ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.345 by root, Wed Nov 10 14:36:42 2010 UTC vs.
Revision 1.399 by root, Mon Apr 2 23:14:41 2012 UTC

…		…
58	ev_timer_start (loop, &timeout_watcher);	58	ev_timer_start (loop, &timeout_watcher);
59		59
60	// now wait for events to arrive	60	// now wait for events to arrive
61	ev_run (loop, 0);	61	ev_run (loop, 0);
62		62
63	// unloop was called, so exit	63	// break was called, so exit
64	return 0;	64	return 0;
65	}	65	}
66		66
67	=head1 ABOUT THIS DOCUMENT	67	=head1 ABOUT THIS DOCUMENT
68		68
…		…
174	=item ev_tstamp ev_time ()	174	=item ev_tstamp ev_time ()
175		175
176	Returns the current time as libev would use it. Please note that the	176	Returns the current time as libev would use it. Please note that the
177	C<ev_now> function is usually faster and also often returns the timestamp	177	C<ev_now> function is usually faster and also often returns the timestamp
178	you actually want to know. Also interesting is the combination of	178	you actually want to know. Also interesting is the combination of
179	C<ev_update_now> and C<ev_now>.	179	C<ev_now_update> and C<ev_now>.
180		180
181	=item ev_sleep (ev_tstamp interval)	181	=item ev_sleep (ev_tstamp interval)
182		182
183	Sleep for the given interval: The current thread will be blocked until	183	Sleep for the given interval: The current thread will be blocked
184	either it is interrupted or the given time interval has passed. Basically	184	until either it is interrupted or the given time interval has
		185	passed (approximately - it might return a bit earlier even if not
		186	interrupted). Returns immediately if C<< interval <= 0 >>.
		187
185	this is a sub-second-resolution C<sleep ()>.	188	Basically this is a sub-second-resolution C<sleep ()>.
		189
		190	The range of the C<interval> is limited - libev only guarantees to work
		191	with sleep times of up to one day (C<< interval <= 86400 >>).
186		192
187	=item int ev_version_major ()	193	=item int ev_version_major ()
188		194
189	=item int ev_version_minor ()	195	=item int ev_version_minor ()
190		196
…		…
299	}	305	}
300		306
301	...	307	...
302	ev_set_syserr_cb (fatal_error);	308	ev_set_syserr_cb (fatal_error);
303		309
		310	=item ev_feed_signal (int signum)
		311
		312	This function can be used to "simulate" a signal receive. It is completely
		313	safe to call this function at any time, from any context, including signal
		314	handlers or random threads.
		315
		316	Its main use is to customise signal handling in your process, especially
		317	in the presence of threads. For example, you could block signals
		318	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		319	creating any loops), and in one thread, use C<sigwait> or any other
		320	mechanism to wait for signals, then "deliver" them to libev by calling
		321	C<ev_feed_signal>.
		322
304	=back	323	=back
305		324
306	=head1 FUNCTIONS CONTROLLING EVENT LOOPS	325	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
307		326
308	An event loop is described by a C<struct ev_loop *> (the C<struct> is	327	An event loop is described by a C<struct ev_loop *> (the C<struct> is
…		…
419		438
420	Signalfd will not be used by default as this changes your signal mask, and	439	Signalfd will not be used by default as this changes your signal mask, and
421	there are a lot of shoddy libraries and programs (glib's threadpool for	440	there are a lot of shoddy libraries and programs (glib's threadpool for
422	example) that can't properly initialise their signal masks.	441	example) that can't properly initialise their signal masks.
423		442
		443	=item C<EVFLAG_NOSIGMASK>
		444
		445	When this flag is specified, then libev will avoid to modify the signal
		446	mask. Specifically, this means you have to make sure signals are unblocked
		447	when you want to receive them.
		448
		449	This behaviour is useful when you want to do your own signal handling, or
		450	want to handle signals only in specific threads and want to avoid libev
		451	unblocking the signals.
		452
		453	It's also required by POSIX in a threaded program, as libev calls
		454	C<sigprocmask>, whose behaviour is officially unspecified.
		455
		456	This flag's behaviour will become the default in future versions of libev.
		457
424	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	458	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
425		459
426	This is your standard select(2) backend. Not I<completely> standard, as	460	This is your standard select(2) backend. Not I<completely> standard, as
427	libev tries to roll its own fd_set with no limits on the number of fds,	461	libev tries to roll its own fd_set with no limits on the number of fds,
428	but if that fails, expect a fairly low limit on the number of fds when	462	but if that fails, expect a fairly low limit on the number of fds when
…		…
455	=item C<EVBACKEND_EPOLL> (value 4, Linux)	489	=item C<EVBACKEND_EPOLL> (value 4, Linux)
456		490
457	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9	491	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
458	kernels).	492	kernels).
459		493
460	For few fds, this backend is a bit little slower than poll and select,	494	For few fds, this backend is a bit little slower than poll and select, but
461	but it scales phenomenally better. While poll and select usually scale	495	it scales phenomenally better. While poll and select usually scale like
462	like O(total_fds) where n is the total number of fds (or the highest fd),	496	O(total_fds) where total_fds is the total number of fds (or the highest
463	epoll scales either O(1) or O(active_fds).	497	fd), epoll scales either O(1) or O(active_fds).
464		498
465	The epoll mechanism deserves honorable mention as the most misdesigned	499	The epoll mechanism deserves honorable mention as the most misdesigned
466	of the more advanced event mechanisms: mere annoyances include silently	500	of the more advanced event mechanisms: mere annoyances include silently
467	dropping file descriptors, requiring a system call per change per file	501	dropping file descriptors, requiring a system call per change per file
468	descriptor (and unnecessary guessing of parameters), problems with dup,	502	descriptor (and unnecessary guessing of parameters), problems with dup,
…		…
471	0.1ms) and so on. The biggest issue is fork races, however - if a program	505	0.1ms) and so on. The biggest issue is fork races, however - if a program
472	forks then I<both> parent and child process have to recreate the epoll	506	forks then I<both> parent and child process have to recreate the epoll
473	set, which can take considerable time (one syscall per file descriptor)	507	set, which can take considerable time (one syscall per file descriptor)
474	and is of course hard to detect.	508	and is of course hard to detect.
475		509
476	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but	510	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
477	of course I<doesn't>, and epoll just loves to report events for totally	511	but of course I<doesn't>, and epoll just loves to report events for
478	I<different> file descriptors (even already closed ones, so one cannot	512	totally I<different> file descriptors (even already closed ones, so
479	even remove them from the set) than registered in the set (especially	513	one cannot even remove them from the set) than registered in the set
480	on SMP systems). Libev tries to counter these spurious notifications by	514	(especially on SMP systems). Libev tries to counter these spurious
481	employing an additional generation counter and comparing that against the	515	notifications by employing an additional generation counter and comparing
482	events to filter out spurious ones, recreating the set when required. Last	516	that against the events to filter out spurious ones, recreating the set
		517	when required. Epoll also erroneously rounds down timeouts, but gives you
		518	no way to know when and by how much, so sometimes you have to busy-wait
		519	because epoll returns immediately despite a nonzero timeout. And last
483	not least, it also refuses to work with some file descriptors which work	520	not least, it also refuses to work with some file descriptors which work
484	perfectly fine with C<select> (files, many character devices...).	521	perfectly fine with C<select> (files, many character devices...).
485		522
486	Epoll is truly the train wreck analog among event poll mechanisms.	523	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		524	cobbled together in a hurry, no thought to design or interaction with
		525	others. Oh, the pain, will it ever stop...
487		526
488	While stopping, setting and starting an I/O watcher in the same iteration	527	While stopping, setting and starting an I/O watcher in the same iteration
489	will result in some caching, there is still a system call per such	528	will result in some caching, there is still a system call per such
490	incident (because the same I<file descriptor> could point to a different	529	incident (because the same I<file descriptor> could point to a different
491	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed	530	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
…		…
528		567
529	It scales in the same way as the epoll backend, but the interface to the	568	It scales in the same way as the epoll backend, but the interface to the
530	kernel is more efficient (which says nothing about its actual speed, of	569	kernel is more efficient (which says nothing about its actual speed, of
531	course). While stopping, setting and starting an I/O watcher does never	570	course). While stopping, setting and starting an I/O watcher does never
532	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	571	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
533	two event changes per incident. Support for C<fork ()> is very bad (but	572	two event changes per incident. Support for C<fork ()> is very bad (you
534	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect	573	might have to leak fd's on fork, but it's more sane than epoll) and it
535	cases	574	drops fds silently in similarly hard-to-detect cases
536		575
537	This backend usually performs well under most conditions.	576	This backend usually performs well under most conditions.
538		577
539	While nominally embeddable in other event loops, this doesn't work	578	While nominally embeddable in other event loops, this doesn't work
540	everywhere, so you might need to test for this. And since it is broken	579	everywhere, so you might need to test for this. And since it is broken
…		…
557	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	596	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
558		597
559	This uses the Solaris 10 event port mechanism. As with everything on Solaris,	598	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
560	it's really slow, but it still scales very well (O(active_fds)).	599	it's really slow, but it still scales very well (O(active_fds)).
561		600
562	Please note that Solaris event ports can deliver a lot of spurious
563	notifications, so you need to use non-blocking I/O or other means to avoid
564	blocking when no data (or space) is available.
565
566	While this backend scales well, it requires one system call per active	601	While this backend scales well, it requires one system call per active
567	file descriptor per loop iteration. For small and medium numbers of file	602	file descriptor per loop iteration. For small and medium numbers of file
568	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	603	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
569	might perform better.	604	might perform better.
570		605
571	On the positive side, with the exception of the spurious readiness	606	On the positive side, this backend actually performed fully to
572	notifications, this backend actually performed fully to specification
573	in all tests and is fully embeddable, which is a rare feat among the	607	specification in all tests and is fully embeddable, which is a rare feat
574	OS-specific backends (I vastly prefer correctness over speed hacks).	608	among the OS-specific backends (I vastly prefer correctness over speed
		609	hacks).
		610
		611	On the negative side, the interface is I<bizarre> - so bizarre that
		612	even sun itself gets it wrong in their code examples: The event polling
		613	function sometimes returns events to the caller even though an error
		614	occurred, but with no indication whether it has done so or not (yes, it's
		615	even documented that way) - deadly for edge-triggered interfaces where you
		616	absolutely have to know whether an event occurred or not because you have
		617	to re-arm the watcher.
