ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.202 by root, Fri Oct 24 08:30:01 2008 UTC vs.
Revision 1.259 by root, Sun Jul 19 01:36:34 2009 UTC

…		…
8		8
9	=head2 EXAMPLE PROGRAM	9	=head2 EXAMPLE PROGRAM
10		10
11	// a single header file is required	11	// a single header file is required
12	#include <ev.h>	12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
13		15
14	// every watcher type has its own typedef'd struct	16	// every watcher type has its own typedef'd struct
15	// with the name ev_TYPE	17	// with the name ev_TYPE
16	ev_io stdin_watcher;	18	ev_io stdin_watcher;
17	ev_timer timeout_watcher;	19	ev_timer timeout_watcher;
…		…
41		43
42	int	44	int
43	main (void)	45	main (void)
44	{	46	{
45	// use the default event loop unless you have special needs	47	// use the default event loop unless you have special needs
46	ev_loop *loop = ev_default_loop (0);	48	struct ev_loop *loop = ev_default_loop (0);
47		49
48	// initialise an io watcher, then start it	50	// initialise an io watcher, then start it
49	// this one will watch for stdin to become readable	51	// this one will watch for stdin to become readable
50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);	52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
51	ev_io_start (loop, &stdin_watcher);	53	ev_io_start (loop, &stdin_watcher);
…		…
60		62
61	// unloop was called, so exit	63	// unloop was called, so exit
62	return 0;	64	return 0;
63	}	65	}
64		66
65	=head1 DESCRIPTION	67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
66		70
67	The newest version of this document is also available as an html-formatted	71	The newest version of this document is also available as an html-formatted
68	web page you might find easier to navigate when reading it for the first	72	web page you might find easier to navigate when reading it for the first
69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.	73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 ABOUT LIBEV
70		84
71	Libev is an event loop: you register interest in certain events (such as a	85	Libev is an event loop: you register interest in certain events (such as a
72	file descriptor being readable or a timeout occurring), and it will manage	86	file descriptor being readable or a timeout occurring), and it will manage
73	these event sources and provide your program with events.	87	these event sources and provide your program with events.
74		88
…		…
108	name C<loop> (which is always of type C<ev_loop *>) will not have	122	name C<loop> (which is always of type C<ev_loop *>) will not have
109	this argument.	123	this argument.
110		124
111	=head2 TIME REPRESENTATION	125	=head2 TIME REPRESENTATION
112		126
113	Libev represents time as a single floating point number, representing the	127	Libev represents time as a single floating point number, representing
114	(fractional) number of seconds since the (POSIX) epoch (somewhere near	128	the (fractional) number of seconds since the (POSIX) epoch (somewhere
115	the beginning of 1970, details are complicated, don't ask). This type is	129	near the beginning of 1970, details are complicated, don't ask). This
116	called C<ev_tstamp>, which is what you should use too. It usually aliases	130	type is called C<ev_tstamp>, which is what you should use too. It usually
117	to the C<double> type in C, and when you need to do any calculations on	131	aliases to the C<double> type in C. When you need to do any calculations
118	it, you should treat it as some floating point value. Unlike the name	132	on it, you should treat it as some floating point value. Unlike the name
119	component C<stamp> might indicate, it is also used for time differences	133	component C<stamp> might indicate, it is also used for time differences
120	throughout libev.	134	throughout libev.
121		135
122	=head1 ERROR HANDLING	136	=head1 ERROR HANDLING
123		137
…		…
298	If you don't know what event loop to use, use the one returned from this	312	If you don't know what event loop to use, use the one returned from this
299	function.	313	function.
300		314
301	Note that this function is I<not> thread-safe, so if you want to use it	315	Note that this function is I<not> thread-safe, so if you want to use it
302	from multiple threads, you have to lock (note also that this is unlikely,	316	from multiple threads, you have to lock (note also that this is unlikely,
303	as loops cannot bes hared easily between threads anyway).	317	as loops cannot be shared easily between threads anyway).
304		318
305	The default loop is the only loop that can handle C<ev_signal> and	319	The default loop is the only loop that can handle C<ev_signal> and
306	C<ev_child> watchers, and to do this, it always registers a handler	320	C<ev_child> watchers, and to do this, it always registers a handler
307	for C<SIGCHLD>. If this is a problem for your application you can either	321	for C<SIGCHLD>. If this is a problem for your application you can either
308	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you	322	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
…		…
384	=item C<EVBACKEND_EPOLL> (value 4, Linux)	398	=item C<EVBACKEND_EPOLL> (value 4, Linux)
385		399
386	For few fds, this backend is a bit little slower than poll and select,	400	For few fds, this backend is a bit little slower than poll and select,
387	but it scales phenomenally better. While poll and select usually scale	401	but it scales phenomenally better. While poll and select usually scale
388	like O(total_fds) where n is the total number of fds (or the highest fd),	402	like O(total_fds) where n is the total number of fds (or the highest fd),
389	epoll scales either O(1) or O(active_fds). The epoll design has a number	403	epoll scales either O(1) or O(active_fds).
390	of shortcomings, such as silently dropping events in some hard-to-detect	404
391	cases and requiring a system call per fd change, no fork support and bad	405	The epoll mechanism deserves honorable mention as the most misdesigned
392	support for dup.	406	of the more advanced event mechanisms: mere annoyances include silently
		407	dropping file descriptors, requiring a system call per change per file
		408	descriptor (and unnecessary guessing of parameters), problems with dup and
		409	so on. The biggest issue is fork races, however - if a program forks then
		410	I<both> parent and child process have to recreate the epoll set, which can
		411	take considerable time (one syscall per file descriptor) and is of course
		412	hard to detect.
		413
		414	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		415	of course I<doesn't>, and epoll just loves to report events for totally
		416	I<different> file descriptors (even already closed ones, so one cannot
		417	even remove them from the set) than registered in the set (especially
		418	on SMP systems). Libev tries to counter these spurious notifications by
		419	employing an additional generation counter and comparing that against the
		420	events to filter out spurious ones, recreating the set when required.
393		421
394	While stopping, setting and starting an I/O watcher in the same iteration	422	While stopping, setting and starting an I/O watcher in the same iteration
395	will result in some caching, there is still a system call per such incident	423	will result in some caching, there is still a system call per such
396	(because the fd could point to a different file description now), so its	424	incident (because the same I<file descriptor> could point to a different
397	best to avoid that. Also, C<dup ()>'ed file descriptors might not work	425	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
398	very well if you register events for both fds.	426	file descriptors might not work very well if you register events for both
399		427	file descriptors.
400	Please note that epoll sometimes generates spurious notifications, so you
401	need to use non-blocking I/O or other means to avoid blocking when no data
402	(or space) is available.
403		428
404	Best performance from this backend is achieved by not unregistering all	429	Best performance from this backend is achieved by not unregistering all
405	watchers for a file descriptor until it has been closed, if possible,	430	watchers for a file descriptor until it has been closed, if possible,
406	i.e. keep at least one watcher active per fd at all times. Stopping and	431	i.e. keep at least one watcher active per fd at all times. Stopping and
407	starting a watcher (without re-setting it) also usually doesn't cause	432	starting a watcher (without re-setting it) also usually doesn't cause
408	extra overhead.	433	extra overhead. A fork can both result in spurious notifications as well
		434	as in libev having to destroy and recreate the epoll object, which can
		435	take considerable time and thus should be avoided.
		436
		437	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		438	faster than epoll for maybe up to a hundred file descriptors, depending on
		439	the usage. So sad.
409		440
410	While nominally embeddable in other event loops, this feature is broken in	441	While nominally embeddable in other event loops, this feature is broken in
411	all kernel versions tested so far.	442	all kernel versions tested so far.
412		443
413	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	444	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
414	C<EVBACKEND_POLL>.	445	C<EVBACKEND_POLL>.
415		446
416	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	447	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
417		448
418	Kqueue deserves special mention, as at the time of this writing, it was	449	Kqueue deserves special mention, as at the time of this writing, it
419	broken on all BSDs except NetBSD (usually it doesn't work reliably with	450	was broken on all BSDs except NetBSD (usually it doesn't work reliably
420	anything but sockets and pipes, except on Darwin, where of course it's	451	with anything but sockets and pipes, except on Darwin, where of course
421	completely useless). For this reason it's not being "auto-detected" unless	452	it's completely useless). Unlike epoll, however, whose brokenness
422	you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or	453	is by design, these kqueue bugs can (and eventually will) be fixed
423	libev was compiled on a known-to-be-good (-enough) system like NetBSD.	454	without API changes to existing programs. For this reason it's not being
		455	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		456	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		457	system like NetBSD.
424		458
425	You still can embed kqueue into a normal poll or select backend and use it	459	You still can embed kqueue into a normal poll or select backend and use it
426	only for sockets (after having made sure that sockets work with kqueue on	460	only for sockets (after having made sure that sockets work with kqueue on
427	the target platform). See C<ev_embed> watchers for more info.	461	the target platform). See C<ev_embed> watchers for more info.
428		462
429	It scales in the same way as the epoll backend, but the interface to the	463	It scales in the same way as the epoll backend, but the interface to the
430	kernel is more efficient (which says nothing about its actual speed, of	464	kernel is more efficient (which says nothing about its actual speed, of
431	course). While stopping, setting and starting an I/O watcher does never	465	course). While stopping, setting and starting an I/O watcher does never
432	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	466	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
433	two event changes per incident. Support for C<fork ()> is very bad and it	467	two event changes per incident. Support for C<fork ()> is very bad (but
434	drops fds silently in similarly hard-to-detect cases.	468	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		469	cases
435		470
436	This backend usually performs well under most conditions.	471	This backend usually performs well under most conditions.
437		472
438	While nominally embeddable in other event loops, this doesn't work	473	While nominally embeddable in other event loops, this doesn't work
439	everywhere, so you might need to test for this. And since it is broken	474	everywhere, so you might need to test for this. And since it is broken
440	almost everywhere, you should only use it when you have a lot of sockets	475	almost everywhere, you should only use it when you have a lot of sockets
441	(for which it usually works), by embedding it into another event loop	476	(for which it usually works), by embedding it into another event loop
442	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,	477	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
443	using it only for sockets.	478	also broken on OS X)) and, did I mention it, using it only for sockets.
444		479
445	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with	480	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
446	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with	481	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
447	C<NOTE_EOF>.	482	C<NOTE_EOF>.
448		483
…		…
468	might perform better.	503	might perform better.
469		504
470	On the positive side, with the exception of the spurious readiness	505	On the positive side, with the exception of the spurious readiness
471	notifications, this backend actually performed fully to specification	506	notifications, this backend actually performed fully to specification
472	in all tests and is fully embeddable, which is a rare feat among the	507	in all tests and is fully embeddable, which is a rare feat among the
473	OS-specific backends.	508	OS-specific backends (I vastly prefer correctness over speed hacks).
474		509
475	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	510	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
476	C<EVBACKEND_POLL>.	511	C<EVBACKEND_POLL>.
477		512
478	=item C<EVBACKEND_ALL>	513	=item C<EVBACKEND_ALL>
…		…
531	responsibility to either stop all watchers cleanly yourself I<before>	566	responsibility to either stop all watchers cleanly yourself I<before>
532	calling this function, or cope with the fact afterwards (which is usually	567	calling this function, or cope with the fact afterwards (which is usually
533	the easiest thing, you can just ignore the watchers and/or C<free ()> them	568	the easiest thing, you can just ignore the watchers and/or C<free ()> them
534	for example).	569	for example).
535		570
536	Note that certain global state, such as signal state, will not be freed by	571	Note that certain global state, such as signal state (and installed signal
537	this function, and related watchers (such as signal and child watchers)	572	handlers), will not be freed by this function, and related watchers (such
538	would need to be stopped manually.	573	as signal and child watchers) would need to be stopped manually.
