909	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2730		2910
2731	Prepare and check watchers are usually (but not always) used in pairs:	2911	Prepare and check watchers are often (but not always) used in pairs:
2732	prepare watchers get invoked before the process blocks and check watchers	2912	prepare watchers get invoked before the process blocks and check watchers
2733	afterwards.	2913	afterwards.
2734		2914
2735	You I<must not> call C<ev_run> or similar functions that enter	2915	You I<must not> call C<ev_run> (or similar functions that enter the
2736	the current event loop from either C<ev_prepare> or C<ev_check>	2916	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
2737	watchers. Other loops than the current one are fine, however. The	2917	C<ev_check> watchers. Other loops than the current one are fine,
2738	rationale behind this is that you do not need to check for recursion in	2918	however. The rationale behind this is that you do not need to check
2739	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,	2919	for recursion in those watchers, i.e. the sequence will always be
2740	C<ev_check> so if you have one watcher of each kind they will always be	2920	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
2741	called in pairs bracketing the blocking call.	2921	kind they will always be called in pairs bracketing the blocking call.
2742		2922
2743	Their main purpose is to integrate other event mechanisms into libev and	2923	Their main purpose is to integrate other event mechanisms into libev and
2744	their use is somewhat advanced. They could be used, for example, to track	2924	their use is somewhat advanced. They could be used, for example, to track
2745	variable changes, implement your own watchers, integrate net-snmp or a	2925	variable changes, implement your own watchers, integrate net-snmp or a
2746	coroutine library and lots more. They are also occasionally useful if	2926	coroutine library and lots more. They are also occasionally useful if
…		…
2764	with priority higher than or equal to the event loop and one coroutine	2944	with priority higher than or equal to the event loop and one coroutine
2765	of lower priority, but only once, using idle watchers to keep the event	2945	of lower priority, but only once, using idle watchers to keep the event
2766	loop from blocking if lower-priority coroutines are active, thus mapping	2946	loop from blocking if lower-priority coroutines are active, thus mapping
2767	low-priority coroutines to idle/background tasks).	2947	low-priority coroutines to idle/background tasks).
2768		2948
2769	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	2949	When used for this purpose, it is recommended to give C<ev_check> watchers
2770	priority, to ensure that they are being run before any other watchers	2950	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
2771	after the poll (this doesn't matter for C<ev_prepare> watchers).	2951	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		2952	watchers).
2772		2953
2773	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not	2954	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
2774	activate ("feed") events into libev. While libev fully supports this, they	2955	activate ("feed") events into libev. While libev fully supports this, they
2775	might get executed before other C<ev_check> watchers did their job. As	2956	might get executed before other C<ev_check> watchers did their job. As
2776	C<ev_check> watchers are often used to embed other (non-libev) event	2957	C<ev_check> watchers are often used to embed other (non-libev) event
2777	loops those other event loops might be in an unusable state until their	2958	loops those other event loops might be in an unusable state until their
2778	C<ev_check> watcher ran (always remind yourself to coexist peacefully with	2959	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
2779	others).	2960	others).
		2961
		2962	=head3 Abusing an C<ev_check> watcher for its side-effect
		2963
		2964	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		2965	useful because they are called once per event loop iteration. For
		2966	example, if you want to handle a large number of connections fairly, you
		2967	normally only do a bit of work for each active connection, and if there
		2968	is more work to do, you wait for the next event loop iteration, so other
		2969	connections have a chance of making progress.
		2970
		2971	Using an C<ev_check> watcher is almost enough: it will be called on the
		2972	next event loop iteration. However, that isn't as soon as possible -
		2973	without external events, your C<ev_check> watcher will not be invoked.
		2974
		2975	This is where C<ev_idle> watchers come in handy - all you need is a
		2976	single global idle watcher that is active as long as you have one active
		2977	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		2978	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		2979	invoked. Neither watcher alone can do that.
2780		2980
2781	=head3 Watcher-Specific Functions and Data Members	2981	=head3 Watcher-Specific Functions and Data Members
2782		2982
2783	=over 4	2983	=over 4
2784		2984
…		…
2985		3185
2986	=over 4	3186	=over 4
2987		3187
2988	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)	3188	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
2989		3189
2990	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)	3190	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
2991		3191
2992	Configures the watcher to embed the given loop, which must be	3192	Configures the watcher to embed the given loop, which must be
2993	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be	3193	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
2994	invoked automatically, otherwise it is the responsibility of the callback	3194	invoked automatically, otherwise it is the responsibility of the callback
2995	to invoke it (it will continue to be called until the sweep has been done,	3195	to invoke it (it will continue to be called until the sweep has been done,
…		…
3016	used).	3216	used).
3017		3217
3018	struct ev_loop *loop_hi = ev_default_init (0);	3218	struct ev_loop *loop_hi = ev_default_init (0);
3019	struct ev_loop *loop_lo = 0;	3219	struct ev_loop *loop_lo = 0;
3020	ev_embed embed;	3220	ev_embed embed;
3021		3221
3022	// see if there is a chance of getting one that works	3222	// see if there is a chance of getting one that works
3023	// (remember that a flags value of 0 means autodetection)	3223	// (remember that a flags value of 0 means autodetection)
3024	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()	3224	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
3025	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())	3225	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
3026	: 0;	3226	: 0;
…		…
3040	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).	3240	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
3041		3241
3042	struct ev_loop *loop = ev_default_init (0);	3242	struct ev_loop *loop = ev_default_init (0);
3043	struct ev_loop *loop_socket = 0;	3243	struct ev_loop *loop_socket = 0;
3044	ev_embed embed;	3244	ev_embed embed;
3045		3245
3046	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)	3246	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
3047	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))	3247	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
3048	{	3248	{
3049	ev_embed_init (&embed, 0, loop_socket);	3249	ev_embed_init (&embed, 0, loop_socket);
3050	ev_embed_start (loop, &embed);	3250	ev_embed_start (loop, &embed);
…		…
3058		3258
3059	=head2 C<ev_fork> - the audacity to resume the event loop after a fork	3259	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
3060		3260
3061	Fork watchers are called when a C<fork ()> was detected (usually because	3261	Fork watchers are called when a C<fork ()> was detected (usually because
3062	whoever is a good citizen cared to tell libev about it by calling	3262	whoever is a good citizen cared to tell libev about it by calling
3063	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the	3263	C<ev_loop_fork>). The invocation is done before the event loop blocks next
3064	event loop blocks next and before C<ev_check> watchers are being called,	3264	and before C<ev_check> watchers are being called, and only in the child
3065	and only in the child after the fork. If whoever good citizen calling	3265	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
3066	C<ev_default_fork> cheats and calls it in the wrong process, the fork	3266	and calls it in the wrong process, the fork handlers will be invoked, too,
3067	handlers will be invoked, too, of course.	3267	of course.
3068		3268
3069	=head3 The special problem of life after fork - how is it possible?	3269	=head3 The special problem of life after fork - how is it possible?