		618
		619	Fortunately libev seems to be able to work around these idiocies.
575		620
576	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	621	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
577	C<EVBACKEND_POLL>.	622	C<EVBACKEND_POLL>.
578		623
579	=item C<EVBACKEND_ALL>	624	=item C<EVBACKEND_ALL>
580		625
581	Try all backends (even potentially broken ones that wouldn't be tried	626	Try all backends (even potentially broken ones that wouldn't be tried
582	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	627	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
583	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	628	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
584		629
585	It is definitely not recommended to use this flag.	630	It is definitely not recommended to use this flag, use whatever
		631	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		632	at all.
		633
		634	=item C<EVBACKEND_MASK>
		635
		636	Not a backend at all, but a mask to select all backend bits from a
		637	C<flags> value, in case you want to mask out any backends from a flags
		638	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
586		639
587	=back	640	=back
588		641
589	If one or more of the backend flags are or'ed into the flags value,	642	If one or more of the backend flags are or'ed into the flags value,
590	then only these backends will be tried (in the reverse order as listed	643	then only these backends will be tried (in the reverse order as listed
…		…
739	without a previous call to C<ev_suspend>.	792	without a previous call to C<ev_suspend>.
740		793
741	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the	794	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
742	event loop time (see C<ev_now_update>).	795	event loop time (see C<ev_now_update>).
743		796
744	=item ev_run (loop, int flags)	797	=item bool ev_run (loop, int flags)
745		798
746	Finally, this is it, the event handler. This function usually is called	799	Finally, this is it, the event handler. This function usually is called
747	after you have initialised all your watchers and you want to start	800	after you have initialised all your watchers and you want to start
748	handling events. It will ask the operating system for any new events, call	801	handling events. It will ask the operating system for any new events, call
749	the watcher callbacks, an then repeat the whole process indefinitely: This	802	the watcher callbacks, and then repeat the whole process indefinitely: This
750	is why event loops are called I<loops>.	803	is why event loops are called I<loops>.
751		804
752	If the flags argument is specified as C<0>, it will keep handling events	805	If the flags argument is specified as C<0>, it will keep handling events
753	until either no event watchers are active anymore or C<ev_break> was	806	until either no event watchers are active anymore or C<ev_break> was
754	called.	807	called.
		808
		809	The return value is false if there are no more active watchers (which
		810	usually means "all jobs done" or "deadlock"), and true in all other cases
		811	(which usually means " you should call C<ev_run> again").
755		812
756	Please note that an explicit C<ev_break> is usually better than	813	Please note that an explicit C<ev_break> is usually better than
757	relying on all watchers to be stopped when deciding when a program has	814	relying on all watchers to be stopped when deciding when a program has
758	finished (especially in interactive programs), but having a program	815	finished (especially in interactive programs), but having a program
759	that automatically loops as long as it has to and no longer by virtue	816	that automatically loops as long as it has to and no longer by virtue
760	of relying on its watchers stopping correctly, that is truly a thing of	817	of relying on its watchers stopping correctly, that is truly a thing of
761	beauty.	818	beauty.
762		819
763	This function is also I<mostly> exception-safe - you can break out of	820	This function is I<mostly> exception-safe - you can break out of a
764	a C<ev_run> call by calling C<longjmp> in a callback, throwing a C++	821	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
765	exception and so on. This does not decrement the C<ev_depth> value, nor	822	exception and so on. This does not decrement the C<ev_depth> value, nor
766	will it clear any outstanding C<EVBREAK_ONE> breaks.	823	will it clear any outstanding C<EVBREAK_ONE> breaks.
767		824
768	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle	825	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
769	those events and any already outstanding ones, but will not wait and	826	those events and any already outstanding ones, but will not wait and
…		…
781	This is useful if you are waiting for some external event in conjunction	838	This is useful if you are waiting for some external event in conjunction
782	with something not expressible using other libev watchers (i.e. "roll your	839	with something not expressible using other libev watchers (i.e. "roll your
783	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is	840	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
784	usually a better approach for this kind of thing.	841	usually a better approach for this kind of thing.
785		842
786	Here are the gory details of what C<ev_run> does:	843	Here are the gory details of what C<ev_run> does (this is for your
		844	understanding, not a guarantee that things will work exactly like this in
		845	future versions):
787		846
788	- Increment loop depth.	847	- Increment loop depth.
789	- Reset the ev_break status.	848	- Reset the ev_break status.
790	- Before the first iteration, call any pending watchers.	849	- Before the first iteration, call any pending watchers.
791	LOOP:	850	LOOP:
…		…
824	anymore.	883	anymore.
825		884
826	... queue jobs here, make sure they register event watchers as long	885	... queue jobs here, make sure they register event watchers as long
827	... as they still have work to do (even an idle watcher will do..)	886	... as they still have work to do (even an idle watcher will do..)
828	ev_run (my_loop, 0);	887	ev_run (my_loop, 0);
829	... jobs done or somebody called unloop. yeah!	888	... jobs done or somebody called break. yeah!
830		889
831	=item ev_break (loop, how)	890	=item ev_break (loop, how)
832		891
833	Can be used to make a call to C<ev_run> return early (but only after it	892	Can be used to make a call to C<ev_run> return early (but only after it
834	has processed all outstanding events). The C<how> argument must be either	893	has processed all outstanding events). The C<how> argument must be either
…		…
867	running when nothing else is active.	926	running when nothing else is active.
868		927
869	ev_signal exitsig;	928	ev_signal exitsig;
870	ev_signal_init (&exitsig, sig_cb, SIGINT);	929	ev_signal_init (&exitsig, sig_cb, SIGINT);
871	ev_signal_start (loop, &exitsig);	930	ev_signal_start (loop, &exitsig);
872	evf_unref (loop);	931	ev_unref (loop);
873		932
874	Example: For some weird reason, unregister the above signal handler again.	933	Example: For some weird reason, unregister the above signal handler again.
875		934
876	ev_ref (loop);	935	ev_ref (loop);
877	ev_signal_stop (loop, &exitsig);	936	ev_signal_stop (loop, &exitsig);
…		…
897	overhead for the actual polling but can deliver many events at once.	956	overhead for the actual polling but can deliver many events at once.
898		957
899	By setting a higher I<io collect interval> you allow libev to spend more	958	By setting a higher I<io collect interval> you allow libev to spend more
900	time collecting I/O events, so you can handle more events per iteration,	959	time collecting I/O events, so you can handle more events per iteration,
901	at the cost of increasing latency. Timeouts (both C<ev_periodic> and	960	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
902	C<ev_timer>) will be not affected. Setting this to a non-null value will	961	C<ev_timer>) will not be affected. Setting this to a non-null value will
903	introduce an additional C<ev_sleep ()> call into most loop iterations. The	962	introduce an additional C<ev_sleep ()> call into most loop iterations. The
904	sleep time ensures that libev will not poll for I/O events more often then	963	sleep time ensures that libev will not poll for I/O events more often then
905	once per this interval, on average.	964	once per this interval, on average (as long as the host time resolution is
		965	good enough).
906		966
907	Likewise, by setting a higher I<timeout collect interval> you allow libev	967	Likewise, by setting a higher I<timeout collect interval> you allow libev
908	to spend more time collecting timeouts, at the expense of increased	968	to spend more time collecting timeouts, at the expense of increased
909	latency/jitter/inexactness (the watcher callback will be called	969	latency/jitter/inexactness (the watcher callback will be called
910	later). C<ev_io> watchers will not be affected. Setting this to a non-null	970	later). C<ev_io> watchers will not be affected. Setting this to a non-null
…		…
964	can be done relatively simply by putting mutex_lock/unlock calls around	1024	can be done relatively simply by putting mutex_lock/unlock calls around
965	each call to a libev function.	1025	each call to a libev function.
966		1026
967	However, C<ev_run> can run an indefinite time, so it is not feasible	1027	However, C<ev_run> can run an indefinite time, so it is not feasible
968	to wait for it to return. One way around this is to wake up the event	1028	to wait for it to return. One way around this is to wake up the event
969	loop via C<ev_break> and C<av_async_send>, another way is to set these	1029	loop via C<ev_break> and C<ev_async_send>, another way is to set these
970	I<release> and I<acquire> callbacks on the loop.	1030	I<release> and I<acquire> callbacks on the loop.
971		1031
972	When set, then C<release> will be called just before the thread is	1032	When set, then C<release> will be called just before the thread is
973	suspended waiting for new events, and C<acquire> is called just	1033	suspended waiting for new events, and C<acquire> is called just
974	afterwards.	1034	afterwards.
…		…
1316	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related	1376	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1317	functions that do not need a watcher.	1377	functions that do not need a watcher.
1318		1378
1319	=back	1379	=back
1320		1380
1321	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1381	See also the L<ASSOCIATING CUSTOM DATA WITH A WATCHER> and L<BUILDING YOUR
1322		1382	OWN COMPOSITE WATCHERS> idioms.
1323	Each watcher has, by default, a member C<void *data> that you can change
1324	and read at any time: libev will completely ignore it. This can be used
1325	to associate arbitrary data with your watcher. If you need more data and
1326	don't want to allocate memory and store a pointer to it in that data
1327	member, you can also "subclass" the watcher type and provide your own
1328	data:
1329
1330	struct my_io
1331	{
1332	ev_io io;
1333	int otherfd;
1334	void *somedata;
1335	struct whatever *mostinteresting;
1336	};
1337
1338	...
1339	struct my_io w;
1340	ev_io_init (&w.io, my_cb, fd, EV_READ);
1341
1342	And since your callback will be called with a pointer to the watcher, you
1343	can cast it back to your own type:
1344
1345	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
1346	{
1347	struct my_io *w = (struct my_io *)w_;
1348	...
1349	}
1350
1351	More interesting and less C-conformant ways of casting your callback type
1352	instead have been omitted.