539		574
540	In general it is not advisable to call this function except in the	575	In general it is not advisable to call this function except in the
541	rare occasion where you really need to free e.g. the signal handling	576	rare occasion where you really need to free e.g. the signal handling
542	pipe fds. If you need dynamically allocated loops it is better to use	577	pipe fds. If you need dynamically allocated loops it is better to use
543	C<ev_loop_new> and C<ev_loop_destroy>).	578	C<ev_loop_new> and C<ev_loop_destroy>).
…		…
586		621
587	This value can sometimes be useful as a generation counter of sorts (it	622	This value can sometimes be useful as a generation counter of sorts (it
588	"ticks" the number of loop iterations), as it roughly corresponds with	623	"ticks" the number of loop iterations), as it roughly corresponds with
589	C<ev_prepare> and C<ev_check> calls.	624	C<ev_prepare> and C<ev_check> calls.
590		625
		626	=item unsigned int ev_loop_depth (loop)
		627
		628	Returns the number of times C<ev_loop> was entered minus the number of
		629	times C<ev_loop> was exited, in other words, the recursion depth.
		630
		631	Outside C<ev_loop>, this number is zero. In a callback, this number is
		632	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		633	in which case it is higher.
		634
		635	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		636	etc.), doesn't count as exit.
		637
591	=item unsigned int ev_backend (loop)	638	=item unsigned int ev_backend (loop)
592		639
593	Returns one of the C<EVBACKEND_*> flags indicating the event backend in	640	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
594	use.	641	use.
595		642
…		…
609		656
610	This function is rarely useful, but when some event callback runs for a	657	This function is rarely useful, but when some event callback runs for a
611	very long time without entering the event loop, updating libev's idea of	658	very long time without entering the event loop, updating libev's idea of
612	the current time is a good idea.	659	the current time is a good idea.
613		660
614	See also "The special problem of time updates" in the C<ev_timer> section.	661	See also L<The special problem of time updates> in the C<ev_timer> section.
		662
		663	=item ev_suspend (loop)
		664
		665	=item ev_resume (loop)
		666
		667	These two functions suspend and resume a loop, for use when the loop is
		668	not used for a while and timeouts should not be processed.
		669
		670	A typical use case would be an interactive program such as a game: When
		671	the user presses C<^Z> to suspend the game and resumes it an hour later it
		672	would be best to handle timeouts as if no time had actually passed while
		673	the program was suspended. This can be achieved by calling C<ev_suspend>
		674	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		675	C<ev_resume> directly afterwards to resume timer processing.
		676
		677	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		678	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		679	will be rescheduled (that is, they will lose any events that would have
		680	occured while suspended).
		681
		682	After calling C<ev_suspend> you B<must not> call I<any> function on the
		683	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		684	without a previous call to C<ev_suspend>.
		685
		686	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		687	event loop time (see C<ev_now_update>).
615		688
616	=item ev_loop (loop, int flags)	689	=item ev_loop (loop, int flags)
617		690
618	Finally, this is it, the event handler. This function usually is called	691	Finally, this is it, the event handler. This function usually is called
619	after you initialised all your watchers and you want to start handling	692	after you initialised all your watchers and you want to start handling
…		…
635	the loop.	708	the loop.
636		709
637	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	710	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
638	necessary) and will handle those and any already outstanding ones. It	711	necessary) and will handle those and any already outstanding ones. It
639	will block your process until at least one new event arrives (which could	712	will block your process until at least one new event arrives (which could
640	be an event internal to libev itself, so there is no guarentee that a	713	be an event internal to libev itself, so there is no guarantee that a
641	user-registered callback will be called), and will return after one	714	user-registered callback will be called), and will return after one
642	iteration of the loop.	715	iteration of the loop.
643		716
644	This is useful if you are waiting for some external event in conjunction	717	This is useful if you are waiting for some external event in conjunction
645	with something not expressible using other libev watchers (i.e. "roll your	718	with something not expressible using other libev watchers (i.e. "roll your
…		…
703		776
704	If you have a watcher you never unregister that should not keep C<ev_loop>	777	If you have a watcher you never unregister that should not keep C<ev_loop>
705	from returning, call ev_unref() after starting, and ev_ref() before	778	from returning, call ev_unref() after starting, and ev_ref() before
706	stopping it.	779	stopping it.
707		780
708	As an example, libev itself uses this for its internal signal pipe: It is	781	As an example, libev itself uses this for its internal signal pipe: It
709	not visible to the libev user and should not keep C<ev_loop> from exiting	782	is not visible to the libev user and should not keep C<ev_loop> from
710	if no event watchers registered by it are active. It is also an excellent	783	exiting if no event watchers registered by it are active. It is also an
711	way to do this for generic recurring timers or from within third-party	784	excellent way to do this for generic recurring timers or from within
712	libraries. Just remember to I<unref after start> and I<ref before stop>	785	third-party libraries. Just remember to I<unref after start> and I<ref
713	(but only if the watcher wasn't active before, or was active before,	786	before stop> (but only if the watcher wasn't active before, or was active
714	respectively).	787	before, respectively. Note also that libev might stop watchers itself
		788	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		789	in the callback).
715		790
716	Example: Create a signal watcher, but keep it from keeping C<ev_loop>	791	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
717	running when nothing else is active.	792	running when nothing else is active.
718		793
719	ev_signal exitsig;	794	ev_signal exitsig;
…		…
748		823
749	By setting a higher I<io collect interval> you allow libev to spend more	824	By setting a higher I<io collect interval> you allow libev to spend more
750	time collecting I/O events, so you can handle more events per iteration,	825	time collecting I/O events, so you can handle more events per iteration,
751	at the cost of increasing latency. Timeouts (both C<ev_periodic> and	826	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
752	C<ev_timer>) will be not affected. Setting this to a non-null value will	827	C<ev_timer>) will be not affected. Setting this to a non-null value will
753	introduce an additional C<ev_sleep ()> call into most loop iterations.	828	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		829	sleep time ensures that libev will not poll for I/O events more often then
		830	once per this interval, on average.
754		831
755	Likewise, by setting a higher I<timeout collect interval> you allow libev	832	Likewise, by setting a higher I<timeout collect interval> you allow libev
756	to spend more time collecting timeouts, at the expense of increased	833	to spend more time collecting timeouts, at the expense of increased
757	latency/jitter/inexactness (the watcher callback will be called	834	latency/jitter/inexactness (the watcher callback will be called
758	later). C<ev_io> watchers will not be affected. Setting this to a non-null	835	later). C<ev_io> watchers will not be affected. Setting this to a non-null
…		…
760		837
761	Many (busy) programs can usually benefit by setting the I/O collect	838	Many (busy) programs can usually benefit by setting the I/O collect
762	interval to a value near C<0.1> or so, which is often enough for	839	interval to a value near C<0.1> or so, which is often enough for
763	interactive servers (of course not for games), likewise for timeouts. It	840	interactive servers (of course not for games), likewise for timeouts. It
764	usually doesn't make much sense to set it to a lower value than C<0.01>,	841	usually doesn't make much sense to set it to a lower value than C<0.01>,
765	as this approaches the timing granularity of most systems.	842	as this approaches the timing granularity of most systems. Note that if
		843	you do transactions with the outside world and you can't increase the
		844	parallelity, then this setting will limit your transaction rate (if you
		845	need to poll once per transaction and the I/O collect interval is 0.01,
		846	then you can't do more than 100 transations per second).
766		847
767	Setting the I<timeout collect interval> can improve the opportunity for	848	Setting the I<timeout collect interval> can improve the opportunity for
768	saving power, as the program will "bundle" timer callback invocations that	849	saving power, as the program will "bundle" timer callback invocations that
769	are "near" in time together, by delaying some, thus reducing the number of	850	are "near" in time together, by delaying some, thus reducing the number of
770	times the process sleeps and wakes up again. Another useful technique to	851	times the process sleeps and wakes up again. Another useful technique to
771	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure	852	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
772	they fire on, say, one-second boundaries only.	853	they fire on, say, one-second boundaries only.
		854
		855	Example: we only need 0.1s timeout granularity, and we wish not to poll
		856	more often than 100 times per second:
		857
		858	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		859	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		860
		861	=item ev_invoke_pending (loop)
		862
		863	This call will simply invoke all pending watchers while resetting their
		864	pending state. Normally, C<ev_loop> does this automatically when required,
		865	but when overriding the invoke callback this call comes handy.
		866
		867	=item int ev_pending_count (loop)
		868
		869	Returns the number of pending watchers - zero indicates that no watchers
		870	are pending.
		871
		872	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		873
		874	This overrides the invoke pending functionality of the loop: Instead of
		875	invoking all pending watchers when there are any, C<ev_loop> will call
		876	this callback instead. This is useful, for example, when you want to
		877	invoke the actual watchers inside another context (another thread etc.).
		878
		879	If you want to reset the callback, use C<ev_invoke_pending> as new
		880	callback.
		881
		882	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		883
		884	Sometimes you want to share the same loop between multiple threads. This
		885	can be done relatively simply by putting mutex_lock/unlock calls around
		886	each call to a libev function.
		887
		888	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		889	wait for it to return. One way around this is to wake up the loop via
		890	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		891	and I<acquire> callbacks on the loop.
		892
		893	When set, then C<release> will be called just before the thread is
		894	suspended waiting for new events, and C<acquire> is called just
		895	afterwards.
		896
		897	Ideally, C<release> will just call your mutex_unlock function, and
		898	C<acquire> will just call the mutex_lock function again.
		899
		900	While event loop modifications are allowed between invocations of
		901	C<release> and C<acquire> (that's their only purpose after all), no
		902	modifications done will affect the event loop, i.e. adding watchers will
		903	have no effect on the set of file descriptors being watched, or the time
		904	waited. USe an C<ev_async> watcher to wake up C<ev_loop> when you want it
		905	to take note of any changes you made.
		906
		907	In theory, threads executing C<ev_loop> will be async-cancel safe between
		908	invocations of C<release> and C<acquire>.
		909
		910	See also the locking example in the C<THREADS> section later in this
		911	document.
		912
		913	=item ev_set_userdata (loop, void *data)
		914
		915	=item ev_userdata (loop)
		916
		917	Set and retrieve a single C<void *> associated with a loop. When
		918	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		919	C<0.>
		920
		921	These two functions can be used to associate arbitrary data with a loop,
		922	and are intended solely for the C<invoke_pending_cb>, C<release> and
		923	C<acquire> callbacks described above, but of course can be (ab-)used for
		924	any other purpose as well.
773		925
774	=item ev_loop_verify (loop)	926	=item ev_loop_verify (loop)
775		927
776	This function only does something when C<EV_VERIFY> support has been	928	This function only does something when C<EV_VERIFY> support has been
777	compiled in, which is the default for non-minimal builds. It tries to go	929	compiled in, which is the default for non-minimal builds. It tries to go
…		…
903		1055
904	=item C<EV_ASYNC>	1056	=item C<EV_ASYNC>
905		1057
906	The given async watcher has been asynchronously notified (see C<ev_async>).	1058	The given async watcher has been asynchronously notified (see C<ev_async>).
907		1059
		1060	=item C<EV_CUSTOM>
		1061
		1062	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1063	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1064
908	=item C<EV_ERROR>	1065	=item C<EV_ERROR>
909		1066
910	An unspecified error has occurred, the watcher has been stopped. This might	1067	An unspecified error has occurred, the watcher has been stopped. This might
911	happen because the watcher could not be properly started because libev	1068	happen because the watcher could not be properly started because libev
912	ran out of memory, a file descriptor was found to be closed or any other	1069	ran out of memory, a file descriptor was found to be closed or any other
…		…
1027	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>	1184	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
1028	(default: C<-2>). Pending watchers with higher priority will be invoked	1185	(default: C<-2>). Pending watchers with higher priority will be invoked
1029	before watchers with lower priority, but priority will not keep watchers	1186	before watchers with lower priority, but priority will not keep watchers
1030	from being executed (except for C<ev_idle> watchers).	1187	from being executed (except for C<ev_idle> watchers).