3070		3270
3071	Most uses of C<fork()> consist of forking, then some simple calls to set	3271	Most uses of C<fork ()> consist of forking, then some simple calls to set
3072	up/change the process environment, followed by a call to C<exec()>. This	3272	up/change the process environment, followed by a call to C<exec()>. This
3073	sequence should be handled by libev without any problems.	3273	sequence should be handled by libev without any problems.
3074		3274
3075	This changes when the application actually wants to do event handling	3275	This changes when the application actually wants to do event handling
3076	in the child, or both parent in child, in effect "continuing" after the	3276	in the child, or both parent in child, in effect "continuing" after the
…		…
3153	atexit (program_exits);	3353	atexit (program_exits);
3154		3354
3155		3355
3156	=head2 C<ev_async> - how to wake up an event loop	3356	=head2 C<ev_async> - how to wake up an event loop
3157		3357
3158	In general, you cannot use an C<ev_run> from multiple threads or other	3358	In general, you cannot use an C<ev_loop> from multiple threads or other
3159	asynchronous sources such as signal handlers (as opposed to multiple event	3359	asynchronous sources such as signal handlers (as opposed to multiple event
3160	loops - those are of course safe to use in different threads).	3360	loops - those are of course safe to use in different threads).
3161		3361
3162	Sometimes, however, you need to wake up an event loop you do not control,	3362	Sometimes, however, you need to wake up an event loop you do not control,
3163	for example because it belongs to another thread. This is what C<ev_async>	3363	for example because it belongs to another thread. This is what C<ev_async>
…		…
3165	it by calling C<ev_async_send>, which is thread- and signal safe.	3365	it by calling C<ev_async_send>, which is thread- and signal safe.
3166		3366
3167	This functionality is very similar to C<ev_signal> watchers, as signals,	3367	This functionality is very similar to C<ev_signal> watchers, as signals,
3168	too, are asynchronous in nature, and signals, too, will be compressed	3368	too, are asynchronous in nature, and signals, too, will be compressed
3169	(i.e. the number of callback invocations may be less than the number of	3369	(i.e. the number of callback invocations may be less than the number of
3170	C<ev_async_sent> calls).	3370	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
3171		3371	of "global async watchers" by using a watcher on an otherwise unused
3172	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not	3372	signal, and C<ev_feed_signal> to signal this watcher from another thread,
3173	just the default loop.	3373	even without knowing which loop owns the signal.
3174		3374
3175	=head3 Queueing	3375	=head3 Queueing
3176		3376
3177	C<ev_async> does not support queueing of data in any way. The reason	3377	C<ev_async> does not support queueing of data in any way. The reason
3178	is that the author does not know of a simple (or any) algorithm for a	3378	is that the author does not know of a simple (or any) algorithm for a
…		…
3270	trust me.	3470	trust me.
3271		3471
3272	=item ev_async_send (loop, ev_async *)	3472	=item ev_async_send (loop, ev_async *)
3273		3473
3274	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3474	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3275	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	3475	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3476	returns.
		3477
3276	C<ev_feed_event>, this call is safe to do from other threads, signal or	3478	Unlike C<ev_feed_event>, this call is safe to do from other threads,
3277	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	3479	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3278	section below on what exactly this means).	3480	embedding section below on what exactly this means).
3279		3481
3280	Note that, as with other watchers in libev, multiple events might get	3482	Note that, as with other watchers in libev, multiple events might get
3281	compressed into a single callback invocation (another way to look at this	3483	compressed into a single callback invocation (another way to look at
3282	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,	3484	this is that C<ev_async> watchers are level-triggered: they are set on
3283	reset when the event loop detects that).	3485	C<ev_async_send>, reset when the event loop detects that).
3284		3486
3285	This call incurs the overhead of a system call only once per event loop	3487	This call incurs the overhead of at most one extra system call per event
3286	iteration, so while the overhead might be noticeable, it doesn't apply to	3488	loop iteration, if the event loop is blocked, and no syscall at all if
3287	repeated calls to C<ev_async_send> for the same event loop.	3489	the event loop (or your program) is processing events. That means that
		3490	repeated calls are basically free (there is no need to avoid calls for
		3491	performance reasons) and that the overhead becomes smaller (typically
		3492	zero) under load.
3288		3493
3289	=item bool = ev_async_pending (ev_async *)	3494	=item bool = ev_async_pending (ev_async *)
3290		3495
3291	Returns a non-zero value when C<ev_async_send> has been called on the	3496	Returns a non-zero value when C<ev_async_send> has been called on the
3292	watcher but the event has not yet been processed (or even noted) by the	3497	watcher but the event has not yet been processed (or even noted) by the
…		…
3347	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3552	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3348		3553
3349	=item ev_feed_fd_event (loop, int fd, int revents)	3554	=item ev_feed_fd_event (loop, int fd, int revents)
3350		3555
3351	Feed an event on the given fd, as if a file descriptor backend detected	3556	Feed an event on the given fd, as if a file descriptor backend detected
3352	the given events it.	3557	the given events.
3353		3558
3354	=item ev_feed_signal_event (loop, int signum)	3559	=item ev_feed_signal_event (loop, int signum)
3355		3560
3356	Feed an event as if the given signal occurred (C<loop> must be the default	3561	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3357	loop!).	3562	which is async-safe.
3358		3563
3359	=back	3564	=back
		3565
		3566
		3567	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3568
		3569	This section explains some common idioms that are not immediately
		3570	obvious. Note that examples are sprinkled over the whole manual, and this
		3571	section only contains stuff that wouldn't fit anywhere else.
		3572
		3573	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3574
		3575	Each watcher has, by default, a C<void *data> member that you can read
		3576	or modify at any time: libev will completely ignore it. This can be used
		3577	to associate arbitrary data with your watcher. If you need more data and
		3578	don't want to allocate memory separately and store a pointer to it in that
		3579	data member, you can also "subclass" the watcher type and provide your own
		3580	data:
		3581
		3582	struct my_io
		3583	{
		3584	ev_io io;
		3585	int otherfd;
		3586	void *somedata;
		3587	struct whatever *mostinteresting;
		3588	};
		3589
		3590	...
		3591	struct my_io w;
		3592	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3593
		3594	And since your callback will be called with a pointer to the watcher, you
		3595	can cast it back to your own type:
		3596
		3597	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3598	{
		3599	struct my_io *w = (struct my_io *)w_;
		3600	...
		3601	}
		3602
		3603	More interesting and less C-conformant ways of casting your callback
		3604	function type instead have been omitted.