1353
1354	Another common scenario is to use some data structure with multiple
1355	embedded watchers:
1356
1357	struct my_biggy
1358	{
1359	int some_data;
1360	ev_timer t1;
1361	ev_timer t2;
1362	}
1363
1364	In this case getting the pointer to C<my_biggy> is a bit more
1365	complicated: Either you store the address of your C<my_biggy> struct
1366	in the C<data> member of the watcher (for woozies), or you need to use
1367	some pointer arithmetic using C<offsetof> inside your watchers (for real
1368	programmers):
1369
1370	#include <stddef.h>
1371
1372	static void
1373	t1_cb (EV_P_ ev_timer *w, int revents)
1374	{
1375	struct my_biggy big = (struct my_biggy *)
1376	(((char *)w) - offsetof (struct my_biggy, t1));
1377	}
1378
1379	static void
1380	t2_cb (EV_P_ ev_timer *w, int revents)
1381	{
1382	struct my_biggy big = (struct my_biggy *)
1383	(((char *)w) - offsetof (struct my_biggy, t2));
1384	}
1385		1383
1386	=head2 WATCHER STATES	1384	=head2 WATCHER STATES
1387		1385
1388	There are various watcher states mentioned throughout this manual -	1386	There are various watcher states mentioned throughout this manual -
1389	active, pending and so on. In this section these states and the rules to	1387	active, pending and so on. In this section these states and the rules to
…		…
1392		1390
1393	=over 4	1391	=over 4
1394		1392
1395	=item initialiased	1393	=item initialiased
1396		1394
1397	Before a watcher can be registered with the event looop it has to be	1395	Before a watcher can be registered with the event loop it has to be
1398	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to	1396	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
1399	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.	1397	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1400		1398
1401	In this state it is simply some block of memory that is suitable for use	1399	In this state it is simply some block of memory that is suitable for
1402	in an event loop. It can be moved around, freed, reused etc. at will.	1400	use in an event loop. It can be moved around, freed, reused etc. at
		1401	will - as long as you either keep the memory contents intact, or call
		1402	C<ev_TYPE_init> again.
1403		1403
1404	=item started/running/active	1404	=item started/running/active
1405		1405
1406	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes	1406	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
1407	property of the event loop, and is actively waiting for events. While in	1407	property of the event loop, and is actively waiting for events. While in
…		…
1435	latter will clear any pending state the watcher might be in, regardless	1435	latter will clear any pending state the watcher might be in, regardless
1436	of whether it was active or not, so stopping a watcher explicitly before	1436	of whether it was active or not, so stopping a watcher explicitly before
1437	freeing it is often a good idea.	1437	freeing it is often a good idea.
1438		1438
1439	While stopped (and not pending) the watcher is essentially in the	1439	While stopped (and not pending) the watcher is essentially in the
1440	initialised state, that is it can be reused, moved, modified in any way	1440	initialised state, that is, it can be reused, moved, modified in any way
1441	you wish.	1441	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1442	it again).
1442		1443
1443	=back	1444	=back
1444		1445
1445	=head2 WATCHER PRIORITY MODELS	1446	=head2 WATCHER PRIORITY MODELS
1446		1447
…		…
1575	In general you can register as many read and/or write event watchers per	1576	In general you can register as many read and/or write event watchers per
1576	fd as you want (as long as you don't confuse yourself). Setting all file	1577	fd as you want (as long as you don't confuse yourself). Setting all file
1577	descriptors to non-blocking mode is also usually a good idea (but not	1578	descriptors to non-blocking mode is also usually a good idea (but not
1578	required if you know what you are doing).	1579	required if you know what you are doing).
1579		1580
1580	If you cannot use non-blocking mode, then force the use of a
1581	known-to-be-good backend (at the time of this writing, this includes only
1582	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
1583	descriptors for which non-blocking operation makes no sense (such as
1584	files) - libev doesn't guarantee any specific behaviour in that case.
1585
1586	Another thing you have to watch out for is that it is quite easy to	1581	Another thing you have to watch out for is that it is quite easy to
1587	receive "spurious" readiness notifications, that is your callback might	1582	receive "spurious" readiness notifications, that is, your callback might
1588	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1583	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1589	because there is no data. Not only are some backends known to create a	1584	because there is no data. It is very easy to get into this situation even
1590	lot of those (for example Solaris ports), it is very easy to get into	1585	with a relatively standard program structure. Thus it is best to always
1591	this situation even with a relatively standard program structure. Thus	1586	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1592	it is best to always use non-blocking I/O: An extra C<read>(2) returning
1593	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1587	preferable to a program hanging until some data arrives.
1594		1588
1595	If you cannot run the fd in non-blocking mode (for example you should	1589	If you cannot run the fd in non-blocking mode (for example you should
1596	not play around with an Xlib connection), then you have to separately	1590	not play around with an Xlib connection), then you have to separately
1597	re-test whether a file descriptor is really ready with a known-to-be good	1591	re-test whether a file descriptor is really ready with a known-to-be good
1598	interface such as poll (fortunately in our Xlib example, Xlib already	1592	interface such as poll (fortunately in the case of Xlib, it already does
1599	does this on its own, so its quite safe to use). Some people additionally	1593	this on its own, so its quite safe to use). Some people additionally
1600	use C<SIGALRM> and an interval timer, just to be sure you won't block	1594	use C<SIGALRM> and an interval timer, just to be sure you won't block
1601	indefinitely.	1595	indefinitely.
1602		1596
1603	But really, best use non-blocking mode.	1597	But really, best use non-blocking mode.
1604		1598
…		…
1632		1626
1633	There is no workaround possible except not registering events	1627	There is no workaround possible except not registering events
1634	for potentially C<dup ()>'ed file descriptors, or to resort to	1628	for potentially C<dup ()>'ed file descriptors, or to resort to
1635	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1629	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1636		1630
		1631	=head3 The special problem of files
		1632
		1633	Many people try to use C<select> (or libev) on file descriptors
		1634	representing files, and expect it to become ready when their program
		1635	doesn't block on disk accesses (which can take a long time on their own).
		1636
		1637	However, this cannot ever work in the "expected" way - you get a readiness
		1638	notification as soon as the kernel knows whether and how much data is
		1639	there, and in the case of open files, that's always the case, so you
		1640	always get a readiness notification instantly, and your read (or possibly
		1641	write) will still block on the disk I/O.
		1642
		1643	Another way to view it is that in the case of sockets, pipes, character
		1644	devices and so on, there is another party (the sender) that delivers data
		1645	on its own, but in the case of files, there is no such thing: the disk
		1646	will not send data on its own, simply because it doesn't know what you
		1647	wish to read - you would first have to request some data.
		1648
		1649	Since files are typically not-so-well supported by advanced notification
		1650	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1651	to files, even though you should not use it. The reason for this is
		1652	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1653	usually a tty, often a pipe, but also sometimes files or special devices
		1654	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1655	F</dev/urandom>), and even though the file might better be served with
		1656	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1657	it "just works" instead of freezing.
		1658
		1659	So avoid file descriptors pointing to files when you know it (e.g. use
		1660	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1661	when you rarely read from a file instead of from a socket, and want to
		1662	reuse the same code path.
		1663
1637	=head3 The special problem of fork	1664	=head3 The special problem of fork
1638		1665
1639	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1666	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1640	useless behaviour. Libev fully supports fork, but needs to be told about	1667	useless behaviour. Libev fully supports fork, but needs to be told about
1641	it in the child.	1668	it in the child if you want to continue to use it in the child.
1642		1669
1643	To support fork in your programs, you either have to call	1670	To support fork in your child processes, you have to call C<ev_loop_fork
1644	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,	1671	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1645	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1672	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1646	C<EVBACKEND_POLL>.
1647		1673
1648	=head3 The special problem of SIGPIPE	1674	=head3 The special problem of SIGPIPE
1649		1675
1650	While not really specific to libev, it is easy to forget about C<SIGPIPE>:	1676	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1651	when writing to a pipe whose other end has been closed, your program gets	1677	when writing to a pipe whose other end has been closed, your program gets
…		…
1749	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1775	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1750	monotonic clock option helps a lot here).	1776	monotonic clock option helps a lot here).
1751		1777
1752	The callback is guaranteed to be invoked only I<after> its timeout has	1778	The callback is guaranteed to be invoked only I<after> its timeout has
1753	passed (not I<at>, so on systems with very low-resolution clocks this	1779	passed (not I<at>, so on systems with very low-resolution clocks this
1754	might introduce a small delay). If multiple timers become ready during the	1780	might introduce a small delay, see "the special problem of being too
		1781	early", below). If multiple timers become ready during the same loop
1755	same loop iteration then the ones with earlier time-out values are invoked	1782	iteration then the ones with earlier time-out values are invoked before
1756	before ones of the same priority with later time-out values (but this is	1783	ones of the same priority with later time-out values (but this is no
1757	no longer true when a callback calls C<ev_run> recursively).	1784	longer true when a callback calls C<ev_run> recursively).
1758		1785
1759	=head3 Be smart about timeouts	1786	=head3 Be smart about timeouts
1760		1787
1761	Many real-world problems involve some kind of timeout, usually for error	1788	Many real-world problems involve some kind of timeout, usually for error
1762	recovery. A typical example is an HTTP request - if the other side hangs,	1789	recovery. A typical example is an HTTP request - if the other side hangs,
…		…
1837		1864
1838	In this case, it would be more efficient to leave the C<ev_timer> alone,	1865	In this case, it would be more efficient to leave the C<ev_timer> alone,
1839	but remember the time of last activity, and check for a real timeout only	1866	but remember the time of last activity, and check for a real timeout only
1840	within the callback:	1867	within the callback:
1841		1868
		1869	ev_tstamp timeout = 60.;
1842	ev_tstamp last_activity; // time of last activity	1870	ev_tstamp last_activity; // time of last activity
		1871	ev_timer timer;
1843		1872
1844	static void	1873	static void
1845	callback (EV_P_ ev_timer *w, int revents)	1874	callback (EV_P_ ev_timer *w, int revents)
1846	{	1875	{
1847	ev_tstamp now = ev_now (EV_A);	1876	// calculate when the timeout would happen
1848	ev_tstamp timeout = last_activity + 60.;	1877	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
1849		1878
1850	// if last_activity + 60. is older than now, we did time out	1879	// if negative, it means we the timeout already occured
1851	if (timeout < now)	1880	if (after < 0.)
1852	{	1881	{
1853	// timeout occurred, take action	1882	// timeout occurred, take action
1854	}	1883	}
1855	else	1884	else
1856	{	1885	{
1857	// callback was invoked, but there was some activity, re-arm	1886	// callback was invoked, but there was some recent
1858	// the watcher to fire in last_activity + 60, which is	1887	// activity. simply restart the timer to time out
1859	// guaranteed to be in the future, so "again" is positive:	1888	// after "after" seconds, which is the earliest time
1860	w->repeat = timeout - now;	1889	// the timeout can occur.