1031		1188
1032	This means that priorities are I<only> used for ordering callback
1033	invocation after new events have been received. This is useful, for
1034	example, to reduce latency after idling, or more often, to bind two
1035	watchers on the same event and make sure one is called first.
1036
1037	If you need to suppress invocation when higher priority events are pending	1189	If you need to suppress invocation when higher priority events are pending
1038	you need to look at C<ev_idle> watchers, which provide this functionality.	1190	you need to look at C<ev_idle> watchers, which provide this functionality.
1039		1191
1040	You I<must not> change the priority of a watcher as long as it is active or	1192	You I<must not> change the priority of a watcher as long as it is active or
1041	pending.	1193	pending.
1042
1043	The default priority used by watchers when no priority has been set is
1044	always C<0>, which is supposed to not be too high and not be too low :).
1045		1194
1046	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is	1195	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
1047	fine, as long as you do not mind that the priority value you query might	1196	fine, as long as you do not mind that the priority value you query might
1048	or might not have been clamped to the valid range.	1197	or might not have been clamped to the valid range.
		1198
		1199	The default priority used by watchers when no priority has been set is
		1200	always C<0>, which is supposed to not be too high and not be too low :).
		1201
		1202	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1203	priorities.
1049		1204
1050	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1205	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
1051		1206
1052	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1207	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1053	C<loop> nor C<revents> need to be valid as long as the watcher callback	1208	C<loop> nor C<revents> need to be valid as long as the watcher callback
…		…
1118	#include <stddef.h>	1273	#include <stddef.h>
1119		1274
1120	static void	1275	static void
1121	t1_cb (EV_P_ ev_timer *w, int revents)	1276	t1_cb (EV_P_ ev_timer *w, int revents)
1122	{	1277	{
1123	struct my_biggy big = (struct my_biggy *	1278	struct my_biggy big = (struct my_biggy *)
1124	(((char *)w) - offsetof (struct my_biggy, t1));	1279	(((char *)w) - offsetof (struct my_biggy, t1));
1125	}	1280	}
1126		1281
1127	static void	1282	static void
1128	t2_cb (EV_P_ ev_timer *w, int revents)	1283	t2_cb (EV_P_ ev_timer *w, int revents)
1129	{	1284	{
1130	struct my_biggy big = (struct my_biggy *	1285	struct my_biggy big = (struct my_biggy *)
1131	(((char *)w) - offsetof (struct my_biggy, t2));	1286	(((char *)w) - offsetof (struct my_biggy, t2));
1132	}	1287	}
		1288
		1289	=head2 WATCHER PRIORITY MODELS
		1290
		1291	Many event loops support I<watcher priorities>, which are usually small
		1292	integers that influence the ordering of event callback invocation
		1293	between watchers in some way, all else being equal.
		1294
		1295	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1296	description for the more technical details such as the actual priority
		1297	range.
		1298
		1299	There are two common ways how these these priorities are being interpreted
		1300	by event loops:
		1301
		1302	In the more common lock-out model, higher priorities "lock out" invocation
		1303	of lower priority watchers, which means as long as higher priority
		1304	watchers receive events, lower priority watchers are not being invoked.
		1305
		1306	The less common only-for-ordering model uses priorities solely to order
		1307	callback invocation within a single event loop iteration: Higher priority
		1308	watchers are invoked before lower priority ones, but they all get invoked
		1309	before polling for new events.
		1310
		1311	Libev uses the second (only-for-ordering) model for all its watchers
		1312	except for idle watchers (which use the lock-out model).
		1313
		1314	The rationale behind this is that implementing the lock-out model for
		1315	watchers is not well supported by most kernel interfaces, and most event
		1316	libraries will just poll for the same events again and again as long as
		1317	their callbacks have not been executed, which is very inefficient in the
		1318	common case of one high-priority watcher locking out a mass of lower
		1319	priority ones.
		1320
		1321	Static (ordering) priorities are most useful when you have two or more
		1322	watchers handling the same resource: a typical usage example is having an
		1323	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1324	timeouts. Under load, data might be received while the program handles
		1325	other jobs, but since timers normally get invoked first, the timeout
		1326	handler will be executed before checking for data. In that case, giving
		1327	the timer a lower priority than the I/O watcher ensures that I/O will be
		1328	handled first even under adverse conditions (which is usually, but not
		1329	always, what you want).
		1330
		1331	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1332	will only be executed when no same or higher priority watchers have
		1333	received events, they can be used to implement the "lock-out" model when
		1334	required.
		1335
		1336	For example, to emulate how many other event libraries handle priorities,
		1337	you can associate an C<ev_idle> watcher to each such watcher, and in
		1338	the normal watcher callback, you just start the idle watcher. The real
		1339	processing is done in the idle watcher callback. This causes libev to
		1340	continously poll and process kernel event data for the watcher, but when
		1341	the lock-out case is known to be rare (which in turn is rare :), this is
		1342	workable.
		1343
		1344	Usually, however, the lock-out model implemented that way will perform
		1345	miserably under the type of load it was designed to handle. In that case,
		1346	it might be preferable to stop the real watcher before starting the
		1347	idle watcher, so the kernel will not have to process the event in case
		1348	the actual processing will be delayed for considerable time.
		1349
		1350	Here is an example of an I/O watcher that should run at a strictly lower
		1351	priority than the default, and which should only process data when no
		1352	other events are pending:
		1353
		1354	ev_idle idle; // actual processing watcher
		1355	ev_io io; // actual event watcher
		1356
		1357	static void
		1358	io_cb (EV_P_ ev_io *w, int revents)
		1359	{
		1360	// stop the I/O watcher, we received the event, but
		1361	// are not yet ready to handle it.
		1362	ev_io_stop (EV_A_ w);
		1363
		1364	// start the idle watcher to ahndle the actual event.
		1365	// it will not be executed as long as other watchers
		1366	// with the default priority are receiving events.
		1367	ev_idle_start (EV_A_ &idle);
		1368	}
		1369
		1370	static void
		1371	idle_cb (EV_P_ ev_idle *w, int revents)
		1372	{
		1373	// actual processing
		1374	read (STDIN_FILENO, ...);
		1375
		1376	// have to start the I/O watcher again, as
		1377	// we have handled the event
		1378	ev_io_start (EV_P_ &io);
		1379	}
		1380
		1381	// initialisation
		1382	ev_idle_init (&idle, idle_cb);
		1383	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1384	ev_io_start (EV_DEFAULT_ &io);
		1385
		1386	In the "real" world, it might also be beneficial to start a timer, so that
		1387	low-priority connections can not be locked out forever under load. This
		1388	enables your program to keep a lower latency for important connections
		1389	during short periods of high load, while not completely locking out less
		1390	important ones.
1133		1391
1134		1392
1135	=head1 WATCHER TYPES	1393	=head1 WATCHER TYPES
1136		1394
1137	This section describes each watcher in detail, but will not repeat	1395	This section describes each watcher in detail, but will not repeat
…		…
1163	descriptors to non-blocking mode is also usually a good idea (but not	1421	descriptors to non-blocking mode is also usually a good idea (but not
1164	required if you know what you are doing).	1422	required if you know what you are doing).
1165		1423
1166	If you cannot use non-blocking mode, then force the use of a	1424	If you cannot use non-blocking mode, then force the use of a
1167	known-to-be-good backend (at the time of this writing, this includes only	1425	known-to-be-good backend (at the time of this writing, this includes only
1168	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).	1426	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
		1427	descriptors for which non-blocking operation makes no sense (such as
		1428	files) - libev doesn't guarentee any specific behaviour in that case.
1169		1429
1170	Another thing you have to watch out for is that it is quite easy to	1430	Another thing you have to watch out for is that it is quite easy to
1171	receive "spurious" readiness notifications, that is your callback might	1431	receive "spurious" readiness notifications, that is your callback might
1172	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1432	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1173	because there is no data. Not only are some backends known to create a	1433	because there is no data. Not only are some backends known to create a
…		…
1294	year, it will still time out after (roughly) one hour. "Roughly" because	1554	year, it will still time out after (roughly) one hour. "Roughly" because
1295	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1555	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1296	monotonic clock option helps a lot here).	1556	monotonic clock option helps a lot here).
1297		1557
1298	The callback is guaranteed to be invoked only I<after> its timeout has	1558	The callback is guaranteed to be invoked only I<after> its timeout has
1299	passed, but if multiple timers become ready during the same loop iteration	1559	passed (not I<at>, so on systems with very low-resolution clocks this
1300	then order of execution is undefined.	1560	might introduce a small delay). If multiple timers become ready during the
		1561	same loop iteration then the ones with earlier time-out values are invoked
		1562	before ones of the same priority with later time-out values (but this is
		1563	no longer true when a callback calls C<ev_loop> recursively).
1301		1564
1302	=head3 Be smart about timeouts	1565	=head3 Be smart about timeouts
1303		1566
1304	Many real-world problems involve some kind of timeout, usually for error	1567	Many real-world problems involve some kind of timeout, usually for error
1305	recovery. A typical example is an HTTP request - if the other side hangs,	1568	recovery. A typical example is an HTTP request - if the other side hangs,
…		…
1349	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>	1612	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
1350	member and C<ev_timer_again>.	1613	member and C<ev_timer_again>.
1351		1614
1352	At start:	1615	At start:
1353		1616
1354	ev_timer_init (timer, callback);	1617	ev_init (timer, callback);
1355	timer->repeat = 60.;	1618	timer->repeat = 60.;
1356	ev_timer_again (loop, timer);	1619	ev_timer_again (loop, timer);
1357		1620
1358	Each time there is some activity:	1621	Each time there is some activity:
1359		1622
…		…
1398	else	1661	else
1399	{	1662	{
1400	// callback was invoked, but there was some activity, re-arm	1663	// callback was invoked, but there was some activity, re-arm
1401	// the watcher to fire in last_activity + 60, which is	1664	// the watcher to fire in last_activity + 60, which is
1402	// guaranteed to be in the future, so "again" is positive:	1665	// guaranteed to be in the future, so "again" is positive:
1403	w->again = timeout - now;	1666	w->repeat = timeout - now;
1404	ev_timer_again (EV_A_ w);	1667	ev_timer_again (EV_A_ w);
1405	}	1668	}
1406	}	1669	}
1407		1670
1408	To summarise the callback: first calculate the real timeout (defined	1671	To summarise the callback: first calculate the real timeout (defined
…		…
1421		1684
1422	To start the timer, simply initialise the watcher and set C<last_activity>	1685	To start the timer, simply initialise the watcher and set C<last_activity>
1423	to the current time (meaning we just have some activity :), then call the	1686	to the current time (meaning we just have some activity :), then call the
1424	callback, which will "do the right thing" and start the timer:	1687	callback, which will "do the right thing" and start the timer:
1425		1688
1426	ev_timer_init (timer, callback);	1689	ev_init (timer, callback);
1427	last_activity = ev_now (loop);	1690	last_activity = ev_now (loop);
1428	callback (loop, timer, EV_TIMEOUT);	1691	callback (loop, timer, EV_TIMEOUT);
1429		1692
1430	And when there is some activity, simply store the current time in	1693	And when there is some activity, simply store the current time in
1431	C<last_activity>, no libev calls at all:	1694	C<last_activity>, no libev calls at all:
…		…
1492		1755
1493	If the event loop is suspended for a long time, you can also force an	1756	If the event loop is suspended for a long time, you can also force an
1494	update of the time returned by C<ev_now ()> by calling C<ev_now_update	1757	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1495	()>.	1758	()>.
1496		1759
		1760	=head3 The special problems of suspended animation
		1761
		1762	When you leave the server world it is quite customary to hit machines that
		1763	can suspend/hibernate - what happens to the clocks during such a suspend?