		3605
		3606	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3607
		3608	Another common scenario is to use some data structure with multiple
		3609	embedded watchers, in effect creating your own watcher that combines
		3610	multiple libev event sources into one "super-watcher":
		3611
		3612	struct my_biggy
		3613	{
		3614	int some_data;
		3615	ev_timer t1;
		3616	ev_timer t2;
		3617	}
		3618
		3619	In this case getting the pointer to C<my_biggy> is a bit more
		3620	complicated: Either you store the address of your C<my_biggy> struct in
		3621	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3622	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3623	real programmers):
		3624
		3625	#include <stddef.h>
		3626
		3627	static void
		3628	t1_cb (EV_P_ ev_timer *w, int revents)
		3629	{
		3630	struct my_biggy big = (struct my_biggy *)
		3631	(((char *)w) - offsetof (struct my_biggy, t1));
		3632	}
		3633
		3634	static void
		3635	t2_cb (EV_P_ ev_timer *w, int revents)
		3636	{
		3637	struct my_biggy big = (struct my_biggy *)
		3638	(((char *)w) - offsetof (struct my_biggy, t2));
		3639	}
		3640
		3641	=head2 AVOIDING FINISHING BEFORE RETURNING
		3642
		3643	Often you have structures like this in event-based programs:
		3644
		3645	callback ()
		3646	{
		3647	free (request);
		3648	}
		3649
		3650	request = start_new_request (..., callback);
		3651
		3652	The intent is to start some "lengthy" operation. The C<request> could be
		3653	used to cancel the operation, or do other things with it.
		3654
		3655	It's not uncommon to have code paths in C<start_new_request> that
		3656	immediately invoke the callback, for example, to report errors. Or you add
		3657	some caching layer that finds that it can skip the lengthy aspects of the
		3658	operation and simply invoke the callback with the result.
		3659
		3660	The problem here is that this will happen I<before> C<start_new_request>
		3661	has returned, so C<request> is not set.
		3662
		3663	Even if you pass the request by some safer means to the callback, you
		3664	might want to do something to the request after starting it, such as
		3665	canceling it, which probably isn't working so well when the callback has
		3666	already been invoked.
		3667
		3668	A common way around all these issues is to make sure that
		3669	C<start_new_request> I<always> returns before the callback is invoked. If
		3670	C<start_new_request> immediately knows the result, it can artificially
		3671	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3672	example, or more sneakily, by reusing an existing (stopped) watcher and
		3673	pushing it into the pending queue:
		3674
		3675	ev_set_cb (watcher, callback);
		3676	ev_feed_event (EV_A_ watcher, 0);
		3677
		3678	This way, C<start_new_request> can safely return before the callback is
		3679	invoked, while not delaying callback invocation too much.
		3680
		3681	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3682
		3683	Often (especially in GUI toolkits) there are places where you have
		3684	I<modal> interaction, which is most easily implemented by recursively
		3685	invoking C<ev_run>.
		3686
		3687	This brings the problem of exiting - a callback might want to finish the
		3688	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3689	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3690	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3691	other combination: In these cases, a simple C<ev_break> will not work.
		3692
		3693	The solution is to maintain "break this loop" variable for each C<ev_run>
		3694	invocation, and use a loop around C<ev_run> until the condition is
		3695	triggered, using C<EVRUN_ONCE>:
		3696
		3697	// main loop
		3698	int exit_main_loop = 0;
		3699
		3700	while (!exit_main_loop)
		3701	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3702
		3703	// in a modal watcher
		3704	int exit_nested_loop = 0;
		3705
		3706	while (!exit_nested_loop)
		3707	ev_run (EV_A_ EVRUN_ONCE);
		3708
		3709	To exit from any of these loops, just set the corresponding exit variable:
		3710
		3711	// exit modal loop
		3712	exit_nested_loop = 1;
		3713
		3714	// exit main program, after modal loop is finished
		3715	exit_main_loop = 1;
		3716
		3717	// exit both
		3718	exit_main_loop = exit_nested_loop = 1;
		3719
		3720	=head2 THREAD LOCKING EXAMPLE
		3721
		3722	Here is a fictitious example of how to run an event loop in a different
		3723	thread from where callbacks are being invoked and watchers are
		3724	created/added/removed.
		3725
		3726	For a real-world example, see the C<EV::Loop::Async> perl module,
		3727	which uses exactly this technique (which is suited for many high-level
		3728	languages).
		3729
		3730	The example uses a pthread mutex to protect the loop data, a condition
		3731	variable to wait for callback invocations, an async watcher to notify the
		3732	event loop thread and an unspecified mechanism to wake up the main thread.
		3733
		3734	First, you need to associate some data with the event loop:
		3735
		3736	typedef struct {
		3737	mutex_t lock; /* global loop lock */
		3738	ev_async async_w;
		3739	thread_t tid;
		3740	cond_t invoke_cv;
		3741	} userdata;
		3742
		3743	void prepare_loop (EV_P)
		3744	{
		3745	// for simplicity, we use a static userdata struct.
		3746	static userdata u;
		3747
		3748	ev_async_init (&u->async_w, async_cb);
		3749	ev_async_start (EV_A_ &u->async_w);
		3750
		3751	pthread_mutex_init (&u->lock, 0);
		3752	pthread_cond_init (&u->invoke_cv, 0);
		3753
		3754	// now associate this with the loop
		3755	ev_set_userdata (EV_A_ u);
		3756	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3757	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3758
		3759	// then create the thread running ev_run
		3760	pthread_create (&u->tid, 0, l_run, EV_A);
		3761	}
		3762
		3763	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3764	solely to wake up the event loop so it takes notice of any new watchers
		3765	that might have been added:
		3766
		3767	static void
		3768	async_cb (EV_P_ ev_async *w, int revents)
		3769	{
		3770	// just used for the side effects
		3771	}
		3772
		3773	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3774	protecting the loop data, respectively.
		3775
		3776	static void
		3777	l_release (EV_P)
		3778	{
		3779	userdata *u = ev_userdata (EV_A);
		3780	pthread_mutex_unlock (&u->lock);
		3781	}
		3782
		3783	static void
		3784	l_acquire (EV_P)
		3785	{
		3786	userdata *u = ev_userdata (EV_A);
		3787	pthread_mutex_lock (&u->lock);
		3788	}
		3789
		3790	The event loop thread first acquires the mutex, and then jumps straight
		3791	into C<ev_run>:
		3792
		3793	void *
		3794	l_run (void *thr_arg)
		3795	{
		3796	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3797
		3798	l_acquire (EV_A);
		3799	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3800	ev_run (EV_A_ 0);
		3801	l_release (EV_A);
		3802
		3803	return 0;
		3804	}
		3805
		3806	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3807	signal the main thread via some unspecified mechanism (signals? pipe
		3808	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3809	have been called (in a while loop because a) spurious wakeups are possible
		3810	and b) skipping inter-thread-communication when there are no pending
		3811	watchers is very beneficial):
		3812
		3813	static void
		3814	l_invoke (EV_P)
		3815	{
		3816	userdata *u = ev_userdata (EV_A);
		3817
		3818	while (ev_pending_count (EV_A))
		3819	{
		3820	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3821	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3822	}
		3823	}
		3824
		3825	Now, whenever the main thread gets told to invoke pending watchers, it
		3826	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3827	thread to continue:
		3828
		3829	static void
		3830	real_invoke_pending (EV_P)
		3831	{
		3832	userdata *u = ev_userdata (EV_A);
		3833
		3834	pthread_mutex_lock (&u->lock);
		3835	ev_invoke_pending (EV_A);
		3836	pthread_cond_signal (&u->invoke_cv);
		3837	pthread_mutex_unlock (&u->lock);
		3838	}
		3839
		3840	Whenever you want to start/stop a watcher or do other modifications to an
		3841	event loop, you will now have to lock:
		3842
		3843	ev_timer timeout_watcher;
		3844	userdata *u = ev_userdata (EV_A);
		3845
		3846	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3847
		3848	pthread_mutex_lock (&u->lock);
		3849	ev_timer_start (EV_A_ &timeout_watcher);
		3850	ev_async_send (EV_A_ &u->async_w);
		3851	pthread_mutex_unlock (&u->lock);
		3852
		3853	Note that sending the C<ev_async> watcher is required because otherwise
		3854	an event loop currently blocking in the kernel will have no knowledge
		3855	about the newly added timer. By waking up the loop it will pick up any new
		3856	watchers in the next event loop iteration.