		1890	ev_timer_set (w, after, 0.);
1861	ev_timer_again (EV_A_ w);	1891	ev_timer_start (EV_A_ w);
1862	}	1892	}
1863	}	1893	}
1864		1894
1865	To summarise the callback: first calculate the real timeout (defined	1895	To summarise the callback: first calculate in how many seconds the
1866	as "60 seconds after the last activity"), then check if that time has	1896	timeout will occur (by calculating the absolute time when it would occur,
1867	been reached, which means something I<did>, in fact, time out. Otherwise	1897	C<last_activity + timeout>, and subtracting the current time, C<ev_now
1868	the callback was invoked too early (C<timeout> is in the future), so	1898	(EV_A)> from that).
1869	re-schedule the timer to fire at that future time, to see if maybe we have
1870	a timeout then.
1871		1899
1872	Note how C<ev_timer_again> is used, taking advantage of the	1900	If this value is negative, then we are already past the timeout, i.e. we
1873	C<ev_timer_again> optimisation when the timer is already running.	1901	timed out, and need to do whatever is needed in this case.
		1902
		1903	Otherwise, we now the earliest time at which the timeout would trigger,
		1904	and simply start the timer with this timeout value.
		1905
		1906	In other words, each time the callback is invoked it will check whether
		1907	the timeout cocured. If not, it will simply reschedule itself to check
		1908	again at the earliest time it could time out. Rinse. Repeat.
1874		1909
1875	This scheme causes more callback invocations (about one every 60 seconds	1910	This scheme causes more callback invocations (about one every 60 seconds
1876	minus half the average time between activity), but virtually no calls to	1911	minus half the average time between activity), but virtually no calls to
1877	libev to change the timeout.	1912	libev to change the timeout.
1878		1913
1879	To start the timer, simply initialise the watcher and set C<last_activity>	1914	To start the machinery, simply initialise the watcher and set
1880	to the current time (meaning we just have some activity :), then call the	1915	C<last_activity> to the current time (meaning there was some activity just
1881	callback, which will "do the right thing" and start the timer:	1916	now), then call the callback, which will "do the right thing" and start
		1917	the timer:
1882		1918
		1919	last_activity = ev_now (EV_A);
1883	ev_init (timer, callback);	1920	ev_init (&timer, callback);
1884	last_activity = ev_now (loop);	1921	callback (EV_A_ &timer, 0);
1885	callback (loop, timer, EV_TIMER);
1886		1922
1887	And when there is some activity, simply store the current time in	1923	When there is some activity, simply store the current time in
1888	C<last_activity>, no libev calls at all:	1924	C<last_activity>, no libev calls at all:
1889		1925
		1926	if (activity detected)
1890	last_activity = ev_now (loop);	1927	last_activity = ev_now (EV_A);
		1928
		1929	When your timeout value changes, then the timeout can be changed by simply
		1930	providing a new value, stopping the timer and calling the callback, which
		1931	will agaion do the right thing (for example, time out immediately :).
		1932
		1933	timeout = new_value;
		1934	ev_timer_stop (EV_A_ &timer);
		1935	callback (EV_A_ &timer, 0);
1891		1936
1892	This technique is slightly more complex, but in most cases where the	1937	This technique is slightly more complex, but in most cases where the
1893	time-out is unlikely to be triggered, much more efficient.	1938	time-out is unlikely to be triggered, much more efficient.
1894
1895	Changing the timeout is trivial as well (if it isn't hard-coded in the
1896	callback :) - just change the timeout and invoke the callback, which will
1897	fix things for you.
1898		1939
1899	=item 4. Wee, just use a double-linked list for your timeouts.	1940	=item 4. Wee, just use a double-linked list for your timeouts.
1900		1941
1901	If there is not one request, but many thousands (millions...), all	1942	If there is not one request, but many thousands (millions...), all
1902	employing some kind of timeout with the same timeout value, then one can	1943	employing some kind of timeout with the same timeout value, then one can
…		…
1929	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is	1970	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
1930	rather complicated, but extremely efficient, something that really pays	1971	rather complicated, but extremely efficient, something that really pays
1931	off after the first million or so of active timers, i.e. it's usually	1972	off after the first million or so of active timers, i.e. it's usually
1932	overkill :)	1973	overkill :)
1933		1974
		1975	=head3 The special problem of being too early
		1976
		1977	If you ask a timer to call your callback after three seconds, then
		1978	you expect it to be invoked after three seconds - but of course, this
		1979	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		1980	guaranteed to any precision by libev - imagine somebody suspending the
		1981	process with a STOP signal for a few hours for example.
		1982
		1983	So, libev tries to invoke your callback as soon as possible I<after> the
		1984	delay has occurred, but cannot guarantee this.
		1985
		1986	A less obvious failure mode is calling your callback too early: many event
		1987	loops compare timestamps with a "elapsed delay >= requested delay", but
		1988	this can cause your callback to be invoked much earlier than you would
		1989	expect.
		1990
		1991	To see why, imagine a system with a clock that only offers full second
		1992	resolution (think windows if you can't come up with a broken enough OS
		1993	yourself). If you schedule a one-second timer at the time 500.9, then the
		1994	event loop will schedule your timeout to elapse at a system time of 500
		1995	(500.9 truncated to the resolution) + 1, or 501.
		1996
		1997	If an event library looks at the timeout 0.1s later, it will see "501 >=
		1998	501" and invoke the callback 0.1s after it was started, even though a
		1999	one-second delay was requested - this is being "too early", despite best
		2000	intentions.
		2001
		2002	This is the reason why libev will never invoke the callback if the elapsed
		2003	delay equals the requested delay, but only when the elapsed delay is
		2004	larger than the requested delay. In the example above, libev would only invoke
		2005	the callback at system time 502, or 1.1s after the timer was started.
		2006
		2007	So, while libev cannot guarantee that your callback will be invoked
		2008	exactly when requested, it I<can> and I<does> guarantee that the requested
		2009	delay has actually elapsed, or in other words, it always errs on the "too
		2010	late" side of things.
		2011
1934	=head3 The special problem of time updates	2012	=head3 The special problem of time updates
1935		2013
1936	Establishing the current time is a costly operation (it usually takes at	2014	Establishing the current time is a costly operation (it usually takes
1937	least two system calls): EV therefore updates its idea of the current	2015	at least one system call): EV therefore updates its idea of the current
1938	time only before and after C<ev_run> collects new events, which causes a	2016	time only before and after C<ev_run> collects new events, which causes a
1939	growing difference between C<ev_now ()> and C<ev_time ()> when handling	2017	growing difference between C<ev_now ()> and C<ev_time ()> when handling
1940	lots of events in one iteration.	2018	lots of events in one iteration.
1941		2019
1942	The relative timeouts are calculated relative to the C<ev_now ()>	2020	The relative timeouts are calculated relative to the C<ev_now ()>
…		…
1948	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2026	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1949		2027
1950	If the event loop is suspended for a long time, you can also force an	2028	If the event loop is suspended for a long time, you can also force an
1951	update of the time returned by C<ev_now ()> by calling C<ev_now_update	2029	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1952	()>.	2030	()>.
		2031
		2032	=head3 The special problem of unsynchronised clocks
		2033
		2034	Modern systems have a variety of clocks - libev itself uses the normal
		2035	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2036	jumps).
		2037
		2038	Neither of these clocks is synchronised with each other or any other clock
		2039	on the system, so C<ev_time ()> might return a considerably different time
		2040	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2041	a call to C<gettimeofday> might return a second count that is one higher
		2042	than a directly following call to C<time>.
		2043
		2044	The moral of this is to only compare libev-related timestamps with
		2045	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2046	a second or so.
		2047
		2048	One more problem arises due to this lack of synchronisation: if libev uses
		2049	the system monotonic clock and you compare timestamps from C<ev_time>
		2050	or C<ev_now> from when you started your timer and when your callback is
		2051	invoked, you will find that sometimes the callback is a bit "early".
		2052
		2053	This is because C<ev_timer>s work in real time, not wall clock time, so
		2054	libev makes sure your callback is not invoked before the delay happened,
		2055	I<measured according to the real time>, not the system clock.
		2056
		2057	If your timeouts are based on a physical timescale (e.g. "time out this
		2058	connection after 100 seconds") then this shouldn't bother you as it is
		2059	exactly the right behaviour.
		2060
		2061	If you want to compare wall clock/system timestamps to your timers, then
		2062	you need to use C<ev_periodic>s, as these are based on the wall clock
		2063	time, where your comparisons will always generate correct results.
1953		2064
1954	=head3 The special problems of suspended animation	2065	=head3 The special problems of suspended animation
1955		2066
1956	When you leave the server world it is quite customary to hit machines that	2067	When you leave the server world it is quite customary to hit machines that
1957	can suspend/hibernate - what happens to the clocks during such a suspend?	2068	can suspend/hibernate - what happens to the clocks during such a suspend?
…		…
2001	keep up with the timer (because it takes longer than those 10 seconds to	2112	keep up with the timer (because it takes longer than those 10 seconds to
2002	do stuff) the timer will not fire more than once per event loop iteration.	2113	do stuff) the timer will not fire more than once per event loop iteration.
2003		2114
2004	=item ev_timer_again (loop, ev_timer *)	2115	=item ev_timer_again (loop, ev_timer *)
2005		2116
2006	This will act as if the timer timed out and restart it again if it is	2117	This will act as if the timer timed out, and restarts it again if it is
2007	repeating. The exact semantics are:	2118	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2119	timeout to the C<repeat> value and calling C<ev_timer_start>.
2008		2120
		2121	The exact semantics are as in the following rules, all of which will be
		2122	applied to the watcher:
		2123
		2124	=over 4
		2125
2009	If the timer is pending, its pending status is cleared.	2126	=item If the timer is pending, the pending status is always cleared.
2010		2127
2011	If the timer is started but non-repeating, stop it (as if it timed out).	2128	=item If the timer is started but non-repeating, stop it (as if it timed
		2129	out, without invoking it).
2012		2130
2013	If the timer is repeating, either start it if necessary (with the	2131	=item If the timer is repeating, make the C<repeat> value the new timeout
2014	C<repeat> value), or reset the running timer to the C<repeat> value.	2132	and start the timer, if necessary.
		2133
		2134	=back
2015		2135
2016	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a	2136	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
2017	usage example.	2137	usage example.