		1764
		1765	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1766	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1767	to run until the system is suspended, but they will not advance while the
		1768	system is suspended. That means, on resume, it will be as if the program
		1769	was frozen for a few seconds, but the suspend time will not be counted
		1770	towards C<ev_timer> when a monotonic clock source is used. The real time
		1771	clock advanced as expected, but if it is used as sole clocksource, then a
		1772	long suspend would be detected as a time jump by libev, and timers would
		1773	be adjusted accordingly.
		1774
		1775	I would not be surprised to see different behaviour in different between
		1776	operating systems, OS versions or even different hardware.
		1777
		1778	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1779	time jump in the monotonic clocks and the realtime clock. If the program
		1780	is suspended for a very long time, and monotonic clock sources are in use,
		1781	then you can expect C<ev_timer>s to expire as the full suspension time
		1782	will be counted towards the timers. When no monotonic clock source is in
		1783	use, then libev will again assume a timejump and adjust accordingly.
		1784
		1785	It might be beneficial for this latter case to call C<ev_suspend>
		1786	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1787	deterministic behaviour in this case (you can do nothing against
		1788	C<SIGSTOP>).
		1789
1497	=head3 Watcher-Specific Functions and Data Members	1790	=head3 Watcher-Specific Functions and Data Members
1498		1791
1499	=over 4	1792	=over 4
1500		1793
1501	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1794	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
…		…
1524	If the timer is started but non-repeating, stop it (as if it timed out).	1817	If the timer is started but non-repeating, stop it (as if it timed out).
1525		1818
1526	If the timer is repeating, either start it if necessary (with the	1819	If the timer is repeating, either start it if necessary (with the
1527	C<repeat> value), or reset the running timer to the C<repeat> value.	1820	C<repeat> value), or reset the running timer to the C<repeat> value.
1528		1821
1529	This sounds a bit complicated, see "Be smart about timeouts", above, for a	1822	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
1530	usage example.	1823	usage example.
		1824
		1825	=item ev_timer_remaining (loop, ev_timer *)
		1826
		1827	Returns the remaining time until a timer fires. If the timer is active,
		1828	then this time is relative to the current event loop time, otherwise it's
		1829	the timeout value currently configured.
		1830
		1831	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1832	C<5>. When the timer is started and one second passes, C<ev_timer_remain>
		1833	will return C<4>. When the timer expires and is restarted, it will return
		1834	roughly C<7> (likely slightly less as callback invocation takes some time,
		1835	too), and so on.
1531		1836
1532	=item ev_tstamp repeat [read-write]	1837	=item ev_tstamp repeat [read-write]
1533		1838
1534	The current C<repeat> value. Will be used each time the watcher times out	1839	The current C<repeat> value. Will be used each time the watcher times out
1535	or C<ev_timer_again> is called, and determines the next timeout (if any),	1840	or C<ev_timer_again> is called, and determines the next timeout (if any),
…		…
1573	=head2 C<ev_periodic> - to cron or not to cron?	1878	=head2 C<ev_periodic> - to cron or not to cron?
1574		1879
1575	Periodic watchers are also timers of a kind, but they are very versatile	1880	Periodic watchers are also timers of a kind, but they are very versatile
1576	(and unfortunately a bit complex).	1881	(and unfortunately a bit complex).
1577		1882
1578	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1883	Unlike C<ev_timer>, periodic watchers are not based on real time (or
1579	but on wall clock time (absolute time). You can tell a periodic watcher	1884	relative time, the physical time that passes) but on wall clock time
1580	to trigger after some specific point in time. For example, if you tell a	1885	(absolute time, the thing you can read on your calender or clock). The
1581	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()	1886	difference is that wall clock time can run faster or slower than real
1582	+ 10.>, that is, an absolute time not a delay) and then reset your system	1887	time, and time jumps are not uncommon (e.g. when you adjust your
1583	clock to January of the previous year, then it will take more than year	1888	wrist-watch).
1584	to trigger the event (unlike an C<ev_timer>, which would still trigger
1585	roughly 10 seconds later as it uses a relative timeout).
1586		1889
		1890	You can tell a periodic watcher to trigger after some specific point
		1891	in time: for example, if you tell a periodic watcher to trigger "in 10
		1892	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1893	not a delay) and then reset your system clock to January of the previous
		1894	year, then it will take a year or more to trigger the event (unlike an
		1895	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1896	it, as it uses a relative timeout).
		1897
1587	C<ev_periodic>s can also be used to implement vastly more complex timers,	1898	C<ev_periodic> watchers can also be used to implement vastly more complex
1588	such as triggering an event on each "midnight, local time", or other	1899	timers, such as triggering an event on each "midnight, local time", or
1589	complicated rules.	1900	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1901	those cannot react to time jumps.
1590		1902
1591	As with timers, the callback is guaranteed to be invoked only when the	1903	As with timers, the callback is guaranteed to be invoked only when the
1592	time (C<at>) has passed, but if multiple periodic timers become ready	1904	point in time where it is supposed to trigger has passed. If multiple
1593	during the same loop iteration, then order of execution is undefined.	1905	timers become ready during the same loop iteration then the ones with
		1906	earlier time-out values are invoked before ones with later time-out values
		1907	(but this is no longer true when a callback calls C<ev_loop> recursively).
1594		1908
1595	=head3 Watcher-Specific Functions and Data Members	1909	=head3 Watcher-Specific Functions and Data Members
1596		1910
1597	=over 4	1911	=over 4
1598		1912
1599	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1913	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1600		1914
1601	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1915	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1602		1916
1603	Lots of arguments, lets sort it out... There are basically three modes of	1917	Lots of arguments, let's sort it out... There are basically three modes of
1604	operation, and we will explain them from simplest to most complex:	1918	operation, and we will explain them from simplest to most complex:
1605		1919
1606	=over 4	1920	=over 4
1607		1921
1608	=item * absolute timer (at = time, interval = reschedule_cb = 0)	1922	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
1609		1923
1610	In this configuration the watcher triggers an event after the wall clock	1924	In this configuration the watcher triggers an event after the wall clock
1611	time C<at> has passed. It will not repeat and will not adjust when a time	1925	time C<offset> has passed. It will not repeat and will not adjust when a
1612	jump occurs, that is, if it is to be run at January 1st 2011 then it will	1926	time jump occurs, that is, if it is to be run at January 1st 2011 then it
1613	only run when the system clock reaches or surpasses this time.	1927	will be stopped and invoked when the system clock reaches or surpasses
		1928	this point in time.
1614		1929
1615	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	1930	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
1616		1931
1617	In this mode the watcher will always be scheduled to time out at the next	1932	In this mode the watcher will always be scheduled to time out at the next
1618	C<at + N * interval> time (for some integer N, which can also be negative)	1933	C<offset + N * interval> time (for some integer N, which can also be
1619	and then repeat, regardless of any time jumps.	1934	negative) and then repeat, regardless of any time jumps. The C<offset>
		1935	argument is merely an offset into the C<interval> periods.
1620		1936
1621	This can be used to create timers that do not drift with respect to the	1937	This can be used to create timers that do not drift with respect to the
1622	system clock, for example, here is a C<ev_periodic> that triggers each	1938	system clock, for example, here is an C<ev_periodic> that triggers each
1623	hour, on the hour:	1939	hour, on the hour (with respect to UTC):
1624		1940
1625	ev_periodic_set (&periodic, 0., 3600., 0);	1941	ev_periodic_set (&periodic, 0., 3600., 0);
1626		1942
1627	This doesn't mean there will always be 3600 seconds in between triggers,	1943	This doesn't mean there will always be 3600 seconds in between triggers,
1628	but only that the callback will be called when the system time shows a	1944	but only that the callback will be called when the system time shows a
1629	full hour (UTC), or more correctly, when the system time is evenly divisible	1945	full hour (UTC), or more correctly, when the system time is evenly divisible
1630	by 3600.	1946	by 3600.
1631		1947
1632	Another way to think about it (for the mathematically inclined) is that	1948	Another way to think about it (for the mathematically inclined) is that
1633	C<ev_periodic> will try to run the callback in this mode at the next possible	1949	C<ev_periodic> will try to run the callback in this mode at the next possible
1634	time where C<time = at (mod interval)>, regardless of any time jumps.	1950	time where C<time = offset (mod interval)>, regardless of any time jumps.
1635		1951
1636	For numerical stability it is preferable that the C<at> value is near	1952	For numerical stability it is preferable that the C<offset> value is near
1637	C<ev_now ()> (the current time), but there is no range requirement for	1953	C<ev_now ()> (the current time), but there is no range requirement for
1638	this value, and in fact is often specified as zero.	1954	this value, and in fact is often specified as zero.
1639		1955
1640	Note also that there is an upper limit to how often a timer can fire (CPU	1956	Note also that there is an upper limit to how often a timer can fire (CPU
1641	speed for example), so if C<interval> is very small then timing stability	1957	speed for example), so if C<interval> is very small then timing stability
1642	will of course deteriorate. Libev itself tries to be exact to be about one	1958	will of course deteriorate. Libev itself tries to be exact to be about one
1643	millisecond (if the OS supports it and the machine is fast enough).	1959	millisecond (if the OS supports it and the machine is fast enough).
1644		1960
1645	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)	1961	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
1646		1962
1647	In this mode the values for C<interval> and C<at> are both being	1963	In this mode the values for C<interval> and C<offset> are both being
1648	ignored. Instead, each time the periodic watcher gets scheduled, the	1964	ignored. Instead, each time the periodic watcher gets scheduled, the
1649	reschedule callback will be called with the watcher as first, and the	1965	reschedule callback will be called with the watcher as first, and the
1650	current time as second argument.	1966	current time as second argument.
1651		1967
1652	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	1968	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
1653	ever, or make ANY event loop modifications whatsoever>.	1969	or make ANY other event loop modifications whatsoever, unless explicitly
		1970	allowed by documentation here>.
1654		1971
1655	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop	1972	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
1656	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the	1973	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
1657	only event loop modification you are allowed to do).	1974	only event loop modification you are allowed to do).
1658		1975
…		…
1688	a different time than the last time it was called (e.g. in a crond like	2005	a different time than the last time it was called (e.g. in a crond like
1689	program when the crontabs have changed).	2006	program when the crontabs have changed).
1690		2007
1691	=item ev_tstamp ev_periodic_at (ev_periodic *)	2008	=item ev_tstamp ev_periodic_at (ev_periodic *)
1692		2009
1693	When active, returns the absolute time that the watcher is supposed to	2010	When active, returns the absolute time that the watcher is supposed
1694	trigger next.	2011	to trigger next. This is not the same as the C<offset> argument to
		2012	C<ev_periodic_set>, but indeed works even in interval and manual
		2013	rescheduling modes.
1695		2014
1696	=item ev_tstamp offset [read-write]	2015	=item ev_tstamp offset [read-write]
1697		2016
1698	When repeating, this contains the offset value, otherwise this is the	2017	When repeating, this contains the offset value, otherwise this is the
1699	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	2018	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2019	although libev might modify this value for better numerical stability).
1700		2020
1701	Can be modified any time, but changes only take effect when the periodic	2021	Can be modified any time, but changes only take effect when the periodic
1702	timer fires or C<ev_periodic_again> is being called.	2022	timer fires or C<ev_periodic_again> is being called.
1703		2023
1704	=item ev_tstamp interval [read-write]	2024	=item ev_tstamp interval [read-write]
…		…
1756	Signal watchers will trigger an event when the process receives a specific	2076	Signal watchers will trigger an event when the process receives a specific
1757	signal one or more times. Even though signals are very asynchronous, libev	2077	signal one or more times. Even though signals are very asynchronous, libev
1758	will try it's best to deliver signals synchronously, i.e. as part of the	2078	will try it's best to deliver signals synchronously, i.e. as part of the
1759	normal event processing, like any other event.	2079	normal event processing, like any other event.
1760		2080
		2081	Note that only the default loop supports registering signal watchers
		2082	currently.