		3857
		3858	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3859
		3860	While the overhead of a callback that e.g. schedules a thread is small, it
		3861	is still an overhead. If you embed libev, and your main usage is with some
		3862	kind of threads or coroutines, you might want to customise libev so that
		3863	doesn't need callbacks anymore.
		3864
		3865	Imagine you have coroutines that you can switch to using a function
		3866	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3867	and that due to some magic, the currently active coroutine is stored in a
		3868	global called C<current_coro>. Then you can build your own "wait for libev
		3869	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3870	the differing C<;> conventions):
		3871
		3872	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3873	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3874
		3875	That means instead of having a C callback function, you store the
		3876	coroutine to switch to in each watcher, and instead of having libev call
		3877	your callback, you instead have it switch to that coroutine.
		3878
		3879	A coroutine might now wait for an event with a function called
		3880	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3881	matter when, or whether the watcher is active or not when this function is
		3882	called):
		3883
		3884	void
		3885	wait_for_event (ev_watcher *w)
		3886	{
		3887	ev_set_cb (w, current_coro);
		3888	switch_to (libev_coro);
		3889	}
		3890
		3891	That basically suspends the coroutine inside C<wait_for_event> and
		3892	continues the libev coroutine, which, when appropriate, switches back to
		3893	this or any other coroutine.
		3894
		3895	You can do similar tricks if you have, say, threads with an event queue -
		3896	instead of storing a coroutine, you store the queue object and instead of
		3897	switching to a coroutine, you push the watcher onto the queue and notify
		3898	any waiters.
		3899
		3900	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		3901	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3902
		3903	// my_ev.h
		3904	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3905	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
		3906	#include "../libev/ev.h"
		3907
		3908	// my_ev.c
		3909	#define EV_H "my_ev.h"
		3910	#include "../libev/ev.c"
		3911
		3912	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3913	F<my_ev.c> into your project. When properly specifying include paths, you
		3914	can even use F<ev.h> as header file name directly.
3360		3915
3361		3916
3362	=head1 LIBEVENT EMULATION	3917	=head1 LIBEVENT EMULATION
3363		3918
3364	Libev offers a compatibility emulation layer for libevent. It cannot	3919	Libev offers a compatibility emulation layer for libevent. It cannot
3365	emulate the internals of libevent, so here are some usage hints:	3920	emulate the internals of libevent, so here are some usage hints:
3366		3921
3367	=over 4	3922	=over 4
		3923
		3924	=item * Only the libevent-1.4.1-beta API is being emulated.
		3925
		3926	This was the newest libevent version available when libev was implemented,
		3927	and is still mostly unchanged in 2010.
3368		3928
3369	=item * Use it by including <event.h>, as usual.	3929	=item * Use it by including <event.h>, as usual.
3370		3930
3371	=item * The following members are fully supported: ev_base, ev_callback,	3931	=item * The following members are fully supported: ev_base, ev_callback,
3372	ev_arg, ev_fd, ev_res, ev_events.	3932	ev_arg, ev_fd, ev_res, ev_events.
…		…
3378	=item * Priorities are not currently supported. Initialising priorities	3938	=item * Priorities are not currently supported. Initialising priorities
3379	will fail and all watchers will have the same priority, even though there	3939	will fail and all watchers will have the same priority, even though there
3380	is an ev_pri field.	3940	is an ev_pri field.
3381		3941
3382	=item * In libevent, the last base created gets the signals, in libev, the	3942	=item * In libevent, the last base created gets the signals, in libev, the
3383	first base created (== the default loop) gets the signals.	3943	base that registered the signal gets the signals.
3384		3944
3385	=item * Other members are not supported.	3945	=item * Other members are not supported.
3386		3946
3387	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3947	=item * The libev emulation is I<not> ABI compatible to libevent, you need
3388	to use the libev header file and library.	3948	to use the libev header file and library.
3389		3949
3390	=back	3950	=back
3391		3951
3392	=head1 C++ SUPPORT	3952	=head1 C++ SUPPORT
		3953
		3954	=head2 C API
		3955
		3956	The normal C API should work fine when used from C++: both ev.h and the
		3957	libev sources can be compiled as C++. Therefore, code that uses the C API
		3958	will work fine.
		3959
		3960	Proper exception specifications might have to be added to callbacks passed
		3961	to libev: exceptions may be thrown only from watcher callbacks, all
		3962	other callbacks (allocator, syserr, loop acquire/release and periodic
		3963	reschedule callbacks) must not throw exceptions, and might need a C<throw
		3964	()> specification. If you have code that needs to be compiled as both C
		3965	and C++ you can use the C<EV_THROW> macro for this:
		3966
		3967	static void
		3968	fatal_error (const char *msg) EV_THROW
		3969	{
		3970	perror (msg);
		3971	abort ();
		3972	}
		3973
		3974	...
		3975	ev_set_syserr_cb (fatal_error);
		3976
		3977	The only API functions that can currently throw exceptions are C<ev_run>,
		3978	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		3979	because it runs cleanup watchers).
		3980
		3981	Throwing exceptions in watcher callbacks is only supported if libev itself
		3982	is compiled with a C++ compiler or your C and C++ environments allow
		3983	throwing exceptions through C libraries (most do).
		3984
		3985	=head2 C++ API
3393		3986
3394	Libev comes with some simplistic wrapper classes for C++ that mainly allow	3987	Libev comes with some simplistic wrapper classes for C++ that mainly allow
3395	you to use some convenience methods to start/stop watchers and also change	3988	you to use some convenience methods to start/stop watchers and also change
3396	the callback model to a model using method callbacks on objects.	3989	the callback model to a model using method callbacks on objects.