2018		2138
2019	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)	2139	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
…		…
2141		2261
2142	Another way to think about it (for the mathematically inclined) is that	2262	Another way to think about it (for the mathematically inclined) is that
2143	C<ev_periodic> will try to run the callback in this mode at the next possible	2263	C<ev_periodic> will try to run the callback in this mode at the next possible
2144	time where C<time = offset (mod interval)>, regardless of any time jumps.	2264	time where C<time = offset (mod interval)>, regardless of any time jumps.
2145		2265
2146	For numerical stability it is preferable that the C<offset> value is near	2266	The C<interval> I<MUST> be positive, and for numerical stability, the
2147	C<ev_now ()> (the current time), but there is no range requirement for	2267	interval value should be higher than C<1/8192> (which is around 100
2148	this value, and in fact is often specified as zero.	2268	microseconds) and C<offset> should be higher than C<0> and should have
		2269	at most a similar magnitude as the current time (say, within a factor of
		2270	ten). Typical values for offset are, in fact, C<0> or something between
		2271	C<0> and C<interval>, which is also the recommended range.
2149		2272
2150	Note also that there is an upper limit to how often a timer can fire (CPU	2273	Note also that there is an upper limit to how often a timer can fire (CPU
2151	speed for example), so if C<interval> is very small then timing stability	2274	speed for example), so if C<interval> is very small then timing stability
2152	will of course deteriorate. Libev itself tries to be exact to be about one	2275	will of course deteriorate. Libev itself tries to be exact to be about one
2153	millisecond (if the OS supports it and the machine is fast enough).	2276	millisecond (if the OS supports it and the machine is fast enough).
…		…
2296	=head3 The special problem of inheritance over fork/execve/pthread_create	2419	=head3 The special problem of inheritance over fork/execve/pthread_create
2297		2420
2298	Both the signal mask (C<sigprocmask>) and the signal disposition	2421	Both the signal mask (C<sigprocmask>) and the signal disposition
2299	(C<sigaction>) are unspecified after starting a signal watcher (and after	2422	(C<sigaction>) are unspecified after starting a signal watcher (and after
2300	stopping it again), that is, libev might or might not block the signal,	2423	stopping it again), that is, libev might or might not block the signal,
2301	and might or might not set or restore the installed signal handler.	2424	and might or might not set or restore the installed signal handler (but
		2425	see C<EVFLAG_NOSIGMASK>).
2302		2426
2303	While this does not matter for the signal disposition (libev never	2427	While this does not matter for the signal disposition (libev never
2304	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on	2428	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
2305	C<execve>), this matters for the signal mask: many programs do not expect	2429	C<execve>), this matters for the signal mask: many programs do not expect
2306	certain signals to be blocked.	2430	certain signals to be blocked.
…		…
2319	I<has> to modify the signal mask, at least temporarily.	2443	I<has> to modify the signal mask, at least temporarily.
2320		2444
2321	So I can't stress this enough: I<If you do not reset your signal mask when	2445	So I can't stress this enough: I<If you do not reset your signal mask when
2322	you expect it to be empty, you have a race condition in your code>. This	2446	you expect it to be empty, you have a race condition in your code>. This
2323	is not a libev-specific thing, this is true for most event libraries.	2447	is not a libev-specific thing, this is true for most event libraries.
		2448
		2449	=head3 The special problem of threads signal handling
		2450
		2451	POSIX threads has problematic signal handling semantics, specifically,
		2452	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2453	threads in a process block signals, which is hard to achieve.
		2454
		2455	When you want to use sigwait (or mix libev signal handling with your own
		2456	for the same signals), you can tackle this problem by globally blocking
		2457	all signals before creating any threads (or creating them with a fully set
		2458	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2459	loops. Then designate one thread as "signal receiver thread" which handles
		2460	these signals. You can pass on any signals that libev might be interested
		2461	in by calling C<ev_feed_signal>.
2324		2462
2325	=head3 Watcher-Specific Functions and Data Members	2463	=head3 Watcher-Specific Functions and Data Members
2326		2464
2327	=over 4	2465	=over 4
2328		2466
…		…
3163	atexit (program_exits);	3301	atexit (program_exits);
3164		3302
3165		3303
3166	=head2 C<ev_async> - how to wake up an event loop	3304	=head2 C<ev_async> - how to wake up an event loop
3167		3305
3168	In general, you cannot use an C<ev_run> from multiple threads or other	3306	In general, you cannot use an C<ev_loop> from multiple threads or other
3169	asynchronous sources such as signal handlers (as opposed to multiple event	3307	asynchronous sources such as signal handlers (as opposed to multiple event
3170	loops - those are of course safe to use in different threads).	3308	loops - those are of course safe to use in different threads).
3171		3309
3172	Sometimes, however, you need to wake up an event loop you do not control,	3310	Sometimes, however, you need to wake up an event loop you do not control,
3173	for example because it belongs to another thread. This is what C<ev_async>	3311	for example because it belongs to another thread. This is what C<ev_async>
…		…
3175	it by calling C<ev_async_send>, which is thread- and signal safe.	3313	it by calling C<ev_async_send>, which is thread- and signal safe.
3176		3314
3177	This functionality is very similar to C<ev_signal> watchers, as signals,	3315	This functionality is very similar to C<ev_signal> watchers, as signals,
3178	too, are asynchronous in nature, and signals, too, will be compressed	3316	too, are asynchronous in nature, and signals, too, will be compressed
3179	(i.e. the number of callback invocations may be less than the number of	3317	(i.e. the number of callback invocations may be less than the number of
3180	C<ev_async_sent> calls).	3318	C<ev_async_sent> calls). In fact, you could use signal watchers as a kind
3181		3319	of "global async watchers" by using a watcher on an otherwise unused
3182	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not	3320	signal, and C<ev_feed_signal> to signal this watcher from another thread,
3183	just the default loop.	3321	even without knowing which loop owns the signal.
3184		3322
3185	=head3 Queueing	3323	=head3 Queueing
3186		3324
3187	C<ev_async> does not support queueing of data in any way. The reason	3325	C<ev_async> does not support queueing of data in any way. The reason
3188	is that the author does not know of a simple (or any) algorithm for a	3326	is that the author does not know of a simple (or any) algorithm for a
…		…
3280	trust me.	3418	trust me.
3281		3419
3282	=item ev_async_send (loop, ev_async *)	3420	=item ev_async_send (loop, ev_async *)
3283		3421
3284	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3422	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
3285	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	3423	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3424	returns.
		3425
3286	C<ev_feed_event>, this call is safe to do from other threads, signal or	3426	Unlike C<ev_feed_event>, this call is safe to do from other threads,
3287	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	3427	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3288	section below on what exactly this means).	3428	embedding section below on what exactly this means).
3289		3429
3290	Note that, as with other watchers in libev, multiple events might get	3430	Note that, as with other watchers in libev, multiple events might get
3291	compressed into a single callback invocation (another way to look at this	3431	compressed into a single callback invocation (another way to look at
3292	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,	3432	this is that C<ev_async> watchers are level-triggered: they are set on
3293	reset when the event loop detects that).	3433	C<ev_async_send>, reset when the event loop detects that).
3294		3434
3295	This call incurs the overhead of a system call only once per event loop	3435	This call incurs the overhead of at most one extra system call per event
3296	iteration, so while the overhead might be noticeable, it doesn't apply to	3436	loop iteration, if the event loop is blocked, and no syscall at all if
3297	repeated calls to C<ev_async_send> for the same event loop.	3437	the event loop (or your program) is processing events. That means that
		3438	repeated calls are basically free (there is no need to avoid calls for
		3439	performance reasons) and that the overhead becomes smaller (typically
		3440	zero) under load.
3298		3441
3299	=item bool = ev_async_pending (ev_async *)	3442	=item bool = ev_async_pending (ev_async *)
3300		3443
3301	Returns a non-zero value when C<ev_async_send> has been called on the	3444	Returns a non-zero value when C<ev_async_send> has been called on the
3302	watcher but the event has not yet been processed (or even noted) by the	3445	watcher but the event has not yet been processed (or even noted) by the
…		…
3357	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3500	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3358		3501
3359	=item ev_feed_fd_event (loop, int fd, int revents)	3502	=item ev_feed_fd_event (loop, int fd, int revents)
3360		3503
3361	Feed an event on the given fd, as if a file descriptor backend detected	3504	Feed an event on the given fd, as if a file descriptor backend detected
3362	the given events it.	3505	the given events.
3363		3506
3364	=item ev_feed_signal_event (loop, int signum)	3507	=item ev_feed_signal_event (loop, int signum)
3365		3508
3366	Feed an event as if the given signal occurred (C<loop> must be the default	3509	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3367	loop!).	3510	which is async-safe.
3368		3511
3369	=back	3512	=back
3370		3513
3371		3514
3372	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)	3515	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
3373		3516
3374	This section explains some common idioms that are not immediately	3517	This section explains some common idioms that are not immediately
3375	obvious. Note that examples are sprinkled over the whole manual, and this	3518	obvious. Note that examples are sprinkled over the whole manual, and this
3376	section only contains stuff that wouldn't fit anywhere else.	3519	section only contains stuff that wouldn't fit anywhere else.
3377		3520
3378	=over 4	3521	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
3379		3522
3380	=item Model/nested event loop invocations and exit conditions.	3523	Each watcher has, by default, a C<void *data> member that you can read
		3524	or modify at any time: libev will completely ignore it. This can be used
		3525	to associate arbitrary data with your watcher. If you need more data and
		3526	don't want to allocate memory separately and store a pointer to it in that
		3527	data member, you can also "subclass" the watcher type and provide your own
		3528	data:
		3529
		3530	struct my_io
		3531	{
		3532	ev_io io;
		3533	int otherfd;
		3534	void *somedata;
		3535	struct whatever *mostinteresting;
		3536	};
		3537
		3538	...
		3539	struct my_io w;
		3540	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3541
		3542	And since your callback will be called with a pointer to the watcher, you
		3543	can cast it back to your own type:
		3544
		3545	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3546	{
		3547	struct my_io *w = (struct my_io *)w_;
		3548	...
		3549	}
		3550
		3551	More interesting and less C-conformant ways of casting your callback
		3552	function type instead have been omitted.