		2083
1761	If you want signals asynchronously, just use C<sigaction> as you would	2084	If you want signals asynchronously, just use C<sigaction> as you would
1762	do without libev and forget about sharing the signal. You can even use	2085	do without libev and forget about sharing the signal. You can even use
1763	C<ev_async> from a signal handler to synchronously wake up an event loop.	2086	C<ev_async> from a signal handler to synchronously wake up an event loop.
1764		2087
1765	You can configure as many watchers as you like per signal. Only when the	2088	You can configure as many watchers as you like per signal. Only when the
1766	first watcher gets started will libev actually register a signal handler	2089	first watcher gets started will libev actually register something with
1767	with the kernel (thus it coexists with your own signal handlers as long as	2090	the kernel (thus it coexists with your own signal handlers as long as you
1768	you don't register any with libev for the same signal). Similarly, when	2091	don't register any with libev for the same signal).
1769	the last signal watcher for a signal is stopped, libev will reset the	2092
1770	signal handler to SIG_DFL (regardless of what it was set to before).	2093	Both the signal mask state (C<sigprocmask>) and the signal handler state
		2094	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2095	sotpping it again), that is, libev might or might not block the signal,
		2096	and might or might not set or restore the installed signal handler.
1771		2097
1772	If possible and supported, libev will install its handlers with	2098	If possible and supported, libev will install its handlers with
1773	C<SA_RESTART> behaviour enabled, so system calls should not be unduly	2099	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
1774	interrupted. If you have a problem with system calls getting interrupted by	2100	not be unduly interrupted. If you have a problem with system calls getting
1775	signals you can block all signals in an C<ev_check> watcher and unblock	2101	interrupted by signals you can block all signals in an C<ev_check> watcher
1776	them in an C<ev_prepare> watcher.	2102	and unblock them in an C<ev_prepare> watcher.
1777		2103
1778	=head3 Watcher-Specific Functions and Data Members	2104	=head3 Watcher-Specific Functions and Data Members
1779		2105
1780	=over 4	2106	=over 4
1781		2107
…		…
1813	some child status changes (most typically when a child of yours dies or	2139	some child status changes (most typically when a child of yours dies or
1814	exits). It is permissible to install a child watcher I<after> the child	2140	exits). It is permissible to install a child watcher I<after> the child
1815	has been forked (which implies it might have already exited), as long	2141	has been forked (which implies it might have already exited), as long
1816	as the event loop isn't entered (or is continued from a watcher), i.e.,	2142	as the event loop isn't entered (or is continued from a watcher), i.e.,
1817	forking and then immediately registering a watcher for the child is fine,	2143	forking and then immediately registering a watcher for the child is fine,
1818	but forking and registering a watcher a few event loop iterations later is	2144	but forking and registering a watcher a few event loop iterations later or
1819	not.	2145	in the next callback invocation is not.
1820		2146
1821	Only the default event loop is capable of handling signals, and therefore	2147	Only the default event loop is capable of handling signals, and therefore
1822	you can only register child watchers in the default event loop.	2148	you can only register child watchers in the default event loop.
1823		2149
		2150	Due to some design glitches inside libev, child watchers will always be
		2151	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2152	libev)
		2153
1824	=head3 Process Interaction	2154	=head3 Process Interaction
1825		2155
1826	Libev grabs C<SIGCHLD> as soon as the default event loop is	2156	Libev grabs C<SIGCHLD> as soon as the default event loop is
1827	initialised. This is necessary to guarantee proper behaviour even if	2157	initialised. This is necessary to guarantee proper behaviour even if the
1828	the first child watcher is started after the child exits. The occurrence	2158	first child watcher is started after the child exits. The occurrence
1829	of C<SIGCHLD> is recorded asynchronously, but child reaping is done	2159	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
1830	synchronously as part of the event loop processing. Libev always reaps all	2160	synchronously as part of the event loop processing. Libev always reaps all
1831	children, even ones not watched.	2161	children, even ones not watched.
1832		2162
1833	=head3 Overriding the Built-In Processing	2163	=head3 Overriding the Built-In Processing
…		…
1843	=head3 Stopping the Child Watcher	2173	=head3 Stopping the Child Watcher
1844		2174
1845	Currently, the child watcher never gets stopped, even when the	2175	Currently, the child watcher never gets stopped, even when the
1846	child terminates, so normally one needs to stop the watcher in the	2176	child terminates, so normally one needs to stop the watcher in the
1847	callback. Future versions of libev might stop the watcher automatically	2177	callback. Future versions of libev might stop the watcher automatically
1848	when a child exit is detected.	2178	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2179	problem).
1849		2180
1850	=head3 Watcher-Specific Functions and Data Members	2181	=head3 Watcher-Specific Functions and Data Members
1851		2182
1852	=over 4	2183	=over 4
1853		2184
…		…
1910		2241
1911		2242
1912	=head2 C<ev_stat> - did the file attributes just change?	2243	=head2 C<ev_stat> - did the file attributes just change?
1913		2244
1914	This watches a file system path for attribute changes. That is, it calls	2245	This watches a file system path for attribute changes. That is, it calls
1915	C<stat> regularly (or when the OS says it changed) and sees if it changed	2246	C<stat> on that path in regular intervals (or when the OS says it changed)
1916	compared to the last time, invoking the callback if it did.	2247	and sees if it changed compared to the last time, invoking the callback if
		2248	it did.
1917		2249
1918	The path does not need to exist: changing from "path exists" to "path does	2250	The path does not need to exist: changing from "path exists" to "path does
1919	not exist" is a status change like any other. The condition "path does	2251	not exist" is a status change like any other. The condition "path does not
1920	not exist" is signified by the C<st_nlink> field being zero (which is	2252	exist" (or more correctly "path cannot be stat'ed") is signified by the
1921	otherwise always forced to be at least one) and all the other fields of	2253	C<st_nlink> field being zero (which is otherwise always forced to be at
1922	the stat buffer having unspecified contents.	2254	least one) and all the other fields of the stat buffer having unspecified
		2255	contents.
1923		2256
1924	The path I<should> be absolute and I<must not> end in a slash. If it is	2257	The path I<must not> end in a slash or contain special components such as
		2258	C<.> or C<..>. The path I<should> be absolute: If it is relative and
1925	relative and your working directory changes, the behaviour is undefined.	2259	your working directory changes, then the behaviour is undefined.
1926		2260
1927	Since there is no standard kernel interface to do this, the portable	2261	Since there is no portable change notification interface available, the
1928	implementation simply calls C<stat (2)> regularly on the path to see if	2262	portable implementation simply calls C<stat(2)> regularly on the path
1929	it changed somehow. You can specify a recommended polling interval for	2263	to see if it changed somehow. You can specify a recommended polling
1930	this case. If you specify a polling interval of C<0> (highly recommended!)	2264	interval for this case. If you specify a polling interval of C<0> (highly
1931	then a I<suitable, unspecified default> value will be used (which	2265	recommended!) then a I<suitable, unspecified default> value will be used
1932	you can expect to be around five seconds, although this might change	2266	(which you can expect to be around five seconds, although this might
1933	dynamically). Libev will also impose a minimum interval which is currently	2267	change dynamically). Libev will also impose a minimum interval which is
1934	around C<0.1>, but thats usually overkill.	2268	currently around C<0.1>, but that's usually overkill.
1935		2269
1936	This watcher type is not meant for massive numbers of stat watchers,	2270	This watcher type is not meant for massive numbers of stat watchers,
1937	as even with OS-supported change notifications, this can be	2271	as even with OS-supported change notifications, this can be
1938	resource-intensive.	2272	resource-intensive.
1939		2273
1940	At the time of this writing, the only OS-specific interface implemented	2274	At the time of this writing, the only OS-specific interface implemented
1941	is the Linux inotify interface (implementing kqueue support is left as	2275	is the Linux inotify interface (implementing kqueue support is left as an
1942	an exercise for the reader. Note, however, that the author sees no way	2276	exercise for the reader. Note, however, that the author sees no way of
1943	of implementing C<ev_stat> semantics with kqueue).	2277	implementing C<ev_stat> semantics with kqueue, except as a hint).
1944		2278
1945	=head3 ABI Issues (Largefile Support)	2279	=head3 ABI Issues (Largefile Support)
1946		2280
1947	Libev by default (unless the user overrides this) uses the default	2281	Libev by default (unless the user overrides this) uses the default
1948	compilation environment, which means that on systems with large file	2282	compilation environment, which means that on systems with large file
1949	support disabled by default, you get the 32 bit version of the stat	2283	support disabled by default, you get the 32 bit version of the stat
1950	structure. When using the library from programs that change the ABI to	2284	structure. When using the library from programs that change the ABI to
1951	use 64 bit file offsets the programs will fail. In that case you have to	2285	use 64 bit file offsets the programs will fail. In that case you have to
1952	compile libev with the same flags to get binary compatibility. This is	2286	compile libev with the same flags to get binary compatibility. This is
1953	obviously the case with any flags that change the ABI, but the problem is	2287	obviously the case with any flags that change the ABI, but the problem is
1954	most noticeably disabled with ev_stat and large file support.	2288	most noticeably displayed with ev_stat and large file support.
1955		2289
1956	The solution for this is to lobby your distribution maker to make large	2290	The solution for this is to lobby your distribution maker to make large
1957	file interfaces available by default (as e.g. FreeBSD does) and not	2291	file interfaces available by default (as e.g. FreeBSD does) and not
1958	optional. Libev cannot simply switch on large file support because it has	2292	optional. Libev cannot simply switch on large file support because it has
1959	to exchange stat structures with application programs compiled using the	2293	to exchange stat structures with application programs compiled using the
1960	default compilation environment.	2294	default compilation environment.
1961		2295
1962	=head3 Inotify and Kqueue	2296	=head3 Inotify and Kqueue
1963		2297
1964	When C<inotify (7)> support has been compiled into libev (generally	2298	When C<inotify (7)> support has been compiled into libev and present at
1965	only available with Linux 2.6.25 or above due to bugs in earlier	2299	runtime, it will be used to speed up change detection where possible. The
1966	implementations) and present at runtime, it will be used to speed up	2300	inotify descriptor will be created lazily when the first C<ev_stat>
1967	change detection where possible. The inotify descriptor will be created	2301	watcher is being started.
1968	lazily when the first C<ev_stat> watcher is being started.
1969		2302
1970	Inotify presence does not change the semantics of C<ev_stat> watchers	2303	Inotify presence does not change the semantics of C<ev_stat> watchers
1971	except that changes might be detected earlier, and in some cases, to avoid	2304	except that changes might be detected earlier, and in some cases, to avoid
1972	making regular C<stat> calls. Even in the presence of inotify support	2305	making regular C<stat> calls. Even in the presence of inotify support
1973	there are many cases where libev has to resort to regular C<stat> polling,	2306	there are many cases where libev has to resort to regular C<stat> polling,
1974	but as long as the path exists, libev usually gets away without polling.	2307	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2308	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2309	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2310	xfs are fully working) libev usually gets away without polling.
1975		2311
1976	There is no support for kqueue, as apparently it cannot be used to	2312	There is no support for kqueue, as apparently it cannot be used to
1977	implement this functionality, due to the requirement of having a file	2313	implement this functionality, due to the requirement of having a file
1978	descriptor open on the object at all times, and detecting renames, unlinks	2314	descriptor open on the object at all times, and detecting renames, unlinks
1979	etc. is difficult.	2315	etc. is difficult.
1980		2316
		2317	=head3 C<stat ()> is a synchronous operation
		2318
		2319	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2320	the process. The exception are C<ev_stat> watchers - those call C<stat
		2321	()>, which is a synchronous operation.
		2322
		2323	For local paths, this usually doesn't matter: unless the system is very
		2324	busy or the intervals between stat's are large, a stat call will be fast,
		2325	as the path data is usually in memory already (except when starting the
		2326	watcher).