3397		3990
3398	To use it,	3991	To use it,
3399		3992
3400	#include <ev++.h>	3993	#include <ev++.h>
3401		3994
3402	This automatically includes F<ev.h> and puts all of its definitions (many	3995	This automatically includes F<ev.h> and puts all of its definitions (many
3403	of them macros) into the global namespace. All C++ specific things are	3996	of them macros) into the global namespace. All C++ specific things are
3404	put into the C<ev> namespace. It should support all the same embedding	3997	put into the C<ev> namespace. It should support all the same embedding
…		…
3407	Care has been taken to keep the overhead low. The only data member the C++	4000	Care has been taken to keep the overhead low. The only data member the C++
3408	classes add (compared to plain C-style watchers) is the event loop pointer	4001	classes add (compared to plain C-style watchers) is the event loop pointer
3409	that the watcher is associated with (or no additional members at all if	4002	that the watcher is associated with (or no additional members at all if
3410	you disable C<EV_MULTIPLICITY> when embedding libev).	4003	you disable C<EV_MULTIPLICITY> when embedding libev).
3411		4004
3412	Currently, functions, and static and non-static member functions can be	4005	Currently, functions, static and non-static member functions and classes
3413	used as callbacks. Other types should be easy to add as long as they only	4006	with C<operator ()> can be used as callbacks. Other types should be easy
3414	need one additional pointer for context. If you need support for other	4007	to add as long as they only need one additional pointer for context. If
3415	types of functors please contact the author (preferably after implementing	4008	you need support for other types of functors please contact the author
3416	it).	4009	(preferably after implementing it).
		4010
		4011	For all this to work, your C++ compiler either has to use the same calling
		4012	conventions as your C compiler (for static member functions), or you have
		4013	to embed libev and compile libev itself as C++.
3417		4014
3418	Here is a list of things available in the C<ev> namespace:	4015	Here is a list of things available in the C<ev> namespace:
3419		4016
3420	=over 4	4017	=over 4
3421		4018
…		…
3431	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.	4028	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
3432		4029
3433	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of	4030	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
3434	the same name in the C<ev> namespace, with the exception of C<ev_signal>	4031	the same name in the C<ev> namespace, with the exception of C<ev_signal>
3435	which is called C<ev::sig> to avoid clashes with the C<signal> macro	4032	which is called C<ev::sig> to avoid clashes with the C<signal> macro
3436	defines by many implementations.	4033	defined by many implementations.
3437		4034
3438	All of those classes have these methods:	4035	All of those classes have these methods:
3439		4036
3440	=over 4	4037	=over 4
3441		4038
…		…
3503	void operator() (ev::io &w, int revents)	4100	void operator() (ev::io &w, int revents)
3504	{	4101	{
3505	...	4102	...
3506	}	4103	}
3507	}	4104	}
3508		4105
3509	myfunctor f;	4106	myfunctor f;
3510		4107
3511	ev::io w;	4108	ev::io w;
3512	w.set (&f);	4109	w.set (&f);
3513		4110
…		…
3531	Associates a different C<struct ev_loop> with this watcher. You can only	4128	Associates a different C<struct ev_loop> with this watcher. You can only
3532	do this when the watcher is inactive (and not pending either).	4129	do this when the watcher is inactive (and not pending either).
3533		4130
3534	=item w->set ([arguments])	4131	=item w->set ([arguments])
3535		4132
3536	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this	4133	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
3537	method or a suitable start method must be called at least once. Unlike the	4134	with the same arguments. Either this method or a suitable start method
3538	C counterpart, an active watcher gets automatically stopped and restarted	4135	must be called at least once. Unlike the C counterpart, an active watcher
3539	when reconfiguring it with this method.	4136	gets automatically stopped and restarted when reconfiguring it with this
		4137	method.
		4138
		4139	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4140	clashing with the C<set (loop)> method.
3540		4141
3541	=item w->start ()	4142	=item w->start ()
3542		4143
3543	Starts the watcher. Note that there is no C<loop> argument, as the	4144	Starts the watcher. Note that there is no C<loop> argument, as the
3544	constructor already stores the event loop.	4145	constructor already stores the event loop.
…		…
3574	watchers in the constructor.	4175	watchers in the constructor.
3575		4176
3576	class myclass	4177	class myclass
3577	{	4178	{
3578	ev::io io ; void io_cb (ev::io &w, int revents);	4179	ev::io io ; void io_cb (ev::io &w, int revents);
3579	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);	4180	ev::io io2 ; void io2_cb (ev::io &w, int revents);
3580	ev::idle idle; void idle_cb (ev::idle &w, int revents);	4181	ev::idle idle; void idle_cb (ev::idle &w, int revents);
3581		4182
3582	myclass (int fd)	4183	myclass (int fd)
3583	{	4184	{
3584	io .set <myclass, &myclass::io_cb > (this);	4185	io .set <myclass, &myclass::io_cb > (this);
…		…
3635	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.	4236	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
3636		4237
3637	=item D	4238	=item D
3638		4239
3639	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	4240	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3640	be found at L<http://proj.llucax.com.ar/wiki/evd>.	4241	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
3641		4242
3642	=item Ocaml	4243	=item Ocaml
3643		4244
3644	Erkki Seppala has written Ocaml bindings for libev, to be found at	4245	Erkki Seppala has written Ocaml bindings for libev, to be found at
3645	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.	4246	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
…		…
3648		4249
3649	Brian Maher has written a partial interface to libev for lua (at the	4250	Brian Maher has written a partial interface to libev for lua (at the
3650	time of this writing, only C<ev_io> and C<ev_timer>), to be found at	4251	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
3651	L<http://github.com/brimworks/lua-ev>.	4252	L<http://github.com/brimworks/lua-ev>.
3652		4253
		4254	=item Javascript
		4255
		4256	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4257
		4258	=item Others
		4259
		4260	There are others, and I stopped counting.
		4261
3653	=back	4262	=back
3654		4263
3655		4264
3656	=head1 MACRO MAGIC	4265	=head1 MACRO MAGIC
3657		4266
…		…
3693	suitable for use with C<EV_A>.	4302	suitable for use with C<EV_A>.
3694		4303
3695	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	4304	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
3696		4305
3697	Similar to the other two macros, this gives you the value of the default	4306	Similar to the other two macros, this gives you the value of the default
3698	loop, if multiple loops are supported ("ev loop default").	4307	loop, if multiple loops are supported ("ev loop default"). The default loop
		4308	will be initialised if it isn't already initialised.
		4309
		4310	For non-multiplicity builds, these macros do nothing, so you always have
		4311	to initialise the loop somewhere.
3699		4312
3700	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>	4313	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3701		4314
3702	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the	4315	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3703	default loop has been initialised (C<UC> == unchecked). Their behaviour	4316	default loop has been initialised (C<UC> == unchecked). Their behaviour
…		…
3848	supported). It will also not define any of the structs usually found in	4461	supported). It will also not define any of the structs usually found in
3849	F<event.h> that are not directly supported by the libev core alone.	4462	F<event.h> that are not directly supported by the libev core alone.