		3553
		3554	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3555
		3556	Another common scenario is to use some data structure with multiple
		3557	embedded watchers, in effect creating your own watcher that combines
		3558	multiple libev event sources into one "super-watcher":
		3559
		3560	struct my_biggy
		3561	{
		3562	int some_data;
		3563	ev_timer t1;
		3564	ev_timer t2;
		3565	}
		3566
		3567	In this case getting the pointer to C<my_biggy> is a bit more
		3568	complicated: Either you store the address of your C<my_biggy> struct in
		3569	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3570	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3571	real programmers):
		3572
		3573	#include <stddef.h>
		3574
		3575	static void
		3576	t1_cb (EV_P_ ev_timer *w, int revents)
		3577	{
		3578	struct my_biggy big = (struct my_biggy *)
		3579	(((char *)w) - offsetof (struct my_biggy, t1));
		3580	}
		3581
		3582	static void
		3583	t2_cb (EV_P_ ev_timer *w, int revents)
		3584	{
		3585	struct my_biggy big = (struct my_biggy *)
		3586	(((char *)w) - offsetof (struct my_biggy, t2));
		3587	}
		3588
		3589	=head2 AVOIDING FINISHING BEFORE RETURNING
		3590
		3591	Often you have structures like this in event-based programs:
		3592
		3593	callback ()
		3594	{
		3595	free (request);
		3596	}
		3597
		3598	request = start_new_request (..., callback);
		3599
		3600	The intent is to start some "lengthy" operation. The C<request> could be
		3601	used to cancel the operation, or do other things with it.
		3602
		3603	It's not uncommon to have code paths in C<start_new_request> that
		3604	immediately invoke the callback, for example, to report errors. Or you add
		3605	some caching layer that finds that it can skip the lengthy aspects of the
		3606	operation and simply invoke the callback with the result.
		3607
		3608	The problem here is that this will happen I<before> C<start_new_request>
		3609	has returned, so C<request> is not set.
		3610
		3611	Even if you pass the request by some safer means to the callback, you
		3612	might want to do something to the request after starting it, such as
		3613	canceling it, which probably isn't working so well when the callback has
		3614	already been invoked.
		3615
		3616	A common way around all these issues is to make sure that
		3617	C<start_new_request> I<always> returns before the callback is invoked. If
		3618	C<start_new_request> immediately knows the result, it can artificially
		3619	delay invoking the callback by e.g. using a C<prepare> or C<idle> watcher
		3620	for example, or more sneakily, by reusing an existing (stopped) watcher
		3621	and pushing it into the pending queue:
		3622
		3623	ev_set_cb (watcher, callback);
		3624	ev_feed_event (EV_A_ watcher, 0);
		3625
		3626	This way, C<start_new_request> can safely return before the callback is
		3627	invoked, while not delaying callback invocation too much.
		3628
		3629	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
3381		3630
3382	Often (especially in GUI toolkits) there are places where you have	3631	Often (especially in GUI toolkits) there are places where you have
3383	I<modal> interaction, which is most easily implemented by recursively	3632	I<modal> interaction, which is most easily implemented by recursively
3384	invoking C<ev_run>.	3633	invoking C<ev_run>.
3385		3634
…		…
3397	int exit_main_loop = 0;	3646	int exit_main_loop = 0;
3398		3647
3399	while (!exit_main_loop)	3648	while (!exit_main_loop)
3400	ev_run (EV_DEFAULT_ EVRUN_ONCE);	3649	ev_run (EV_DEFAULT_ EVRUN_ONCE);
3401		3650
3402	// in a model watcher	3651	// in a modal watcher
3403	int exit_nested_loop = 0;	3652	int exit_nested_loop = 0;
3404		3653
3405	while (!exit_nested_loop)	3654	while (!exit_nested_loop)
3406	ev_run (EV_A_ EVRUN_ONCE);	3655	ev_run (EV_A_ EVRUN_ONCE);
3407		3656
…		…
3414	exit_main_loop = 1;	3663	exit_main_loop = 1;
3415		3664
3416	// exit both	3665	// exit both
3417	exit_main_loop = exit_nested_loop = 1;	3666	exit_main_loop = exit_nested_loop = 1;
3418		3667
3419	=back	3668	=head2 THREAD LOCKING EXAMPLE
		3669
		3670	Here is a fictitious example of how to run an event loop in a different
		3671	thread from where callbacks are being invoked and watchers are
		3672	created/added/removed.
		3673
		3674	For a real-world example, see the C<EV::Loop::Async> perl module,
		3675	which uses exactly this technique (which is suited for many high-level
		3676	languages).
		3677
		3678	The example uses a pthread mutex to protect the loop data, a condition
		3679	variable to wait for callback invocations, an async watcher to notify the
		3680	event loop thread and an unspecified mechanism to wake up the main thread.
		3681
		3682	First, you need to associate some data with the event loop:
		3683
		3684	typedef struct {
		3685	mutex_t lock; /* global loop lock */
		3686	ev_async async_w;
		3687	thread_t tid;
		3688	cond_t invoke_cv;
		3689	} userdata;
		3690
		3691	void prepare_loop (EV_P)
		3692	{
		3693	// for simplicity, we use a static userdata struct.
		3694	static userdata u;
		3695
		3696	ev_async_init (&u->async_w, async_cb);
		3697	ev_async_start (EV_A_ &u->async_w);
		3698
		3699	pthread_mutex_init (&u->lock, 0);
		3700	pthread_cond_init (&u->invoke_cv, 0);
		3701
		3702	// now associate this with the loop
		3703	ev_set_userdata (EV_A_ u);
		3704	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3705	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3706
		3707	// then create the thread running ev_run
		3708	pthread_create (&u->tid, 0, l_run, EV_A);
		3709	}
		3710
		3711	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3712	solely to wake up the event loop so it takes notice of any new watchers
		3713	that might have been added:
		3714
		3715	static void
		3716	async_cb (EV_P_ ev_async *w, int revents)
		3717	{
		3718	// just used for the side effects
		3719	}
		3720
		3721	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3722	protecting the loop data, respectively.
		3723
		3724	static void
		3725	l_release (EV_P)
		3726	{
		3727	userdata *u = ev_userdata (EV_A);
		3728	pthread_mutex_unlock (&u->lock);
		3729	}
		3730
		3731	static void
		3732	l_acquire (EV_P)
		3733	{
		3734	userdata *u = ev_userdata (EV_A);
		3735	pthread_mutex_lock (&u->lock);
		3736	}
		3737
		3738	The event loop thread first acquires the mutex, and then jumps straight
		3739	into C<ev_run>:
		3740
		3741	void *
		3742	l_run (void *thr_arg)
		3743	{
		3744	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3745
		3746	l_acquire (EV_A);
		3747	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3748	ev_run (EV_A_ 0);
		3749	l_release (EV_A);
		3750
		3751	return 0;
		3752	}
		3753
		3754	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3755	signal the main thread via some unspecified mechanism (signals? pipe
		3756	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3757	have been called (in a while loop because a) spurious wakeups are possible
		3758	and b) skipping inter-thread-communication when there are no pending
		3759	watchers is very beneficial):
		3760
		3761	static void
		3762	l_invoke (EV_P)
		3763	{
		3764	userdata *u = ev_userdata (EV_A);
		3765
		3766	while (ev_pending_count (EV_A))
		3767	{
		3768	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3769	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3770	}
		3771	}
		3772
		3773	Now, whenever the main thread gets told to invoke pending watchers, it
		3774	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3775	thread to continue:
		3776
		3777	static void
		3778	real_invoke_pending (EV_P)
		3779	{
		3780	userdata *u = ev_userdata (EV_A);
		3781
		3782	pthread_mutex_lock (&u->lock);
		3783	ev_invoke_pending (EV_A);
		3784	pthread_cond_signal (&u->invoke_cv);
		3785	pthread_mutex_unlock (&u->lock);
		3786	}
		3787
		3788	Whenever you want to start/stop a watcher or do other modifications to an
		3789	event loop, you will now have to lock:
		3790
		3791	ev_timer timeout_watcher;
		3792	userdata *u = ev_userdata (EV_A);
		3793
		3794	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3795
		3796	pthread_mutex_lock (&u->lock);
		3797	ev_timer_start (EV_A_ &timeout_watcher);
		3798	ev_async_send (EV_A_ &u->async_w);
		3799	pthread_mutex_unlock (&u->lock);
		3800
		3801	Note that sending the C<ev_async> watcher is required because otherwise
		3802	an event loop currently blocking in the kernel will have no knowledge
		3803	about the newly added timer. By waking up the loop it will pick up any new
		3804	watchers in the next event loop iteration.
		3805
		3806	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3807
		3808	While the overhead of a callback that e.g. schedules a thread is small, it
		3809	is still an overhead. If you embed libev, and your main usage is with some
		3810	kind of threads or coroutines, you might want to customise libev so that
		3811	doesn't need callbacks anymore.
		3812
		3813	Imagine you have coroutines that you can switch to using a function
		3814	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3815	and that due to some magic, the currently active coroutine is stored in a
		3816	global called C<current_coro>. Then you can build your own "wait for libev
		3817	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3818	the differing C<;> conventions):
		3819
		3820	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3821	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3822
		3823	That means instead of having a C callback function, you store the
		3824	coroutine to switch to in each watcher, and instead of having libev call
		3825	your callback, you instead have it switch to that coroutine.
		3826
		3827	A coroutine might now wait for an event with a function called
		3828	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3829	matter when, or whether the watcher is active or not when this function is
		3830	called):
		3831
		3832	void
		3833	wait_for_event (ev_watcher *w)
		3834	{
		3835	ev_cb_set (w) = current_coro;
		3836	switch_to (libev_coro);
		3837	}
		3838
		3839	That basically suspends the coroutine inside C<wait_for_event> and
		3840	continues the libev coroutine, which, when appropriate, switches back to
		3841	this or any other coroutine.
		3842
		3843	You can do similar tricks if you have, say, threads with an event queue -
		3844	instead of storing a coroutine, you store the queue object and instead of
		3845	switching to a coroutine, you push the watcher onto the queue and notify
		3846	any waiters.
		3847
		3848	To embed libev, see L<EMBEDDING>, but in short, it's easiest to create two
		3849	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3850
		3851	// my_ev.h
		3852	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3853	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
		3854	#include "../libev/ev.h"
		3855
		3856	// my_ev.c
		3857	#define EV_H "my_ev.h"
		3858	#include "../libev/ev.c"
		3859
		3860	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3861	F<my_ev.c> into your project. When properly specifying include paths, you
		3862	can even use F<ev.h> as header file name directly.