		2327
		2328	For networked file systems, calling C<stat ()> can block an indefinite
		2329	time due to network issues, and even under good conditions, a stat call
		2330	often takes multiple milliseconds.
		2331
		2332	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2333	paths, although this is fully supported by libev.
		2334
1981	=head3 The special problem of stat time resolution	2335	=head3 The special problem of stat time resolution
1982		2336
1983	The C<stat ()> system call only supports full-second resolution portably, and	2337	The C<stat ()> system call only supports full-second resolution portably,
1984	even on systems where the resolution is higher, most file systems still	2338	and even on systems where the resolution is higher, most file systems
1985	only support whole seconds.	2339	still only support whole seconds.
1986		2340
1987	That means that, if the time is the only thing that changes, you can	2341	That means that, if the time is the only thing that changes, you can
1988	easily miss updates: on the first update, C<ev_stat> detects a change and	2342	easily miss updates: on the first update, C<ev_stat> detects a change and
1989	calls your callback, which does something. When there is another update	2343	calls your callback, which does something. When there is another update
1990	within the same second, C<ev_stat> will be unable to detect unless the	2344	within the same second, C<ev_stat> will be unable to detect unless the
…		…
2133		2487
2134	=head3 Watcher-Specific Functions and Data Members	2488	=head3 Watcher-Specific Functions and Data Members
2135		2489
2136	=over 4	2490	=over 4
2137		2491
2138	=item ev_idle_init (ev_signal *, callback)	2492	=item ev_idle_init (ev_idle *, callback)
2139		2493
2140	Initialises and configures the idle watcher - it has no parameters of any	2494	Initialises and configures the idle watcher - it has no parameters of any
2141	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2495	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
2142	believe me.	2496	believe me.
2143		2497
…		…
2156	// no longer anything immediate to do.	2510	// no longer anything immediate to do.
2157	}	2511	}
2158		2512
2159	ev_idle *idle_watcher = malloc (sizeof (ev_idle));	2513	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
2160	ev_idle_init (idle_watcher, idle_cb);	2514	ev_idle_init (idle_watcher, idle_cb);
2161	ev_idle_start (loop, idle_cb);	2515	ev_idle_start (loop, idle_watcher);
2162		2516
2163		2517
2164	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	2518	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2165		2519
2166	Prepare and check watchers are usually (but not always) used in pairs:	2520	Prepare and check watchers are usually (but not always) used in pairs:
…		…
2259	struct pollfd fds [nfd];	2613	struct pollfd fds [nfd];
2260	// actual code will need to loop here and realloc etc.	2614	// actual code will need to loop here and realloc etc.
2261	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));	2615	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2262		2616
2263	/* the callback is illegal, but won't be called as we stop during check */	2617	/* the callback is illegal, but won't be called as we stop during check */
2264	ev_timer_init (&tw, 0, timeout * 1e-3);	2618	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
2265	ev_timer_start (loop, &tw);	2619	ev_timer_start (loop, &tw);
2266		2620
2267	// create one ev_io per pollfd	2621	// create one ev_io per pollfd
2268	for (int i = 0; i < nfd; ++i)	2622	for (int i = 0; i < nfd; ++i)
2269	{	2623	{
…		…
2382	some fds have to be watched and handled very quickly (with low latency),	2736	some fds have to be watched and handled very quickly (with low latency),
2383	and even priorities and idle watchers might have too much overhead. In	2737	and even priorities and idle watchers might have too much overhead. In
2384	this case you would put all the high priority stuff in one loop and all	2738	this case you would put all the high priority stuff in one loop and all
2385	the rest in a second one, and embed the second one in the first.	2739	the rest in a second one, and embed the second one in the first.
2386		2740
2387	As long as the watcher is active, the callback will be invoked every time	2741	As long as the watcher is active, the callback will be invoked every
2388	there might be events pending in the embedded loop. The callback must then	2742	time there might be events pending in the embedded loop. The callback
2389	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	2743	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
2390	their callbacks (you could also start an idle watcher to give the embedded	2744	sweep and invoke their callbacks (the callback doesn't need to invoke the
2391	loop strictly lower priority for example). You can also set the callback	2745	C<ev_embed_sweep> function directly, it could also start an idle watcher
2392	to C<0>, in which case the embed watcher will automatically execute the	2746	to give the embedded loop strictly lower priority for example).
2393	embedded loop sweep.
2394		2747
2395	As long as the watcher is started it will automatically handle events. The	2748	You can also set the callback to C<0>, in which case the embed watcher
2396	callback will be invoked whenever some events have been handled. You can	2749	will automatically execute the embedded loop sweep whenever necessary.
2397	set the callback to C<0> to avoid having to specify one if you are not
2398	interested in that.
2399		2750
2400	Also, there have not currently been made special provisions for forking:	2751	Fork detection will be handled transparently while the C<ev_embed> watcher
2401	when you fork, you not only have to call C<ev_loop_fork> on both loops,	2752	is active, i.e., the embedded loop will automatically be forked when the
2402	but you will also have to stop and restart any C<ev_embed> watchers	2753	embedding loop forks. In other cases, the user is responsible for calling
2403	yourself - but you can use a fork watcher to handle this automatically,	2754	C<ev_loop_fork> on the embedded loop.
2404	and future versions of libev might do just that.
2405		2755
2406	Unfortunately, not all backends are embeddable: only the ones returned by	2756	Unfortunately, not all backends are embeddable: only the ones returned by
2407	C<ev_embeddable_backends> are, which, unfortunately, does not include any	2757	C<ev_embeddable_backends> are, which, unfortunately, does not include any
2408	portable one.	2758	portable one.
2409		2759
…		…
2503	event loop blocks next and before C<ev_check> watchers are being called,	2853	event loop blocks next and before C<ev_check> watchers are being called,
2504	and only in the child after the fork. If whoever good citizen calling	2854	and only in the child after the fork. If whoever good citizen calling
2505	C<ev_default_fork> cheats and calls it in the wrong process, the fork	2855	C<ev_default_fork> cheats and calls it in the wrong process, the fork
2506	handlers will be invoked, too, of course.	2856	handlers will be invoked, too, of course.
2507		2857
		2858	=head3 The special problem of life after fork - how is it possible?
		2859
		2860	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2861	up/change the process environment, followed by a call to C<exec()>. This
		2862	sequence should be handled by libev without any problems.
		2863
		2864	This changes when the application actually wants to do event handling
		2865	in the child, or both parent in child, in effect "continuing" after the
		2866	fork.
		2867
		2868	The default mode of operation (for libev, with application help to detect
		2869	forks) is to duplicate all the state in the child, as would be expected
		2870	when I<either> the parent I<or> the child process continues.
		2871
		2872	When both processes want to continue using libev, then this is usually the
		2873	wrong result. In that case, usually one process (typically the parent) is
		2874	supposed to continue with all watchers in place as before, while the other
		2875	process typically wants to start fresh, i.e. without any active watchers.
		2876
		2877	The cleanest and most efficient way to achieve that with libev is to
		2878	simply create a new event loop, which of course will be "empty", and
		2879	use that for new watchers. This has the advantage of not touching more
		2880	memory than necessary, and thus avoiding the copy-on-write, and the
		2881	disadvantage of having to use multiple event loops (which do not support
		2882	signal watchers).
		2883
		2884	When this is not possible, or you want to use the default loop for
		2885	other reasons, then in the process that wants to start "fresh", call
		2886	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2887	the default loop will "orphan" (not stop) all registered watchers, so you
		2888	have to be careful not to execute code that modifies those watchers. Note
		2889	also that in that case, you have to re-register any signal watchers.
		2890
2508	=head3 Watcher-Specific Functions and Data Members	2891	=head3 Watcher-Specific Functions and Data Members
2509		2892
2510	=over 4	2893	=over 4
2511		2894
2512	=item ev_fork_init (ev_signal *, callback)	2895	=item ev_fork_init (ev_signal *, callback)
…		…
2629	=over 4	3012	=over 4
2630		3013
2631	=item ev_async_init (ev_async *, callback)	3014	=item ev_async_init (ev_async *, callback)
2632		3015
2633	Initialises and configures the async watcher - it has no parameters of any	3016	Initialises and configures the async watcher - it has no parameters of any
2634	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,	3017	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
2635	trust me.	3018	trust me.
2636		3019
2637	=item ev_async_send (loop, ev_async *)	3020	=item ev_async_send (loop, ev_async *)
2638		3021
2639	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3022	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2640	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	3023	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2641	C<ev_feed_event>, this call is safe to do from other threads, signal or	3024	C<ev_feed_event>, this call is safe to do from other threads, signal or
2642	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	3025	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
2643	section below on what exactly this means).	3026	section below on what exactly this means).
2644		3027
		3028	Note that, as with other watchers in libev, multiple events might get
		3029	compressed into a single callback invocation (another way to look at this
		3030	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3031	reset when the event loop detects that).
		3032
2645	This call incurs the overhead of a system call only once per loop iteration,	3033	This call incurs the overhead of a system call only once per event loop
2646	so while the overhead might be noticeable, it doesn't apply to repeated	3034	iteration, so while the overhead might be noticeable, it doesn't apply to
2647	calls to C<ev_async_send>.	3035	repeated calls to C<ev_async_send> for the same event loop.
2648		3036
2649	=item bool = ev_async_pending (ev_async *)	3037	=item bool = ev_async_pending (ev_async *)
2650		3038
2651	Returns a non-zero value when C<ev_async_send> has been called on the	3039	Returns a non-zero value when C<ev_async_send> has been called on the
2652	watcher but the event has not yet been processed (or even noted) by the	3040	watcher but the event has not yet been processed (or even noted) by the
…		…
2655	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When	3043	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2656	the loop iterates next and checks for the watcher to have become active,	3044	the loop iterates next and checks for the watcher to have become active,
2657	it will reset the flag again. C<ev_async_pending> can be used to very	3045	it will reset the flag again. C<ev_async_pending> can be used to very
2658	quickly check whether invoking the loop might be a good idea.	3046	quickly check whether invoking the loop might be a good idea.
2659		3047
2660	Not that this does I<not> check whether the watcher itself is pending, only	3048	Not that this does I<not> check whether the watcher itself is pending,
2661	whether it has been requested to make this watcher pending.	3049	only whether it has been requested to make this watcher pending: there
		3050	is a time window between the event loop checking and resetting the async
		3051	notification, and the callback being invoked.
2662		3052
2663	=back	3053	=back
2664		3054
2665		3055
2666	=head1 OTHER FUNCTIONS	3056	=head1 OTHER FUNCTIONS
…		…
2845		3235
2846	myclass obj;	3236	myclass obj;
2847	ev::io iow;	3237	ev::io iow;
2848	iow.set <myclass, &myclass::io_cb> (&obj);	3238	iow.set <myclass, &myclass::io_cb> (&obj);
2849		3239
		3240	=item w->set (object *)
		3241
		3242	This is an B<experimental> feature that might go away in a future version.
		3243
		3244	This is a variation of a method callback - leaving out the method to call
		3245	will default the method to C<operator ()>, which makes it possible to use
		3246	functor objects without having to manually specify the C<operator ()> all
		3247	the time. Incidentally, you can then also leave out the template argument
		3248	list.
		3249
		3250	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3251	int revents)>.
		3252
		3253	See the method-C<set> above for more details.
		3254
		3255	Example: use a functor object as callback.
		3256
		3257	struct myfunctor
		3258	{
		3259	void operator() (ev::io &w, int revents)
		3260	{
		3261	...
		3262	}
		3263	}
		3264
		3265	myfunctor f;
		3266
		3267	ev::io w;
		3268	w.set (&f);
		3269
2850	=item w->set<function> (void *data = 0)	3270	=item w->set<function> (void *data = 0)
2851		3271
2852	Also sets a callback, but uses a static method or plain function as	3272	Also sets a callback, but uses a static method or plain function as
2853	callback. The optional C<data> argument will be stored in the watcher's	3273	callback. The optional C<data> argument will be stored in the watcher's
2854	C<data> member and is free for you to use.	3274	C<data> member and is free for you to use.
…		…
2940	L<http://software.schmorp.de/pkg/EV>.	3360	L<http://software.schmorp.de/pkg/EV>.
2941		3361
2942	=item Python	3362	=item Python
2943		3363
2944	Python bindings can be found at L<http://code.google.com/p/pyev/>. It	3364	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
2945	seems to be quite complete and well-documented. Note, however, that the	3365	seems to be quite complete and well-documented.