3850		4463
3851	In standalone mode, libev will still try to automatically deduce the	4464	In standalone mode, libev will still try to automatically deduce the
3852	configuration, but has to be more conservative.	4465	configuration, but has to be more conservative.
		4466
		4467	=item EV_USE_FLOOR
		4468
		4469	If defined to be C<1>, libev will use the C<floor ()> function for its
		4470	periodic reschedule calculations, otherwise libev will fall back on a
		4471	portable (slower) implementation. If you enable this, you usually have to
		4472	link against libm or something equivalent. Enabling this when the C<floor>
		4473	function is not available will fail, so the safe default is to not enable
		4474	this.
3853		4475
3854	=item EV_USE_MONOTONIC	4476	=item EV_USE_MONOTONIC
3855		4477
3856	If defined to be C<1>, libev will try to detect the availability of the	4478	If defined to be C<1>, libev will try to detect the availability of the
3857	monotonic clock option at both compile time and runtime. Otherwise no	4479	monotonic clock option at both compile time and runtime. Otherwise no
…		…
3942		4564
3943	If programs implement their own fd to handle mapping on win32, then this	4565	If programs implement their own fd to handle mapping on win32, then this
3944	macro can be used to override the C<close> function, useful to unregister	4566	macro can be used to override the C<close> function, useful to unregister
3945	file descriptors again. Note that the replacement function has to close	4567	file descriptors again. Note that the replacement function has to close
3946	the underlying OS handle.	4568	the underlying OS handle.
		4569
		4570	=item EV_USE_WSASOCKET
		4571
		4572	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4573	communication socket, which works better in some environments. Otherwise,
		4574	the normal C<socket> function will be used, which works better in other
		4575	environments.
3947		4576
3948	=item EV_USE_POLL	4577	=item EV_USE_POLL
3949		4578
3950	If defined to be C<1>, libev will compile in support for the C<poll>(2)	4579	If defined to be C<1>, libev will compile in support for the C<poll>(2)
3951	backend. Otherwise it will be enabled on non-win32 platforms. It	4580	backend. Otherwise it will be enabled on non-win32 platforms. It
…		…
3987	If defined to be C<1>, libev will compile in support for the Linux inotify	4616	If defined to be C<1>, libev will compile in support for the Linux inotify
3988	interface to speed up C<ev_stat> watchers. Its actual availability will	4617	interface to speed up C<ev_stat> watchers. Its actual availability will
3989	be detected at runtime. If undefined, it will be enabled if the headers	4618	be detected at runtime. If undefined, it will be enabled if the headers
3990	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.	4619	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
3991		4620
		4621	=item EV_NO_SMP
		4622
		4623	If defined to be C<1>, libev will assume that memory is always coherent
		4624	between threads, that is, threads can be used, but threads never run on
		4625	different cpus (or different cpu cores). This reduces dependencies
		4626	and makes libev faster.
		4627
		4628	=item EV_NO_THREADS
		4629
		4630	If defined to be C<1>, libev will assume that it will never be called from
		4631	different threads (that includes signal handlers), which is a stronger
		4632	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4633	libev faster.
		4634
3992	=item EV_ATOMIC_T	4635	=item EV_ATOMIC_T
3993		4636
3994	Libev requires an integer type (suitable for storing C<0> or C<1>) whose	4637	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
3995	access is atomic with respect to other threads or signal contexts. No such	4638	access is atomic with respect to other threads or signal contexts. No
3996	type is easily found in the C language, so you can provide your own type	4639	such type is easily found in the C language, so you can provide your own
3997	that you know is safe for your purposes. It is used both for signal handler "locking"	4640	type that you know is safe for your purposes. It is used both for signal
3998	as well as for signal and thread safety in C<ev_async> watchers.	4641	handler "locking" as well as for signal and thread safety in C<ev_async>
		4642	watchers.
3999		4643
4000	In the absence of this define, libev will use C<sig_atomic_t volatile>	4644	In the absence of this define, libev will use C<sig_atomic_t volatile>
4001	(from F<signal.h>), which is usually good enough on most platforms.	4645	(from F<signal.h>), which is usually good enough on most platforms.
4002		4646
4003	=item EV_H (h)	4647	=item EV_H (h)
…		…
4030	will have the C<struct ev_loop *> as first argument, and you can create	4674	will have the C<struct ev_loop *> as first argument, and you can create
4031	additional independent event loops. Otherwise there will be no support	4675	additional independent event loops. Otherwise there will be no support
4032	for multiple event loops and there is no first event loop pointer	4676	for multiple event loops and there is no first event loop pointer
4033	argument. Instead, all functions act on the single default loop.	4677	argument. Instead, all functions act on the single default loop.
4034		4678
		4679	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4680	default loop when multiplicity is switched off - you always have to
		4681	initialise the loop manually in this case.
		4682
4035	=item EV_MINPRI	4683	=item EV_MINPRI
4036		4684
4037	=item EV_MAXPRI	4685	=item EV_MAXPRI
4038		4686
4039	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to	4687	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
…		…
4075	#define EV_USE_POLL 1	4723	#define EV_USE_POLL 1
4076	#define EV_CHILD_ENABLE 1	4724	#define EV_CHILD_ENABLE 1
4077	#define EV_ASYNC_ENABLE 1	4725	#define EV_ASYNC_ENABLE 1
4078		4726
4079	The actual value is a bitset, it can be a combination of the following	4727	The actual value is a bitset, it can be a combination of the following
4080	values:	4728	values (by default, all of these are enabled):
4081		4729
4082	=over 4	4730	=over 4
4083		4731
4084	=item C<1> - faster/larger code	4732	=item C<1> - faster/larger code
4085		4733
…		…
4089	code size by roughly 30% on amd64).	4737	code size by roughly 30% on amd64).
4090		4738
4091	When optimising for size, use of compiler flags such as C<-Os> with	4739	When optimising for size, use of compiler flags such as C<-Os> with
4092	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of	4740	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
4093	assertions.	4741	assertions.
		4742
		4743	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4744	(e.g. gcc with C<-Os>).
4094		4745
4095	=item C<2> - faster/larger data structures	4746	=item C<2> - faster/larger data structures
4096		4747
4097	Replaces the small 2-heap for timer management by a faster 4-heap, larger	4748	Replaces the small 2-heap for timer management by a faster 4-heap, larger
4098	hash table sizes and so on. This will usually further increase code size	4749	hash table sizes and so on. This will usually further increase code size
4099	and can additionally have an effect on the size of data structures at	4750	and can additionally have an effect on the size of data structures at
4100	runtime.	4751	runtime.
4101		4752
		4753	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4754	(e.g. gcc with C<-Os>).
		4755
4102	=item C<4> - full API configuration	4756	=item C<4> - full API configuration
4103		4757
4104	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and	4758	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
4105	enables multiplicity (C<EV_MULTIPLICITY>=1).	4759	enables multiplicity (C<EV_MULTIPLICITY>=1).