3420		3863
3421		3864
3422	=head1 LIBEVENT EMULATION	3865	=head1 LIBEVENT EMULATION
3423		3866
3424	Libev offers a compatibility emulation layer for libevent. It cannot	3867	Libev offers a compatibility emulation layer for libevent. It cannot
…		…
3427	=over 4	3870	=over 4
3428		3871
3429	=item * Only the libevent-1.4.1-beta API is being emulated.	3872	=item * Only the libevent-1.4.1-beta API is being emulated.
3430		3873
3431	This was the newest libevent version available when libev was implemented,	3874	This was the newest libevent version available when libev was implemented,
3432	and is still mostly uncanged in 2010.	3875	and is still mostly unchanged in 2010.
3433		3876
3434	=item * Use it by including <event.h>, as usual.	3877	=item * Use it by including <event.h>, as usual.
3435		3878
3436	=item * The following members are fully supported: ev_base, ev_callback,	3879	=item * The following members are fully supported: ev_base, ev_callback,
3437	ev_arg, ev_fd, ev_res, ev_events.	3880	ev_arg, ev_fd, ev_res, ev_events.
…		…
3472	Care has been taken to keep the overhead low. The only data member the C++	3915	Care has been taken to keep the overhead low. The only data member the C++
3473	classes add (compared to plain C-style watchers) is the event loop pointer	3916	classes add (compared to plain C-style watchers) is the event loop pointer
3474	that the watcher is associated with (or no additional members at all if	3917	that the watcher is associated with (or no additional members at all if
3475	you disable C<EV_MULTIPLICITY> when embedding libev).	3918	you disable C<EV_MULTIPLICITY> when embedding libev).
3476		3919
3477	Currently, functions, and static and non-static member functions can be	3920	Currently, functions, static and non-static member functions and classes
3478	used as callbacks. Other types should be easy to add as long as they only	3921	with C<operator ()> can be used as callbacks. Other types should be easy
3479	need one additional pointer for context. If you need support for other	3922	to add as long as they only need one additional pointer for context. If
3480	types of functors please contact the author (preferably after implementing	3923	you need support for other types of functors please contact the author
3481	it).	3924	(preferably after implementing it).
		3925
		3926	For all this to work, your C++ compiler either has to use the same calling
		3927	conventions as your C compiler (for static member functions), or you have
		3928	to embed libev and compile libev itself as C++.
3482		3929
3483	Here is a list of things available in the C<ev> namespace:	3930	Here is a list of things available in the C<ev> namespace:
3484		3931
3485	=over 4	3932	=over 4
3486		3933
…		…
3496	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.	3943	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
3497		3944
3498	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of	3945	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
3499	the same name in the C<ev> namespace, with the exception of C<ev_signal>	3946	the same name in the C<ev> namespace, with the exception of C<ev_signal>
3500	which is called C<ev::sig> to avoid clashes with the C<signal> macro	3947	which is called C<ev::sig> to avoid clashes with the C<signal> macro
3501	defines by many implementations.	3948	defined by many implementations.
3502		3949
3503	All of those classes have these methods:	3950	All of those classes have these methods:
3504		3951
3505	=over 4	3952	=over 4
3506		3953
…		…
3639	watchers in the constructor.	4086	watchers in the constructor.
3640		4087
3641	class myclass	4088	class myclass
3642	{	4089	{
3643	ev::io io ; void io_cb (ev::io &w, int revents);	4090	ev::io io ; void io_cb (ev::io &w, int revents);
3644	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);	4091	ev::io io2 ; void io2_cb (ev::io &w, int revents);
3645	ev::idle idle; void idle_cb (ev::idle &w, int revents);	4092	ev::idle idle; void idle_cb (ev::idle &w, int revents);
3646		4093
3647	myclass (int fd)	4094	myclass (int fd)
3648	{	4095	{
3649	io .set <myclass, &myclass::io_cb > (this);	4096	io .set <myclass, &myclass::io_cb > (this);
…		…
3700	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.	4147	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
3701		4148
3702	=item D	4149	=item D
3703		4150
3704	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	4151	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3705	be found at L<http://proj.llucax.com.ar/wiki/evd>.	4152	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
3706		4153
3707	=item Ocaml	4154	=item Ocaml
3708		4155
3709	Erkki Seppala has written Ocaml bindings for libev, to be found at	4156	Erkki Seppala has written Ocaml bindings for libev, to be found at
3710	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.	4157	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
…		…
3758	suitable for use with C<EV_A>.	4205	suitable for use with C<EV_A>.
3759		4206
3760	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	4207	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
3761		4208
3762	Similar to the other two macros, this gives you the value of the default	4209	Similar to the other two macros, this gives you the value of the default
3763	loop, if multiple loops are supported ("ev loop default").	4210	loop, if multiple loops are supported ("ev loop default"). The default loop
		4211	will be initialised if it isn't already initialised.
		4212
		4213	For non-multiplicity builds, these macros do nothing, so you always have
		4214	to initialise the loop somewhere.
3764		4215
3765	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>	4216	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3766		4217
3767	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the	4218	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3768	default loop has been initialised (C<UC> == unchecked). Their behaviour	4219	default loop has been initialised (C<UC> == unchecked). Their behaviour
…		…
3913	supported). It will also not define any of the structs usually found in	4364	supported). It will also not define any of the structs usually found in
3914	F<event.h> that are not directly supported by the libev core alone.	4365	F<event.h> that are not directly supported by the libev core alone.
3915		4366
3916	In standalone mode, libev will still try to automatically deduce the	4367	In standalone mode, libev will still try to automatically deduce the
3917	configuration, but has to be more conservative.	4368	configuration, but has to be more conservative.
		4369
		4370	=item EV_USE_FLOOR
		4371
		4372	If defined to be C<1>, libev will use the C<floor ()> function for its
		4373	periodic reschedule calculations, otherwise libev will fall back on a
		4374	portable (slower) implementation. If you enable this, you usually have to
		4375	link against libm or something equivalent. Enabling this when the C<floor>
		4376	function is not available will fail, so the safe default is to not enable
		4377	this.
3918		4378
3919	=item EV_USE_MONOTONIC	4379	=item EV_USE_MONOTONIC
3920		4380
3921	If defined to be C<1>, libev will try to detect the availability of the	4381	If defined to be C<1>, libev will try to detect the availability of the
3922	monotonic clock option at both compile time and runtime. Otherwise no	4382	monotonic clock option at both compile time and runtime. Otherwise no
…		…
4052	If defined to be C<1>, libev will compile in support for the Linux inotify	4512	If defined to be C<1>, libev will compile in support for the Linux inotify
4053	interface to speed up C<ev_stat> watchers. Its actual availability will	4513	interface to speed up C<ev_stat> watchers. Its actual availability will
4054	be detected at runtime. If undefined, it will be enabled if the headers	4514	be detected at runtime. If undefined, it will be enabled if the headers
4055	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.	4515	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4056		4516
		4517	=item EV_NO_SMP
		4518
		4519	If defined to be C<1>, libev will assume that memory is always coherent
		4520	between threads, that is, threads can be used, but threads never run on
		4521	different cpus (or different cpu cores). This reduces dependencies
		4522	and makes libev faster.
		4523
		4524	=item EV_NO_THREADS
		4525
		4526	If defined to be C<1>, libev will assume that it will never be called
		4527	from different threads, which is a stronger assumption than C<EV_NO_SMP>,
		4528	above. This reduces dependencies and makes libev faster.
		4529
4057	=item EV_ATOMIC_T	4530	=item EV_ATOMIC_T
4058		4531
4059	Libev requires an integer type (suitable for storing C<0> or C<1>) whose	4532	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
4060	access is atomic with respect to other threads or signal contexts. No such	4533	access is atomic and serialised with respect to other threads or signal
4061	type is easily found in the C language, so you can provide your own type	4534	contexts. No such type is easily found in the C language, so you can
4062	that you know is safe for your purposes. It is used both for signal handler "locking"	4535	provide your own type that you know is safe for your purposes. It is used
4063	as well as for signal and thread safety in C<ev_async> watchers.	4536	both for signal handler "locking" as well as for signal and thread safety
		4537	in C<ev_async> watchers.
4064		4538
4065	In the absence of this define, libev will use C<sig_atomic_t volatile>	4539	In the absence of this define, libev will use C<sig_atomic_t volatile>
4066	(from F<signal.h>), which is usually good enough on most platforms.	4540	(from F<signal.h>), which is usually good enough on most platforms,
		4541	although strictly speaking using a type that also implies a memory fence
		4542	is required.
4067		4543
4068	=item EV_H (h)	4544	=item EV_H (h)
4069		4545
4070	The name of the F<ev.h> header file used to include it. The default if	4546	The name of the F<ev.h> header file used to include it. The default if
4071	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be	4547	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
…		…
4095	will have the C<struct ev_loop *> as first argument, and you can create	4571	will have the C<struct ev_loop *> as first argument, and you can create
4096	additional independent event loops. Otherwise there will be no support	4572	additional independent event loops. Otherwise there will be no support
4097	for multiple event loops and there is no first event loop pointer	4573	for multiple event loops and there is no first event loop pointer
4098	argument. Instead, all functions act on the single default loop.	4574	argument. Instead, all functions act on the single default loop.
4099		4575
		4576	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4577	default loop when multiplicity is switched off - you always have to
		4578	initialise the loop manually in this case.
		4579
4100	=item EV_MINPRI	4580	=item EV_MINPRI
4101		4581
4102	=item EV_MAXPRI	4582	=item EV_MAXPRI
4103		4583
4104	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to	4584	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
…		…
4201		4681
4202	With an intelligent-enough linker (gcc+binutils are intelligent enough	4682	With an intelligent-enough linker (gcc+binutils are intelligent enough
4203	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by	4683	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4204	your program might be left out as well - a binary starting a timer and an	4684	your program might be left out as well - a binary starting a timer and an
4205	I/O watcher then might come out at only 5Kb.	4685	I/O watcher then might come out at only 5Kb.
		4686
		4687	=item EV_API_STATIC
		4688
		4689	If this symbol is defined (by default it is not), then all identifiers
		4690	will have static linkage. This means that libev will not export any
		4691	identifiers, and you cannot link against libev anymore. This can be useful
		4692	when you embed libev, only want to use libev functions in a single file,
		4693	and do not want its identifiers to be visible.