2946	patch they require for libev is outright dangerous as it breaks the ABI
2947	for everybody else, and therefore, should never be applied in an installed
2948	libev (if python requires an incompatible ABI then it needs to embed
2949	libev).
2950		3366
2951	=item Ruby	3367	=item Ruby
2952		3368
2953	Tony Arcieri has written a ruby extension that offers access to a subset	3369	Tony Arcieri has written a ruby extension that offers access to a subset
2954	of the libev API and adds file handle abstractions, asynchronous DNS and	3370	of the libev API and adds file handle abstractions, asynchronous DNS and
2955	more on top of it. It can be found via gem servers. Its homepage is at	3371	more on top of it. It can be found via gem servers. Its homepage is at
2956	L<http://rev.rubyforge.org/>.	3372	L<http://rev.rubyforge.org/>.
		3373
		3374	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3375	makes rev work even on mingw.
		3376
		3377	=item Haskell
		3378
		3379	A haskell binding to libev is available at
		3380	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
2957		3381
2958	=item D	3382	=item D
2959		3383
2960	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	3384	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
2961	be found at L<http://proj.llucax.com.ar/wiki/evd>.	3385	be found at L<http://proj.llucax.com.ar/wiki/evd>.
…		…
3072		3496
3073	#define EV_STANDALONE 1	3497	#define EV_STANDALONE 1
3074	#include "ev.h"	3498	#include "ev.h"
3075		3499
3076	Both header files and implementation files can be compiled with a C++	3500	Both header files and implementation files can be compiled with a C++
3077	compiler (at least, thats a stated goal, and breakage will be treated	3501	compiler (at least, that's a stated goal, and breakage will be treated
3078	as a bug).	3502	as a bug).
3079		3503
3080	You need the following files in your source tree, or in a directory	3504	You need the following files in your source tree, or in a directory
3081	in your include path (e.g. in libev/ when using -Ilibev):	3505	in your include path (e.g. in libev/ when using -Ilibev):
3082		3506
…		…
3138	keeps libev from including F<config.h>, and it also defines dummy	3562	keeps libev from including F<config.h>, and it also defines dummy
3139	implementations for some libevent functions (such as logging, which is not	3563	implementations for some libevent functions (such as logging, which is not
3140	supported). It will also not define any of the structs usually found in	3564	supported). It will also not define any of the structs usually found in
3141	F<event.h> that are not directly supported by the libev core alone.	3565	F<event.h> that are not directly supported by the libev core alone.
3142		3566
		3567	In stanbdalone mode, libev will still try to automatically deduce the
		3568	configuration, but has to be more conservative.
		3569
3143	=item EV_USE_MONOTONIC	3570	=item EV_USE_MONOTONIC
3144		3571
3145	If defined to be C<1>, libev will try to detect the availability of the	3572	If defined to be C<1>, libev will try to detect the availability of the
3146	monotonic clock option at both compile time and runtime. Otherwise no use	3573	monotonic clock option at both compile time and runtime. Otherwise no
3147	of the monotonic clock option will be attempted. If you enable this, you	3574	use of the monotonic clock option will be attempted. If you enable this,
3148	usually have to link against librt or something similar. Enabling it when	3575	you usually have to link against librt or something similar. Enabling it
3149	the functionality isn't available is safe, though, although you have	3576	when the functionality isn't available is safe, though, although you have
3150	to make sure you link against any libraries where the C<clock_gettime>	3577	to make sure you link against any libraries where the C<clock_gettime>
3151	function is hiding in (often F<-lrt>).	3578	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
3152		3579
3153	=item EV_USE_REALTIME	3580	=item EV_USE_REALTIME
3154		3581
3155	If defined to be C<1>, libev will try to detect the availability of the	3582	If defined to be C<1>, libev will try to detect the availability of the
3156	real-time clock option at compile time (and assume its availability at	3583	real-time clock option at compile time (and assume its availability
3157	runtime if successful). Otherwise no use of the real-time clock option will	3584	at runtime if successful). Otherwise no use of the real-time clock
3158	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	3585	option will be attempted. This effectively replaces C<gettimeofday>
3159	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the	3586	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
3160	note about libraries in the description of C<EV_USE_MONOTONIC>, though.	3587	correctness. See the note about libraries in the description of
		3588	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3589	C<EV_USE_CLOCK_SYSCALL>.
		3590
		3591	=item EV_USE_CLOCK_SYSCALL
		3592
		3593	If defined to be C<1>, libev will try to use a direct syscall instead
		3594	of calling the system-provided C<clock_gettime> function. This option
		3595	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3596	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3597	programs needlessly. Using a direct syscall is slightly slower (in
		3598	theory), because no optimised vdso implementation can be used, but avoids
		3599	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3600	higher, as it simplifies linking (no need for C<-lrt>).
3161		3601
3162	=item EV_USE_NANOSLEEP	3602	=item EV_USE_NANOSLEEP
3163		3603
3164	If defined to be C<1>, libev will assume that C<nanosleep ()> is available	3604	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
3165	and will use it for delays. Otherwise it will use C<select ()>.	3605	and will use it for delays. Otherwise it will use C<select ()>.
…		…
3181		3621
3182	=item EV_SELECT_USE_FD_SET	3622	=item EV_SELECT_USE_FD_SET
3183		3623
3184	If defined to C<1>, then the select backend will use the system C<fd_set>	3624	If defined to C<1>, then the select backend will use the system C<fd_set>
3185	structure. This is useful if libev doesn't compile due to a missing	3625	structure. This is useful if libev doesn't compile due to a missing
3186	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on	3626	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
3187	exotic systems. This usually limits the range of file descriptors to some	3627	on exotic systems. This usually limits the range of file descriptors to
3188	low limit such as 1024 or might have other limitations (winsocket only	3628	some low limit such as 1024 or might have other limitations (winsocket
3189	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might	3629	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
3190	influence the size of the C<fd_set> used.	3630	configures the maximum size of the C<fd_set>.
3191		3631
3192	=item EV_SELECT_IS_WINSOCKET	3632	=item EV_SELECT_IS_WINSOCKET
3193		3633
3194	When defined to C<1>, the select backend will assume that	3634	When defined to C<1>, the select backend will assume that
3195	select/socket/connect etc. don't understand file descriptors but	3635	select/socket/connect etc. don't understand file descriptors but
…		…
3345	defined to be C<0>, then they are not.	3785	defined to be C<0>, then they are not.
3346		3786
3347	=item EV_MINIMAL	3787	=item EV_MINIMAL
3348		3788
3349	If you need to shave off some kilobytes of code at the expense of some	3789	If you need to shave off some kilobytes of code at the expense of some
3350	speed, define this symbol to C<1>. Currently this is used to override some	3790	speed (but with the full API), define this symbol to C<1>. Currently this
3351	inlining decisions, saves roughly 30% code size on amd64. It also selects a	3791	is used to override some inlining decisions, saves roughly 30% code size
3352	much smaller 2-heap for timer management over the default 4-heap.	3792	on amd64. It also selects a much smaller 2-heap for timer management over
		3793	the default 4-heap.
		3794
		3795	You can save even more by disabling watcher types you do not need
		3796	and setting C<EV_MAXPRI> == C<EV_MINPRI>. Also, disabling C<assert>
		3797	(C<-DNDEBUG>) will usually reduce code size a lot.
		3798
		3799	Defining C<EV_MINIMAL> to C<2> will additionally reduce the core API to
		3800	provide a bare-bones event library. See C<ev.h> for details on what parts
		3801	of the API are still available, and do not complain if this subset changes
		3802	over time.
3353		3803
3354	=item EV_PID_HASHSIZE	3804	=item EV_PID_HASHSIZE
3355		3805
3356	C<ev_child> watchers use a small hash table to distribute workload by	3806	C<ev_child> watchers use a small hash table to distribute workload by
3357	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more	3807	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
…		…
3543	default loop and triggering an C<ev_async> watcher from the default loop	3993	default loop and triggering an C<ev_async> watcher from the default loop
3544	watcher callback into the event loop interested in the signal.	3994	watcher callback into the event loop interested in the signal.
3545		3995
3546	=back	3996	=back
3547		3997
		3998	=head4 THREAD LOCKING EXAMPLE
		3999
		4000	Here is a fictitious example of how to run an event loop in a different
		4001	thread than where callbacks are being invoked and watchers are
		4002	created/added/removed.
		4003
		4004	For a real-world example, see the C<EV::Loop::Async> perl module,
		4005	which uses exactly this technique (which is suited for many high-level
		4006	languages).
		4007
		4008	The example uses a pthread mutex to protect the loop data, a condition
		4009	variable to wait for callback invocations, an async watcher to notify the
		4010	event loop thread and an unspecified mechanism to wake up the main thread.
		4011
		4012	First, you need to associate some data with the event loop:
		4013
		4014	typedef struct {
		4015	mutex_t lock; /* global loop lock */
		4016	ev_async async_w;
		4017	thread_t tid;
		4018	cond_t invoke_cv;
		4019	} userdata;
		4020
		4021	void prepare_loop (EV_P)
		4022	{
		4023	// for simplicity, we use a static userdata struct.
		4024	static userdata u;
		4025
		4026	ev_async_init (&u->async_w, async_cb);
		4027	ev_async_start (EV_A_ &u->async_w);
		4028
		4029	pthread_mutex_init (&u->lock, 0);
		4030	pthread_cond_init (&u->invoke_cv, 0);
		4031
		4032	// now associate this with the loop
		4033	ev_set_userdata (EV_A_ u);
		4034	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4035	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4036
		4037	// then create the thread running ev_loop
		4038	pthread_create (&u->tid, 0, l_run, EV_A);
		4039	}
		4040
		4041	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4042	solely to wake up the event loop so it takes notice of any new watchers
		4043	that might have been added:
		4044
		4045	static void
		4046	async_cb (EV_P_ ev_async *w, int revents)
		4047	{
		4048	// just used for the side effects
		4049	}
		4050
		4051	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4052	protecting the loop data, respectively.
		4053
		4054	static void
		4055	l_release (EV_P)
		4056	{
		4057	userdata *u = ev_userdata (EV_A);
		4058	pthread_mutex_unlock (&u->lock);
		4059	}
		4060
		4061	static void
		4062	l_acquire (EV_P)
		4063	{
		4064	userdata *u = ev_userdata (EV_A);
		4065	pthread_mutex_lock (&u->lock);
		4066	}
		4067
		4068	The event loop thread first acquires the mutex, and then jumps straight
		4069	into C<ev_loop>:
		4070
		4071	void *
		4072	l_run (void *thr_arg)
		4073	{
		4074	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4075
		4076	l_acquire (EV_A);
		4077	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4078	ev_loop (EV_A_ 0);
		4079	l_release (EV_A);
		4080
		4081	return 0;
		4082	}
		4083
		4084	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4085	signal the main thread via some unspecified mechanism (signals? pipe
		4086	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4087	have been called (in a while loop because a) spurious wakeups are possible
		4088	and b) skipping inter-thread-communication when there are no pending
		4089	watchers is very beneficial):
		4090
		4091	static void
		4092	l_invoke (EV_P)
		4093	{
		4094	userdata *u = ev_userdata (EV_A);
		4095
		4096	while (ev_pending_count (EV_A))
		4097	{
		4098	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4099	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4100	}
		4101	}
		4102
		4103	Now, whenever the main thread gets told to invoke pending watchers, it
		4104	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4105	thread to continue:
		4106
		4107	static void
		4108	real_invoke_pending (EV_P)
		4109	{
		4110	userdata *u = ev_userdata (EV_A);
		4111
		4112	pthread_mutex_lock (&u->lock);
		4113	ev_invoke_pending (EV_A);
		4114	pthread_cond_signal (&u->invoke_cv);
		4115	pthread_mutex_unlock (&u->lock);
		4116	}
		4117
		4118	Whenever you want to start/stop a watcher or do other modifications to an
		4119	event loop, you will now have to lock:
		4120
		4121	ev_timer timeout_watcher;
		4122	userdata *u = ev_userdata (EV_A);
		4123
		4124	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4125
		4126	pthread_mutex_lock (&u->lock);
		4127	ev_timer_start (EV_A_ &timeout_watcher);
		4128	ev_async_send (EV_A_ &u->async_w);
		4129	pthread_mutex_unlock (&u->lock);
		4130
		4131	Note that sending the C<ev_async> watcher is required because otherwise
		4132	an event loop currently blocking in the kernel will have no knowledge
		4133	about the newly added timer. By waking up the loop it will pick up any new
		4134	watchers in the next event loop iteration.
		4135
3548	=head3 COROUTINES	4136	=head3 COROUTINES
3549		4137
3550	Libev is very accommodating to coroutines ("cooperative threads"):	4138	Libev is very accommodating to coroutines ("cooperative threads"):
3551	libev fully supports nesting calls to its functions from different	4139	libev fully supports nesting calls to its functions from different
3552	coroutines (e.g. you can call C<ev_loop> on the same loop from two	4140	coroutines (e.g. you can call C<ev_loop> on the same loop from two
3553	different coroutines, and switch freely between both coroutines running the	4141	different coroutines, and switch freely between both coroutines running
3554	loop, as long as you don't confuse yourself). The only exception is that	4142	the loop, as long as you don't confuse yourself). The only exception is
3555	you must not do this from C<ev_periodic> reschedule callbacks.	4143	that you must not do this from C<ev_periodic> reschedule callbacks.
3556		4144
3557	Care has been taken to ensure that libev does not keep local state inside	4145	Care has been taken to ensure that libev does not keep local state inside
3558	C<ev_loop>, and other calls do not usually allow for coroutine switches as	4146	C<ev_loop>, and other calls do not usually allow for coroutine switches as
3559	they do not clal any callbacks.	4147	they do not call any callbacks.
3560		4148
3561	=head2 COMPILER WARNINGS	4149	=head2 COMPILER WARNINGS
3562		4150
3563	Depending on your compiler and compiler settings, you might get no or a	4151	Depending on your compiler and compiler settings, you might get no or a
3564	lot of warnings when compiling libev code. Some people are apparently	4152	lot of warnings when compiling libev code. Some people are apparently
…		…
3598	==2274== definitely lost: 0 bytes in 0 blocks.	4186	==2274== definitely lost: 0 bytes in 0 blocks.
3599	==2274== possibly lost: 0 bytes in 0 blocks.	4187	==2274== possibly lost: 0 bytes in 0 blocks.
3600	==2274== still reachable: 256 bytes in 1 blocks.	4188	==2274== still reachable: 256 bytes in 1 blocks.
3601		4189
3602	Then there is no memory leak, just as memory accounted to global variables	4190	Then there is no memory leak, just as memory accounted to global variables
3603	is not a memleak - the memory is still being refernced, and didn't leak.	4191	is not a memleak - the memory is still being referenced, and didn't leak.
3604		4192
3605	Similarly, under some circumstances, valgrind might report kernel bugs	4193	Similarly, under some circumstances, valgrind might report kernel bugs
3606	as if it were a bug in libev (e.g. in realloc or in the poll backend,	4194	as if it were a bug in libev (e.g. in realloc or in the poll backend,
3607	although an acceptable workaround has been found here), or it might be	4195	although an acceptable workaround has been found here), or it might be
3608	confused.	4196	confused.
…		…
3637	way (note also that glib is the slowest event library known to man).	4225	way (note also that glib is the slowest event library known to man).
3638		4226
3639	There is no supported compilation method available on windows except	4227	There is no supported compilation method available on windows except
3640	embedding it into other applications.	4228	embedding it into other applications.
3641		4229
		4230	Sensible signal handling is officially unsupported by Microsoft - libev
		4231	tries its best, but under most conditions, signals will simply not work.
		4232
3642	Not a libev limitation but worth mentioning: windows apparently doesn't	4233	Not a libev limitation but worth mentioning: windows apparently doesn't
3643	accept large writes: instead of resulting in a partial write, windows will	4234	accept large writes: instead of resulting in a partial write, windows will
3644	either accept everything or return C<ENOBUFS> if the buffer is too large,	4235	either accept everything or return C<ENOBUFS> if the buffer is too large,
3645	so make sure you only write small amounts into your sockets (less than a	4236	so make sure you only write small amounts into your sockets (less than a
3646	megabyte seems safe, but this apparently depends on the amount of memory	4237	megabyte seems safe, but this apparently depends on the amount of memory
…		…
3650	the abysmal performance of winsockets, using a large number of sockets	4241	the abysmal performance of winsockets, using a large number of sockets
3651	is not recommended (and not reasonable). If your program needs to use	4242	is not recommended (and not reasonable). If your program needs to use
3652	more than a hundred or so sockets, then likely it needs to use a totally	4243	more than a hundred or so sockets, then likely it needs to use a totally
3653	different implementation for windows, as libev offers the POSIX readiness	4244	different implementation for windows, as libev offers the POSIX readiness
3654	notification model, which cannot be implemented efficiently on windows	4245	notification model, which cannot be implemented efficiently on windows
3655	(Microsoft monopoly games).	4246	(due to Microsoft monopoly games).
3656		4247
3657	A typical way to use libev under windows is to embed it (see the embedding	4248	A typical way to use libev under windows is to embed it (see the embedding
3658	section for details) and use the following F<evwrap.h> header file instead	4249	section for details) and use the following F<evwrap.h> header file instead
3659	of F<ev.h>:	4250	of F<ev.h>:
3660		4251
…		…
3696		4287
3697	Early versions of winsocket's select only supported waiting for a maximum	4288	Early versions of winsocket's select only supported waiting for a maximum
3698	of C<64> handles (probably owning to the fact that all windows kernels	4289	of C<64> handles (probably owning to the fact that all windows kernels
3699	can only wait for C<64> things at the same time internally; Microsoft	4290	can only wait for C<64> things at the same time internally; Microsoft
3700	recommends spawning a chain of threads and wait for 63 handles and the	4291	recommends spawning a chain of threads and wait for 63 handles and the
3701	previous thread in each. Great).	4292	previous thread in each. Sounds great!).
3702		4293
3703	Newer versions support more handles, but you need to define C<FD_SETSIZE>	4294	Newer versions support more handles, but you need to define C<FD_SETSIZE>
3704	to some high number (e.g. C<2048>) before compiling the winsocket select	4295	to some high number (e.g. C<2048>) before compiling the winsocket select
3705	call (which might be in libev or elsewhere, for example, perl does its own	4296	call (which might be in libev or elsewhere, for example, perl and many
3706	select emulation on windows).	4297	other interpreters do their own select emulation on windows).
3707		4298
3708	Another limit is the number of file descriptors in the Microsoft runtime	4299	Another limit is the number of file descriptors in the Microsoft runtime
3709	libraries, which by default is C<64> (there must be a hidden I<64> fetish	4300	libraries, which by default is C<64> (there must be a hidden I<64>
3710	or something like this inside Microsoft). You can increase this by calling	4301	fetish or something like this inside Microsoft). You can increase this
3711	C<_setmaxstdio>, which can increase this limit to C<2048> (another	4302	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
3712	arbitrary limit), but is broken in many versions of the Microsoft runtime	4303	(another arbitrary limit), but is broken in many versions of the Microsoft
3713	libraries.
3714
3715	This might get you to about C<512> or C<2048> sockets (depending on	4304	runtime libraries. This might get you to about C<512> or C<2048> sockets
3716	windows version and/or the phase of the moon). To get more, you need to	4305	(depending on windows version and/or the phase of the moon). To get more,
3717	wrap all I/O functions and provide your own fd management, but the cost of	4306	you need to wrap all I/O functions and provide your own fd management, but
3718	calling select (O(n²)) will likely make this unworkable.	4307	the cost of calling select (O(n²)) will likely make this unworkable.
3719		4308
3720	=back	4309	=back
3721		4310
3722	=head2 PORTABILITY REQUIREMENTS	4311	=head2 PORTABILITY REQUIREMENTS
3723		4312
…		…
3766	=item C<double> must hold a time value in seconds with enough accuracy	4355	=item C<double> must hold a time value in seconds with enough accuracy
3767		4356
3768	The type C<double> is used to represent timestamps. It is required to	4357	The type C<double> is used to represent timestamps. It is required to
3769	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	4358	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
3770	enough for at least into the year 4000. This requirement is fulfilled by	4359	enough for at least into the year 4000. This requirement is fulfilled by
3771	implementations implementing IEEE 754 (basically all existing ones).	4360	implementations implementing IEEE 754, which is basically all existing
		4361	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4362	2200.
3772		4363
3773	=back	4364	=back
3774		4365
3775	If you know of other additional requirements drop me a note.	4366	If you know of other additional requirements drop me a note.
3776		4367
…		…
3844	involves iterating over all running async watchers or all signal numbers.	4435	involves iterating over all running async watchers or all signal numbers.
3845		4436
3846	=back	4437	=back
3847		4438
3848		4439
		4440	=head1 GLOSSARY
		4441
		4442	=over 4
		4443
		4444	=item active
		4445
		4446	A watcher is active as long as it has been started (has been attached to
		4447	an event loop) but not yet stopped (disassociated from the event loop).
		4448
		4449	=item application
		4450
		4451	In this document, an application is whatever is using libev.
		4452
		4453	=item callback
		4454
		4455	The address of a function that is called when some event has been
		4456	detected. Callbacks are being passed the event loop, the watcher that
		4457	received the event, and the actual event bitset.
		4458
		4459	=item callback invocation
		4460
		4461	The act of calling the callback associated with a watcher.
		4462
		4463	=item event
		4464
		4465	A change of state of some external event, such as data now being available
		4466	for reading on a file descriptor, time having passed or simply not having
		4467	any other events happening anymore.
		4468
		4469	In libev, events are represented as single bits (such as C<EV_READ> or
		4470	C<EV_TIMEOUT>).
		4471
		4472	=item event library
		4473
		4474	A software package implementing an event model and loop.
		4475
		4476	=item event loop
		4477
		4478	An entity that handles and processes external events and converts them
		4479	into callback invocations.
		4480
		4481	=item event model
		4482
		4483	The model used to describe how an event loop handles and processes
		4484	watchers and events.
		4485
		4486	=item pending
		4487
		4488	A watcher is pending as soon as the corresponding event has been detected,
		4489	and stops being pending as soon as the watcher will be invoked or its
		4490	pending status is explicitly cleared by the application.
		4491
		4492	A watcher can be pending, but not active. Stopping a watcher also clears
		4493	its pending status.
		4494
		4495	=item real time
		4496
		4497	The physical time that is observed. It is apparently strictly monotonic :)
		4498
		4499	=item wall-clock time
		4500
		4501	The time and date as shown on clocks. Unlike real time, it can actually
		4502	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4503	clock.
		4504
		4505	=item watcher
		4506
		4507	A data structure that describes interest in certain events. Watchers need
		4508	to be started (attached to an event loop) before they can receive events.
		4509
		4510	=item watcher invocation
		4511
		4512	The act of calling the callback associated with a watcher.
		4513
		4514	=back
		4515
3849	=head1 AUTHOR	4516	=head1 AUTHOR
3850		4517
3851	Marc Lehmann <libev@schmorp.de>.	4518	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
3852		4519

Diff Legend

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