4106		4760
…		…
4136		4790
4137	With an intelligent-enough linker (gcc+binutils are intelligent enough	4791	With an intelligent-enough linker (gcc+binutils are intelligent enough
4138	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by	4792	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4139	your program might be left out as well - a binary starting a timer and an	4793	your program might be left out as well - a binary starting a timer and an
4140	I/O watcher then might come out at only 5Kb.	4794	I/O watcher then might come out at only 5Kb.
		4795
		4796	=item EV_API_STATIC
		4797
		4798	If this symbol is defined (by default it is not), then all identifiers
		4799	will have static linkage. This means that libev will not export any
		4800	identifiers, and you cannot link against libev anymore. This can be useful
		4801	when you embed libev, only want to use libev functions in a single file,
		4802	and do not want its identifiers to be visible.
		4803
		4804	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4805	wants to use libev.
		4806
		4807	This option only works when libev is compiled with a C compiler, as C++
		4808	doesn't support the required declaration syntax.
4141		4809
4142	=item EV_AVOID_STDIO	4810	=item EV_AVOID_STDIO
4143		4811
4144	If this is set to C<1> at compiletime, then libev will avoid using stdio	4812	If this is set to C<1> at compiletime, then libev will avoid using stdio
4145	functions (printf, scanf, perror etc.). This will increase the code size	4813	functions (printf, scanf, perror etc.). This will increase the code size
…		…
4289	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4957	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4290		4958
4291	#include "ev_cpp.h"	4959	#include "ev_cpp.h"
4292	#include "ev.c"	4960	#include "ev.c"
4293		4961
4294	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES	4962	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
4295		4963
4296	=head2 THREADS AND COROUTINES	4964	=head2 THREADS AND COROUTINES
4297		4965
4298	=head3 THREADS	4966	=head3 THREADS
4299		4967
…		…
4350	default loop and triggering an C<ev_async> watcher from the default loop	5018	default loop and triggering an C<ev_async> watcher from the default loop
4351	watcher callback into the event loop interested in the signal.	5019	watcher callback into the event loop interested in the signal.
4352		5020
4353	=back	5021	=back
4354		5022
4355	=head4 THREAD LOCKING EXAMPLE	5023	See also L</THREAD LOCKING EXAMPLE>.
4356
4357	Here is a fictitious example of how to run an event loop in a different
4358	thread than where callbacks are being invoked and watchers are
4359	created/added/removed.
4360
4361	For a real-world example, see the C<EV::Loop::Async> perl module,
4362	which uses exactly this technique (which is suited for many high-level
4363	languages).
4364
4365	The example uses a pthread mutex to protect the loop data, a condition
4366	variable to wait for callback invocations, an async watcher to notify the
4367	event loop thread and an unspecified mechanism to wake up the main thread.
4368
4369	First, you need to associate some data with the event loop:
4370
4371	typedef struct {
4372	mutex_t lock; /* global loop lock */
4373	ev_async async_w;
4374	thread_t tid;
4375	cond_t invoke_cv;
4376	} userdata;
4377
4378	void prepare_loop (EV_P)
4379	{
4380	// for simplicity, we use a static userdata struct.
4381	static userdata u;
4382
4383	ev_async_init (&u->async_w, async_cb);
4384	ev_async_start (EV_A_ &u->async_w);
4385
4386	pthread_mutex_init (&u->lock, 0);
4387	pthread_cond_init (&u->invoke_cv, 0);
4388
4389	// now associate this with the loop
4390	ev_set_userdata (EV_A_ u);
4391	ev_set_invoke_pending_cb (EV_A_ l_invoke);
4392	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
4393
4394	// then create the thread running ev_loop
4395	pthread_create (&u->tid, 0, l_run, EV_A);
4396	}
4397
4398	The callback for the C<ev_async> watcher does nothing: the watcher is used
4399	solely to wake up the event loop so it takes notice of any new watchers
4400	that might have been added:
4401
4402	static void
4403	async_cb (EV_P_ ev_async *w, int revents)
4404	{
4405	// just used for the side effects
4406	}
4407
4408	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
4409	protecting the loop data, respectively.
4410
4411	static void
4412	l_release (EV_P)
4413	{
4414	userdata *u = ev_userdata (EV_A);
4415	pthread_mutex_unlock (&u->lock);
4416	}
4417
4418	static void
4419	l_acquire (EV_P)
4420	{
4421	userdata *u = ev_userdata (EV_A);
4422	pthread_mutex_lock (&u->lock);
4423	}
4424
4425	The event loop thread first acquires the mutex, and then jumps straight
4426	into C<ev_run>:
4427
4428	void *
4429	l_run (void *thr_arg)
4430	{
4431	struct ev_loop *loop = (struct ev_loop *)thr_arg;
4432
4433	l_acquire (EV_A);
4434	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
4435	ev_run (EV_A_ 0);
4436	l_release (EV_A);
4437
4438	return 0;
4439	}
4440
4441	Instead of invoking all pending watchers, the C<l_invoke> callback will
4442	signal the main thread via some unspecified mechanism (signals? pipe
4443	writes? C<Async::Interrupt>?) and then waits until all pending watchers
4444	have been called (in a while loop because a) spurious wakeups are possible
4445	and b) skipping inter-thread-communication when there are no pending
4446	watchers is very beneficial):
4447
4448	static void
4449	l_invoke (EV_P)
4450	{
4451	userdata *u = ev_userdata (EV_A);
4452
4453	while (ev_pending_count (EV_A))
4454	{
4455	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
4456	pthread_cond_wait (&u->invoke_cv, &u->lock);
4457	}
4458	}
4459
4460	Now, whenever the main thread gets told to invoke pending watchers, it
4461	will grab the lock, call C<ev_invoke_pending> and then signal the loop
4462	thread to continue:
4463
4464	static void
4465	real_invoke_pending (EV_P)
4466	{
4467	userdata *u = ev_userdata (EV_A);
4468
4469	pthread_mutex_lock (&u->lock);
4470	ev_invoke_pending (EV_A);
4471	pthread_cond_signal (&u->invoke_cv);
4472	pthread_mutex_unlock (&u->lock);
4473	}
4474
4475	Whenever you want to start/stop a watcher or do other modifications to an
4476	event loop, you will now have to lock:
4477
4478	ev_timer timeout_watcher;
4479	userdata *u = ev_userdata (EV_A);
4480
4481	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
4482
4483	pthread_mutex_lock (&u->lock);
4484	ev_timer_start (EV_A_ &timeout_watcher);
4485	ev_async_send (EV_A_ &u->async_w);
4486	pthread_mutex_unlock (&u->lock);
4487
4488	Note that sending the C<ev_async> watcher is required because otherwise
4489	an event loop currently blocking in the kernel will have no knowledge
4490	about the newly added timer. By waking up the loop it will pick up any new
4491	watchers in the next event loop iteration.