		4694
		4695	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4696	wants to use libev.
		4697
		4698	This option only works when libev is compiled with a C compiler, as C++
		4699	doesn't support the required declaration syntax.
4206		4700
4207	=item EV_AVOID_STDIO	4701	=item EV_AVOID_STDIO
4208		4702
4209	If this is set to C<1> at compiletime, then libev will avoid using stdio	4703	If this is set to C<1> at compiletime, then libev will avoid using stdio
4210	functions (printf, scanf, perror etc.). This will increase the code size	4704	functions (printf, scanf, perror etc.). This will increase the code size
…		…
4354	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4848	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4355		4849
4356	#include "ev_cpp.h"	4850	#include "ev_cpp.h"
4357	#include "ev.c"	4851	#include "ev.c"
4358		4852
4359	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES	4853	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
4360		4854
4361	=head2 THREADS AND COROUTINES	4855	=head2 THREADS AND COROUTINES
4362		4856
4363	=head3 THREADS	4857	=head3 THREADS
4364		4858
…		…
4415	default loop and triggering an C<ev_async> watcher from the default loop	4909	default loop and triggering an C<ev_async> watcher from the default loop
4416	watcher callback into the event loop interested in the signal.	4910	watcher callback into the event loop interested in the signal.
4417		4911
4418	=back	4912	=back
4419		4913
4420	=head4 THREAD LOCKING EXAMPLE	4914	See also L<THREAD LOCKING EXAMPLE>.
4421
4422	Here is a fictitious example of how to run an event loop in a different
4423	thread than where callbacks are being invoked and watchers are
4424	created/added/removed.
4425
4426	For a real-world example, see the C<EV::Loop::Async> perl module,
4427	which uses exactly this technique (which is suited for many high-level
4428	languages).
4429
4430	The example uses a pthread mutex to protect the loop data, a condition
4431	variable to wait for callback invocations, an async watcher to notify the
4432	event loop thread and an unspecified mechanism to wake up the main thread.
4433
4434	First, you need to associate some data with the event loop:
4435
4436	typedef struct {
4437	mutex_t lock; /* global loop lock */
4438	ev_async async_w;
4439	thread_t tid;
4440	cond_t invoke_cv;
4441	} userdata;
4442
4443	void prepare_loop (EV_P)
4444	{
4445	// for simplicity, we use a static userdata struct.
4446	static userdata u;
4447
4448	ev_async_init (&u->async_w, async_cb);
4449	ev_async_start (EV_A_ &u->async_w);
4450
4451	pthread_mutex_init (&u->lock, 0);
4452	pthread_cond_init (&u->invoke_cv, 0);
4453
4454	// now associate this with the loop
4455	ev_set_userdata (EV_A_ u);
4456	ev_set_invoke_pending_cb (EV_A_ l_invoke);
4457	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
4458
4459	// then create the thread running ev_loop
4460	pthread_create (&u->tid, 0, l_run, EV_A);
4461	}
4462
4463	The callback for the C<ev_async> watcher does nothing: the watcher is used
4464	solely to wake up the event loop so it takes notice of any new watchers
4465	that might have been added:
4466
4467	static void
4468	async_cb (EV_P_ ev_async *w, int revents)
4469	{
4470	// just used for the side effects
4471	}
4472
4473	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
4474	protecting the loop data, respectively.
4475
4476	static void
4477	l_release (EV_P)
4478	{
4479	userdata *u = ev_userdata (EV_A);
4480	pthread_mutex_unlock (&u->lock);
4481	}
4482
4483	static void
4484	l_acquire (EV_P)
4485	{
4486	userdata *u = ev_userdata (EV_A);
4487	pthread_mutex_lock (&u->lock);
4488	}
4489
4490	The event loop thread first acquires the mutex, and then jumps straight
4491	into C<ev_run>:
4492
4493	void *
4494	l_run (void *thr_arg)
4495	{
4496	struct ev_loop *loop = (struct ev_loop *)thr_arg;
4497
4498	l_acquire (EV_A);
4499	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
4500	ev_run (EV_A_ 0);
4501	l_release (EV_A);
4502
4503	return 0;
4504	}
4505
4506	Instead of invoking all pending watchers, the C<l_invoke> callback will
4507	signal the main thread via some unspecified mechanism (signals? pipe
4508	writes? C<Async::Interrupt>?) and then waits until all pending watchers
4509	have been called (in a while loop because a) spurious wakeups are possible
4510	and b) skipping inter-thread-communication when there are no pending
4511	watchers is very beneficial):
4512
4513	static void
4514	l_invoke (EV_P)
4515	{
4516	userdata *u = ev_userdata (EV_A);
4517
4518	while (ev_pending_count (EV_A))
4519	{
4520	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
4521	pthread_cond_wait (&u->invoke_cv, &u->lock);
4522	}
4523	}
4524
4525	Now, whenever the main thread gets told to invoke pending watchers, it
4526	will grab the lock, call C<ev_invoke_pending> and then signal the loop
4527	thread to continue:
4528
4529	static void
4530	real_invoke_pending (EV_P)
4531	{
4532	userdata *u = ev_userdata (EV_A);
4533
4534	pthread_mutex_lock (&u->lock);
4535	ev_invoke_pending (EV_A);
4536	pthread_cond_signal (&u->invoke_cv);
4537	pthread_mutex_unlock (&u->lock);
4538	}
4539
4540	Whenever you want to start/stop a watcher or do other modifications to an
4541	event loop, you will now have to lock:
4542
4543	ev_timer timeout_watcher;
4544	userdata *u = ev_userdata (EV_A);
4545
4546	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
4547
4548	pthread_mutex_lock (&u->lock);
4549	ev_timer_start (EV_A_ &timeout_watcher);
4550	ev_async_send (EV_A_ &u->async_w);
4551	pthread_mutex_unlock (&u->lock);
4552
4553	Note that sending the C<ev_async> watcher is required because otherwise
4554	an event loop currently blocking in the kernel will have no knowledge
4555	about the newly added timer. By waking up the loop it will pick up any new
4556	watchers in the next event loop iteration.
4557		4915
4558	=head3 COROUTINES	4916	=head3 COROUTINES
4559		4917
4560	Libev is very accommodating to coroutines ("cooperative threads"):	4918	Libev is very accommodating to coroutines ("cooperative threads"):
4561	libev fully supports nesting calls to its functions from different	4919	libev fully supports nesting calls to its functions from different
…		…
4726	requires, and its I/O model is fundamentally incompatible with the POSIX	5084	requires, and its I/O model is fundamentally incompatible with the POSIX
4727	model. Libev still offers limited functionality on this platform in	5085	model. Libev still offers limited functionality on this platform in
4728	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	5086	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
4729	descriptors. This only applies when using Win32 natively, not when using	5087	descriptors. This only applies when using Win32 natively, not when using
4730	e.g. cygwin. Actually, it only applies to the microsofts own compilers,	5088	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
4731	as every compielr comes with a slightly differently broken/incompatible	5089	as every compiler comes with a slightly differently broken/incompatible
4732	environment.	5090	environment.
4733		5091
4734	Lifting these limitations would basically require the full	5092	Lifting these limitations would basically require the full
4735	re-implementation of the I/O system. If you are into this kind of thing,	5093	re-implementation of the I/O system. If you are into this kind of thing,
4736	then note that glib does exactly that for you in a very portable way (note	5094	then note that glib does exactly that for you in a very portable way (note
…		…
4869		5227
4870	The type C<double> is used to represent timestamps. It is required to	5228	The type C<double> is used to represent timestamps. It is required to
4871	have at least 51 bits of mantissa (and 9 bits of exponent), which is	5229	have at least 51 bits of mantissa (and 9 bits of exponent), which is
4872	good enough for at least into the year 4000 with millisecond accuracy	5230	good enough for at least into the year 4000 with millisecond accuracy
4873	(the design goal for libev). This requirement is overfulfilled by	5231	(the design goal for libev). This requirement is overfulfilled by
4874	implementations using IEEE 754, which is basically all existing ones. With	5232	implementations using IEEE 754, which is basically all existing ones.
		5233
4875	IEEE 754 doubles, you get microsecond accuracy until at least 2200.	5234	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5235	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5236	is either obsolete or somebody patched it to use C<long double> or
		5237	something like that, just kidding).
4876		5238
4877	=back	5239	=back
4878		5240
4879	If you know of other additional requirements drop me a note.	5241	If you know of other additional requirements drop me a note.
4880		5242
…		…
4942	=item Processing ev_async_send: O(number_of_async_watchers)	5304	=item Processing ev_async_send: O(number_of_async_watchers)
4943		5305
4944	=item Processing signals: O(max_signal_number)	5306	=item Processing signals: O(max_signal_number)
4945		5307
4946	Sending involves a system call I<iff> there were no other C<ev_async_send>	5308	Sending involves a system call I<iff> there were no other C<ev_async_send>
4947	calls in the current loop iteration. Checking for async and signal events	5309	calls in the current loop iteration and the loop is currently
		5310	blocked. Checking for async and signal events involves iterating over all
4948	involves iterating over all running async watchers or all signal numbers.	5311	running async watchers or all signal numbers.
4949		5312
4950	=back	5313	=back
4951		5314
4952		5315
4953	=head1 PORTING FROM LIBEV 3.X TO 4.X	5316	=head1 PORTING FROM LIBEV 3.X TO 4.X
…		…
5070	The physical time that is observed. It is apparently strictly monotonic :)	5433	The physical time that is observed. It is apparently strictly monotonic :)
5071		5434
5072	=item wall-clock time	5435	=item wall-clock time
5073		5436
5074	The time and date as shown on clocks. Unlike real time, it can actually	5437	The time and date as shown on clocks. Unlike real time, it can actually
5075	be wrong and jump forwards and backwards, e.g. when the you adjust your	5438	be wrong and jump forwards and backwards, e.g. when you adjust your
5076	clock.	5439	clock.
5077		5440
5078	=item watcher	5441	=item watcher
5079		5442
5080	A data structure that describes interest in certain events. Watchers need	5443	A data structure that describes interest in certain events. Watchers need
…		…
5083	=back	5446	=back
5084		5447
5085	=head1 AUTHOR	5448	=head1 AUTHOR
5086		5449
5087	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael	5450	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
5088	Magnusson and Emanuele Giaquinta.	5451	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
5089		5452

Diff Legend

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