4492		5024
4493	=head3 COROUTINES	5025	=head3 COROUTINES
4494		5026
4495	Libev is very accommodating to coroutines ("cooperative threads"):	5027	Libev is very accommodating to coroutines ("cooperative threads"):
4496	libev fully supports nesting calls to its functions from different	5028	libev fully supports nesting calls to its functions from different
…		…
4661	requires, and its I/O model is fundamentally incompatible with the POSIX	5193	requires, and its I/O model is fundamentally incompatible with the POSIX
4662	model. Libev still offers limited functionality on this platform in	5194	model. Libev still offers limited functionality on this platform in
4663	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	5195	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
4664	descriptors. This only applies when using Win32 natively, not when using	5196	descriptors. This only applies when using Win32 natively, not when using
4665	e.g. cygwin. Actually, it only applies to the microsofts own compilers,	5197	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
4666	as every compielr comes with a slightly differently broken/incompatible	5198	as every compiler comes with a slightly differently broken/incompatible
4667	environment.	5199	environment.
4668		5200
4669	Lifting these limitations would basically require the full	5201	Lifting these limitations would basically require the full
4670	re-implementation of the I/O system. If you are into this kind of thing,	5202	re-implementation of the I/O system. If you are into this kind of thing,
4671	then note that glib does exactly that for you in a very portable way (note	5203	then note that glib does exactly that for you in a very portable way (note
…		…
4787	thread" or will block signals process-wide, both behaviours would	5319	thread" or will block signals process-wide, both behaviours would
4788	be compatible with libev. Interaction between C<sigprocmask> and	5320	be compatible with libev. Interaction between C<sigprocmask> and
4789	C<pthread_sigmask> could complicate things, however.	5321	C<pthread_sigmask> could complicate things, however.
4790		5322
4791	The most portable way to handle signals is to block signals in all threads	5323	The most portable way to handle signals is to block signals in all threads
4792	except the initial one, and run the default loop in the initial thread as	5324	except the initial one, and run the signal handling loop in the initial
4793	well.	5325	thread as well.
4794		5326
4795	=item C<long> must be large enough for common memory allocation sizes	5327	=item C<long> must be large enough for common memory allocation sizes
4796		5328
4797	To improve portability and simplify its API, libev uses C<long> internally	5329	To improve portability and simplify its API, libev uses C<long> internally
4798	instead of C<size_t> when allocating its data structures. On non-POSIX	5330	instead of C<size_t> when allocating its data structures. On non-POSIX
…		…
4804		5336
4805	The type C<double> is used to represent timestamps. It is required to	5337	The type C<double> is used to represent timestamps. It is required to
4806	have at least 51 bits of mantissa (and 9 bits of exponent), which is	5338	have at least 51 bits of mantissa (and 9 bits of exponent), which is
4807	good enough for at least into the year 4000 with millisecond accuracy	5339	good enough for at least into the year 4000 with millisecond accuracy
4808	(the design goal for libev). This requirement is overfulfilled by	5340	(the design goal for libev). This requirement is overfulfilled by
4809	implementations using IEEE 754, which is basically all existing ones. With	5341	implementations using IEEE 754, which is basically all existing ones.
		5342
4810	IEEE 754 doubles, you get microsecond accuracy until at least 2200.	5343	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5344	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5345	is either obsolete or somebody patched it to use C<long double> or
		5346	something like that, just kidding).
4811		5347
4812	=back	5348	=back
4813		5349
4814	If you know of other additional requirements drop me a note.	5350	If you know of other additional requirements drop me a note.
4815		5351
…		…
4877	=item Processing ev_async_send: O(number_of_async_watchers)	5413	=item Processing ev_async_send: O(number_of_async_watchers)
4878		5414
4879	=item Processing signals: O(max_signal_number)	5415	=item Processing signals: O(max_signal_number)
4880		5416
4881	Sending involves a system call I<iff> there were no other C<ev_async_send>	5417	Sending involves a system call I<iff> there were no other C<ev_async_send>
4882	calls in the current loop iteration. Checking for async and signal events	5418	calls in the current loop iteration and the loop is currently
		5419	blocked. Checking for async and signal events involves iterating over all
4883	involves iterating over all running async watchers or all signal numbers.	5420	running async watchers or all signal numbers.
4884		5421
4885	=back	5422	=back
4886		5423
4887		5424
4888	=head1 PORTING FROM LIBEV 3.X TO 4.X	5425	=head1 PORTING FROM LIBEV 3.X TO 4.X
…		…
4897	=over 4	5434	=over 4
4898		5435
4899	=item C<EV_COMPAT3> backwards compatibility mechanism	5436	=item C<EV_COMPAT3> backwards compatibility mechanism
4900		5437
4901	The backward compatibility mechanism can be controlled by	5438	The backward compatibility mechanism can be controlled by
4902	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>	5439	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
4903	section.	5440	section.
4904		5441
4905	=item C<ev_default_destroy> and C<ev_default_fork> have been removed	5442	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
4906		5443
4907	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:	5444	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
…		…
4950	=over 4	5487	=over 4
4951		5488
4952	=item active	5489	=item active
4953		5490
4954	A watcher is active as long as it has been started and not yet stopped.	5491	A watcher is active as long as it has been started and not yet stopped.
4955	See L<WATCHER STATES> for details.	5492	See L</WATCHER STATES> for details.
4956		5493
4957	=item application	5494	=item application
4958		5495
4959	In this document, an application is whatever is using libev.	5496	In this document, an application is whatever is using libev.
4960		5497
…		…
4996	watchers and events.	5533	watchers and events.
4997		5534
4998	=item pending	5535	=item pending
4999		5536
5000	A watcher is pending as soon as the corresponding event has been	5537	A watcher is pending as soon as the corresponding event has been
5001	detected. See L<WATCHER STATES> for details.	5538	detected. See L</WATCHER STATES> for details.
5002		5539
5003	=item real time	5540	=item real time
5004		5541
5005	The physical time that is observed. It is apparently strictly monotonic :)	5542	The physical time that is observed. It is apparently strictly monotonic :)
5006		5543
5007	=item wall-clock time	5544	=item wall-clock time
5008		5545
5009	The time and date as shown on clocks. Unlike real time, it can actually	5546	The time and date as shown on clocks. Unlike real time, it can actually
5010	be wrong and jump forwards and backwards, e.g. when the you adjust your	5547	be wrong and jump forwards and backwards, e.g. when you adjust your
5011	clock.	5548	clock.
5012		5549
5013	=item watcher	5550	=item watcher
5014		5551
5015	A data structure that describes interest in certain events. Watchers need	5552	A data structure that describes interest in certain events. Watchers need
…		…
5018	=back	5555	=back
5019		5556
5020	=head1 AUTHOR	5557	=head1 AUTHOR
5021		5558
5022	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael	5559	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
5023	Magnusson and Emanuele Giaquinta.	5560	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
5024		5561

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines