ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.343 by root, Wed Nov 10 13:39:10 2010 UTC vs.
Revision 1.388 by root, Tue Dec 20 04:08:35 2011 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
…		…
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
…		…
781	This is useful if you are waiting for some external event in conjunction	834	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	835	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	836	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.	837	usually a better approach for this kind of thing.
785		838
786	Here are the gory details of what C<ev_run> does:	839	Here are the gory details of what C<ev_run> does (this is for your
		840	understanding, not a guarantee that things will work exactly like this in
		841	future versions):
787		842
788	- Increment loop depth.	843	- Increment loop depth.
789	- Reset the ev_break status.	844	- Reset the ev_break status.
790	- Before the first iteration, call any pending watchers.	845	- Before the first iteration, call any pending watchers.
791	LOOP:	846	LOOP:
…		…
824	anymore.	879	anymore.
825		880
826	... queue jobs here, make sure they register event watchers as long	881	... 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..)	882	... as they still have work to do (even an idle watcher will do..)
828	ev_run (my_loop, 0);	883	ev_run (my_loop, 0);
829	... jobs done or somebody called unloop. yeah!	884	... jobs done or somebody called break. yeah!
830		885
831	=item ev_break (loop, how)	886	=item ev_break (loop, how)
832		887
833	Can be used to make a call to C<ev_run> return early (but only after it	888	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	889	has processed all outstanding events). The C<how> argument must be either
…		…
867	running when nothing else is active.	922	running when nothing else is active.
868		923
869	ev_signal exitsig;	924	ev_signal exitsig;
870	ev_signal_init (&exitsig, sig_cb, SIGINT);	925	ev_signal_init (&exitsig, sig_cb, SIGINT);
871	ev_signal_start (loop, &exitsig);	926	ev_signal_start (loop, &exitsig);
872	evf_unref (loop);	927	ev_unref (loop);
873		928
874	Example: For some weird reason, unregister the above signal handler again.	929	Example: For some weird reason, unregister the above signal handler again.
875		930
876	ev_ref (loop);	931	ev_ref (loop);
877	ev_signal_stop (loop, &exitsig);	932	ev_signal_stop (loop, &exitsig);
…		…
897	overhead for the actual polling but can deliver many events at once.	952	overhead for the actual polling but can deliver many events at once.
898		953
899	By setting a higher I<io collect interval> you allow libev to spend more	954	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,	955	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	956	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	957	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	958	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	959	sleep time ensures that libev will not poll for I/O events more often then
905	once per this interval, on average.	960	once per this interval, on average (as long as the host time resolution is
		961	good enough).
906		962
907	Likewise, by setting a higher I<timeout collect interval> you allow libev	963	Likewise, by setting a higher I<timeout collect interval> you allow libev
908	to spend more time collecting timeouts, at the expense of increased	964	to spend more time collecting timeouts, at the expense of increased
909	latency/jitter/inexactness (the watcher callback will be called	965	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	966	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	1020	can be done relatively simply by putting mutex_lock/unlock calls around
965	each call to a libev function.	1021	each call to a libev function.
966		1022
967	However, C<ev_run> can run an indefinite time, so it is not feasible	1023	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	1024	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	1025	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.	1026	I<release> and I<acquire> callbacks on the loop.
971		1027
972	When set, then C<release> will be called just before the thread is	1028	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	1029	suspended waiting for new events, and C<acquire> is called just
974	afterwards.	1030	afterwards.
…		…
989	See also the locking example in the C<THREADS> section later in this	1045	See also the locking example in the C<THREADS> section later in this
990	document.	1046	document.
991		1047
992	=item ev_set_userdata (loop, void *data)	1048	=item ev_set_userdata (loop, void *data)
993		1049
994	=item ev_userdata (loop)	1050	=item void *ev_userdata (loop)
995		1051
996	Set and retrieve a single C<void *> associated with a loop. When	1052	Set and retrieve a single C<void *> associated with a loop. When
997	C<ev_set_userdata> has never been called, then C<ev_userdata> returns	1053	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
998	C<0>.	1054	C<0>.
999		1055
…		…
1316	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related	1372	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
1317	functions that do not need a watcher.	1373	functions that do not need a watcher.
1318		1374
1319	=back	1375	=back
1320		1376
1321	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1377	See also the L<ASSOCIATING CUSTOM DATA WITH A WATCHER> and L<BUILDING YOUR
1322		1378	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		1379
1386	=head2 WATCHER STATES	1380	=head2 WATCHER STATES
1387		1381
1388	There are various watcher states mentioned throughout this manual -	1382	There are various watcher states mentioned throughout this manual -
1389	active, pending and so on. In this section these states and the rules to	1383	active, pending and so on. In this section these states and the rules to
…		…
1392		1386
1393	=over 4	1387	=over 4
1394		1388
1395	=item initialiased	1389	=item initialiased
1396		1390
1397	Before a watcher can be registered with the event looop it has to be	1391	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	1392	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.	1393	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
1400		1394
1401	In this state it is simply some block of memory that is suitable for use	1395	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.	1396	use in an event loop. It can be moved around, freed, reused etc. at
		1397	will - as long as you either keep the memory contents intact, or call
		1398	C<ev_TYPE_init> again.
1403		1399
1404	=item started/running/active	1400	=item started/running/active
1405		1401
1406	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes	1402	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	1403	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	1431	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	1432	of whether it was active or not, so stopping a watcher explicitly before
1437	freeing it is often a good idea.	1433	freeing it is often a good idea.
1438		1434
1439	While stopped (and not pending) the watcher is essentially in the	1435	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	1436	initialised state, that is, it can be reused, moved, modified in any way
1441	you wish.	1437	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1438	it again).
1442		1439
1443	=back	1440	=back
1444		1441
1445	=head2 WATCHER PRIORITY MODELS	1442	=head2 WATCHER PRIORITY MODELS
1446		1443
…		…
1575	In general you can register as many read and/or write event watchers per	1572	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	1573	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	1574	descriptors to non-blocking mode is also usually a good idea (but not
1578	required if you know what you are doing).	1575	required if you know what you are doing).
1579		1576
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	1577	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	1578	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	1579	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	1580	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	1581	with a relatively standard program structure. Thus it is best to always
1591	this situation even with a relatively standard program structure. Thus	1582	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.	1583	preferable to a program hanging until some data arrives.
1594		1584
1595	If you cannot run the fd in non-blocking mode (for example you should	1585	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	1586	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	1587	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	1588	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	1589	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	1590	use C<SIGALRM> and an interval timer, just to be sure you won't block
1601	indefinitely.	1591	indefinitely.
1602		1592
1603	But really, best use non-blocking mode.	1593	But really, best use non-blocking mode.
1604		1594
…		…
1632		1622
1633	There is no workaround possible except not registering events	1623	There is no workaround possible except not registering events
1634	for potentially C<dup ()>'ed file descriptors, or to resort to	1624	for potentially C<dup ()>'ed file descriptors, or to resort to
1635	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1625	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1636		1626
		1627	=head3 The special problem of files
		1628
		1629	Many people try to use C<select> (or libev) on file descriptors
		1630	representing files, and expect it to become ready when their program
		1631	doesn't block on disk accesses (which can take a long time on their own).
		1632
		1633	However, this cannot ever work in the "expected" way - you get a readiness
		1634	notification as soon as the kernel knows whether and how much data is
		1635	there, and in the case of open files, that's always the case, so you
		1636	always get a readiness notification instantly, and your read (or possibly
		1637	write) will still block on the disk I/O.
		1638
		1639	Another way to view it is that in the case of sockets, pipes, character
		1640	devices and so on, there is another party (the sender) that delivers data
		1641	on its own, but in the case of files, there is no such thing: the disk
		1642	will not send data on its own, simply because it doesn't know what you
		1643	wish to read - you would first have to request some data.
		1644
		1645	Since files are typically not-so-well supported by advanced notification
		1646	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1647	to files, even though you should not use it. The reason for this is
		1648	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1649	usually a tty, often a pipe, but also sometimes files or special devices
		1650	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1651	F</dev/urandom>), and even though the file might better be served with
		1652	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1653	it "just works" instead of freezing.
		1654
		1655	So avoid file descriptors pointing to files when you know it (e.g. use
		1656	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1657	when you rarely read from a file instead of from a socket, and want to
		1658	reuse the same code path.
		1659
1637	=head3 The special problem of fork	1660	=head3 The special problem of fork
1638		1661
1639	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1662	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	1663	useless behaviour. Libev fully supports fork, but needs to be told about
1641	it in the child.	1664	it in the child if you want to continue to use it in the child.
1642		1665
1643	To support fork in your programs, you either have to call	1666	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,	1667	()> 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	1668	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1646	C<EVBACKEND_POLL>.
1647		1669
1648	=head3 The special problem of SIGPIPE	1670	=head3 The special problem of SIGPIPE
1649		1671
1650	While not really specific to libev, it is easy to forget about C<SIGPIPE>:	1672	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	1673	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	1771	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1750	monotonic clock option helps a lot here).	1772	monotonic clock option helps a lot here).
1751		1773
1752	The callback is guaranteed to be invoked only I<after> its timeout has	1774	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	1775	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	1776	might introduce a small delay, see "the special problem of being too
		1777	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	1778	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	1779	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).	1780	longer true when a callback calls C<ev_run> recursively).
1758		1781
1759	=head3 Be smart about timeouts	1782	=head3 Be smart about timeouts
1760		1783
1761	Many real-world problems involve some kind of timeout, usually for error	1784	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,	1785	recovery. A typical example is an HTTP request - if the other side hangs,
…		…
1837		1860
1838	In this case, it would be more efficient to leave the C<ev_timer> alone,	1861	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	1862	but remember the time of last activity, and check for a real timeout only
1840	within the callback:	1863	within the callback:
1841		1864
		1865	ev_tstamp timeout = 60.;
1842	ev_tstamp last_activity; // time of last activity	1866	ev_tstamp last_activity; // time of last activity
		1867	ev_timer timer;
1843		1868
1844	static void	1869	static void
1845	callback (EV_P_ ev_timer *w, int revents)	1870	callback (EV_P_ ev_timer *w, int revents)
1846	{	1871	{
1847	ev_tstamp now = ev_now (EV_A);	1872	// calculate when the timeout would happen
1848	ev_tstamp timeout = last_activity + 60.;	1873	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
1849		1874
1850	// if last_activity + 60. is older than now, we did time out	1875	// if negative, it means we the timeout already occured
1851	if (timeout < now)	1876	if (after < 0.)
1852	{	1877	{
1853	// timeout occurred, take action	1878	// timeout occurred, take action
1854	}	1879	}
1855	else	1880	else
1856	{	1881	{
1857	// callback was invoked, but there was some activity, re-arm	1882	// callback was invoked, but there was some recent
1858	// the watcher to fire in last_activity + 60, which is	1883	// activity. simply restart the timer to time out
1859	// guaranteed to be in the future, so "again" is positive:	1884	// after "after" seconds, which is the earliest time
1860	w->repeat = timeout - now;	1885	// the timeout can occur.
		1886	ev_timer_set (w, after, 0.);
1861	ev_timer_again (EV_A_ w);	1887	ev_timer_start (EV_A_ w);
1862	}	1888	}
1863	}	1889	}
1864		1890
1865	To summarise the callback: first calculate the real timeout (defined	1891	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	1892	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	1893	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	1894	(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		1895
1872	Note how C<ev_timer_again> is used, taking advantage of the	1896	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.	1897	timed out, and need to do whatever is needed in this case.
		1898
		1899	Otherwise, we now the earliest time at which the timeout would trigger,
		1900	and simply start the timer with this timeout value.
		1901
		1902	In other words, each time the callback is invoked it will check whether
		1903	the timeout cocured. If not, it will simply reschedule itself to check
		1904	again at the earliest time it could time out. Rinse. Repeat.
1874		1905
1875	This scheme causes more callback invocations (about one every 60 seconds	1906	This scheme causes more callback invocations (about one every 60 seconds
1876	minus half the average time between activity), but virtually no calls to	1907	minus half the average time between activity), but virtually no calls to
1877	libev to change the timeout.	1908	libev to change the timeout.
1878		1909
1879	To start the timer, simply initialise the watcher and set C<last_activity>	1910	To start the machinery, simply initialise the watcher and set
1880	to the current time (meaning we just have some activity :), then call the	1911	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:	1912	now), then call the callback, which will "do the right thing" and start
		1913	the timer:
1882		1914
		1915	last_activity = ev_now (EV_A);
1883	ev_init (timer, callback);	1916	ev_init (&timer, callback);
1884	last_activity = ev_now (loop);	1917	callback (EV_A_ &timer, 0);
1885	callback (loop, timer, EV_TIMER);
1886		1918
1887	And when there is some activity, simply store the current time in	1919	When there is some activity, simply store the current time in
1888	C<last_activity>, no libev calls at all:	1920	C<last_activity>, no libev calls at all:
1889		1921
		1922	if (activity detected)
1890	last_activity = ev_now (loop);	1923	last_activity = ev_now (EV_A);
		1924
		1925	When your timeout value changes, then the timeout can be changed by simply
		1926	providing a new value, stopping the timer and calling the callback, which
		1927	will agaion do the right thing (for example, time out immediately :).
		1928
		1929	timeout = new_value;
		1930	ev_timer_stop (EV_A_ &timer);
		1931	callback (EV_A_ &timer, 0);
1891		1932
1892	This technique is slightly more complex, but in most cases where the	1933	This technique is slightly more complex, but in most cases where the
1893	time-out is unlikely to be triggered, much more efficient.	1934	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		1935
1899	=item 4. Wee, just use a double-linked list for your timeouts.	1936	=item 4. Wee, just use a double-linked list for your timeouts.
1900		1937
1901	If there is not one request, but many thousands (millions...), all	1938	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	1939	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	1966	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	1967	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	1968	off after the first million or so of active timers, i.e. it's usually
1932	overkill :)	1969	overkill :)
1933		1970
		1971	=head3 The special problem of being too early
		1972
		1973	If you ask a timer to call your callback after three seconds, then
		1974	you expect it to be invoked after three seconds - but of course, this
		1975	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		1976	guaranteed to any precision by libev - imagine somebody suspending the
		1977	process with a STOP signal for a few hours for example.
		1978
		1979	So, libev tries to invoke your callback as soon as possible I<after> the
		1980	delay has occurred, but cannot guarantee this.
		1981
		1982	A less obvious failure mode is calling your callback too early: many event
		1983	loops compare timestamps with a "elapsed delay >= requested delay", but
		1984	this can cause your callback to be invoked much earlier than you would
		1985	expect.
		1986
		1987	To see why, imagine a system with a clock that only offers full second
		1988	resolution (think windows if you can't come up with a broken enough OS
		1989	yourself). If you schedule a one-second timer at the time 500.9, then the
		1990	event loop will schedule your timeout to elapse at a system time of 500
		1991	(500.9 truncated to the resolution) + 1, or 501.
		1992
		1993	If an event library looks at the timeout 0.1s later, it will see "501 >=
		1994	501" and invoke the callback 0.1s after it was started, even though a
		1995	one-second delay was requested - this is being "too early", despite best
		1996	intentions.
		1997
		1998	This is the reason why libev will never invoke the callback if the elapsed
		1999	delay equals the requested delay, but only when the elapsed delay is
		2000	larger than the requested delay. In the example above, libev would only invoke
		2001	the callback at system time 502, or 1.1s after the timer was started.
		2002
		2003	So, while libev cannot guarantee that your callback will be invoked
		2004	exactly when requested, it I<can> and I<does> guarantee that the requested
		2005	delay has actually elapsed, or in other words, it always errs on the "too
		2006	late" side of things.
		2007
1934	=head3 The special problem of time updates	2008	=head3 The special problem of time updates
1935		2009
1936	Establishing the current time is a costly operation (it usually takes at	2010	Establishing the current time is a costly operation (it usually takes
1937	least two system calls): EV therefore updates its idea of the current	2011	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	2012	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	2013	growing difference between C<ev_now ()> and C<ev_time ()> when handling
1940	lots of events in one iteration.	2014	lots of events in one iteration.
1941		2015
1942	The relative timeouts are calculated relative to the C<ev_now ()>	2016	The relative timeouts are calculated relative to the C<ev_now ()>
…		…
1948	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2022	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1949		2023
1950	If the event loop is suspended for a long time, you can also force an	2024	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	2025	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1952	()>.	2026	()>.
		2027
		2028	=head3 The special problem of unsynchronised clocks
		2029
		2030	Modern systems have a variety of clocks - libev itself uses the normal
		2031	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2032	jumps).
		2033
		2034	Neither of these clocks is synchronised with each other or any other clock
		2035	on the system, so C<ev_time ()> might return a considerably different time
		2036	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2037	a call to C<gettimeofday> might return a second count that is one higher
		2038	than a directly following call to C<time>.
		2039
		2040	The moral of this is to only compare libev-related timestamps with
		2041	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2042	a second or so.
		2043
		2044	One more problem arises due to this lack of synchronisation: if libev uses
		2045	the system monotonic clock and you compare timestamps from C<ev_time>
		2046	or C<ev_now> from when you started your timer and when your callback is
		2047	invoked, you will find that sometimes the callback is a bit "early".
		2048
		2049	This is because C<ev_timer>s work in real time, not wall clock time, so
		2050	libev makes sure your callback is not invoked before the delay happened,
		2051	I<measured according to the real time>, not the system clock.
		2052
		2053	If your timeouts are based on a physical timescale (e.g. "time out this
		2054	connection after 100 seconds") then this shouldn't bother you as it is
		2055	exactly the right behaviour.
		2056
		2057	If you want to compare wall clock/system timestamps to your timers, then
		2058	you need to use C<ev_periodic>s, as these are based on the wall clock
		2059	time, where your comparisons will always generate correct results.
1953		2060
1954	=head3 The special problems of suspended animation	2061	=head3 The special problems of suspended animation
1955		2062
1956	When you leave the server world it is quite customary to hit machines that	2063	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?	2064	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	2108	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.	2109	do stuff) the timer will not fire more than once per event loop iteration.
2003		2110
2004	=item ev_timer_again (loop, ev_timer *)	2111	=item ev_timer_again (loop, ev_timer *)
2005		2112
2006	This will act as if the timer timed out and restart it again if it is	2113	This will act as if the timer timed out and restarts it again if it is
2007	repeating. The exact semantics are:	2114	repeating. The exact semantics are:
2008		2115
2009	If the timer is pending, its pending status is cleared.	2116	If the timer is pending, its pending status is cleared.
2010		2117
2011	If the timer is started but non-repeating, stop it (as if it timed out).	2118	If the timer is started but non-repeating, stop it (as if it timed out).
…		…
2141		2248
2142	Another way to think about it (for the mathematically inclined) is that	2249	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	2250	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.	2251	time where C<time = offset (mod interval)>, regardless of any time jumps.
2145		2252
2146	For numerical stability it is preferable that the C<offset> value is near	2253	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	2254	interval value should be higher than C<1/8192> (which is around 100
2148	this value, and in fact is often specified as zero.	2255	microseconds) and C<offset> should be higher than C<0> and should have
		2256	at most a similar magnitude as the current time (say, within a factor of
		2257	ten). Typical values for offset are, in fact, C<0> or something between
		2258	C<0> and C<interval>, which is also the recommended range.
2149		2259
2150	Note also that there is an upper limit to how often a timer can fire (CPU	2260	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	2261	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	2262	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).	2263	millisecond (if the OS supports it and the machine is fast enough).
…		…
2296	=head3 The special problem of inheritance over fork/execve/pthread_create	2406	=head3 The special problem of inheritance over fork/execve/pthread_create
2297		2407
2298	Both the signal mask (C<sigprocmask>) and the signal disposition	2408	Both the signal mask (C<sigprocmask>) and the signal disposition
2299	(C<sigaction>) are unspecified after starting a signal watcher (and after	2409	(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,	2410	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.	2411	and might or might not set or restore the installed signal handler (but
		2412	see C<EVFLAG_NOSIGMASK>).
2302		2413
2303	While this does not matter for the signal disposition (libev never	2414	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	2415	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	2416	C<execve>), this matters for the signal mask: many programs do not expect
2306	certain signals to be blocked.	2417	certain signals to be blocked.
…		…
2319	I<has> to modify the signal mask, at least temporarily.	2430	I<has> to modify the signal mask, at least temporarily.
2320		2431
2321	So I can't stress this enough: I<If you do not reset your signal mask when	2432	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	2433	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.	2434	is not a libev-specific thing, this is true for most event libraries.
		2435
		2436	=head3 The special problem of threads signal handling
		2437
		2438	POSIX threads has problematic signal handling semantics, specifically,
		2439	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2440	threads in a process block signals, which is hard to achieve.
		2441
		2442	When you want to use sigwait (or mix libev signal handling with your own
		2443	for the same signals), you can tackle this problem by globally blocking
		2444	all signals before creating any threads (or creating them with a fully set
		2445	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2446	loops. Then designate one thread as "signal receiver thread" which handles
		2447	these signals. You can pass on any signals that libev might be interested
		2448	in by calling C<ev_feed_signal>.
2324		2449
2325	=head3 Watcher-Specific Functions and Data Members	2450	=head3 Watcher-Specific Functions and Data Members
2326		2451
2327	=over 4	2452	=over 4
2328		2453
…		…
3163	atexit (program_exits);	3288	atexit (program_exits);
3164		3289
3165		3290
3166	=head2 C<ev_async> - how to wake up an event loop	3291	=head2 C<ev_async> - how to wake up an event loop
3167		3292
3168	In general, you cannot use an C<ev_run> from multiple threads or other	3293	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	3294	asynchronous sources such as signal handlers (as opposed to multiple event
3170	loops - those are of course safe to use in different threads).	3295	loops - those are of course safe to use in different threads).
3171		3296
3172	Sometimes, however, you need to wake up an event loop you do not control,	3297	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>	3298	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.	3300	it by calling C<ev_async_send>, which is thread- and signal safe.
3176		3301
3177	This functionality is very similar to C<ev_signal> watchers, as signals,	3302	This functionality is very similar to C<ev_signal> watchers, as signals,
3178	too, are asynchronous in nature, and signals, too, will be compressed	3303	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	3304	(i.e. the number of callback invocations may be less than the number of
3180	C<ev_async_sent> calls).	3305	C<ev_async_sent> calls). In fact, you could use signal watchers as a kind
3181		3306	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	3307	signal, and C<ev_feed_signal> to signal this watcher from another thread,
3183	just the default loop.	3308	even without knowing which loop owns the signal.
3184		3309
3185	=head3 Queueing	3310	=head3 Queueing
3186		3311
3187	C<ev_async> does not support queueing of data in any way. The reason	3312	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	3313	is that the author does not know of a simple (or any) algorithm for a
…		…
3280	trust me.	3405	trust me.
3281		3406
3282	=item ev_async_send (loop, ev_async *)	3407	=item ev_async_send (loop, ev_async *)
3283		3408
3284	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3409	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	3410	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3411	returns.
		3412
3286	C<ev_feed_event>, this call is safe to do from other threads, signal or	3413	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	3414	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
3288	section below on what exactly this means).	3415	embedding section below on what exactly this means).
3289		3416
3290	Note that, as with other watchers in libev, multiple events might get	3417	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	3418	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>,	3419	this is that C<ev_async> watchers are level-triggered: they are set on
3293	reset when the event loop detects that).	3420	C<ev_async_send>, reset when the event loop detects that).
3294		3421
3295	This call incurs the overhead of a system call only once per event loop	3422	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	3423	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.	3424	the event loop (or your program) is processing events. That means that
		3425	repeated calls are basically free (there is no need to avoid calls for
		3426	performance reasons) and that the overhead becomes smaller (typically
		3427	zero) under load.
3298		3428
3299	=item bool = ev_async_pending (ev_async *)	3429	=item bool = ev_async_pending (ev_async *)
3300		3430
3301	Returns a non-zero value when C<ev_async_send> has been called on the	3431	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	3432	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);	3487	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
3358		3488
3359	=item ev_feed_fd_event (loop, int fd, int revents)	3489	=item ev_feed_fd_event (loop, int fd, int revents)
3360		3490
3361	Feed an event on the given fd, as if a file descriptor backend detected	3491	Feed an event on the given fd, as if a file descriptor backend detected
3362	the given events it.	3492	the given events.
3363		3493
3364	=item ev_feed_signal_event (loop, int signum)	3494	=item ev_feed_signal_event (loop, int signum)
3365		3495
3366	Feed an event as if the given signal occurred (C<loop> must be the default	3496	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
3367	loop!).	3497	which is async-safe.
3368		3498
3369	=back	3499	=back
		3500
		3501
		3502	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3503
		3504	This section explains some common idioms that are not immediately
		3505	obvious. Note that examples are sprinkled over the whole manual, and this
		3506	section only contains stuff that wouldn't fit anywhere else.
		3507
		3508	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3509
		3510	Each watcher has, by default, a C<void *data> member that you can read
		3511	or modify at any time: libev will completely ignore it. This can be used
		3512	to associate arbitrary data with your watcher. If you need more data and
		3513	don't want to allocate memory separately and store a pointer to it in that
		3514	data member, you can also "subclass" the watcher type and provide your own
		3515	data:
		3516
		3517	struct my_io
		3518	{
		3519	ev_io io;
		3520	int otherfd;
		3521	void *somedata;
		3522	struct whatever *mostinteresting;
		3523	};
		3524
		3525	...
		3526	struct my_io w;
		3527	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3528
		3529	And since your callback will be called with a pointer to the watcher, you
		3530	can cast it back to your own type:
		3531
		3532	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3533	{
		3534	struct my_io *w = (struct my_io *)w_;
		3535	...
		3536	}
		3537
		3538	More interesting and less C-conformant ways of casting your callback
		3539	function type instead have been omitted.
		3540
		3541	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3542
		3543	Another common scenario is to use some data structure with multiple
		3544	embedded watchers, in effect creating your own watcher that combines
		3545	multiple libev event sources into one "super-watcher":
		3546
		3547	struct my_biggy
		3548	{
		3549	int some_data;
		3550	ev_timer t1;
		3551	ev_timer t2;
		3552	}
		3553
		3554	In this case getting the pointer to C<my_biggy> is a bit more
		3555	complicated: Either you store the address of your C<my_biggy> struct in
		3556	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3557	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3558	real programmers):
		3559
		3560	#include <stddef.h>
		3561
		3562	static void
		3563	t1_cb (EV_P_ ev_timer *w, int revents)
		3564	{
		3565	struct my_biggy big = (struct my_biggy *)
		3566	(((char *)w) - offsetof (struct my_biggy, t1));
		3567	}
		3568
		3569	static void
		3570	t2_cb (EV_P_ ev_timer *w, int revents)
		3571	{
		3572	struct my_biggy big = (struct my_biggy *)
		3573	(((char *)w) - offsetof (struct my_biggy, t2));
		3574	}
		3575
		3576	=head2 AVOIDING FINISHING BEFORE RETURNING
		3577
		3578	Often you have structures like this in event-based programs:
		3579
		3580	callback ()
		3581	{
		3582	free (request);
		3583	}
		3584
		3585	request = start_new_request (..., callback);
		3586
		3587	The intent is to start some "lengthy" operation. The C<request> could be
		3588	used to cancel the operation, or do other things with it.
		3589
		3590	It's not uncommon to have code paths in C<start_new_request> that
		3591	immediately invoke the callback, for example, to report errors. Or you add
		3592	some caching layer that finds that it can skip the lengthy aspects of the
		3593	operation and simply invoke the callback with the result.
		3594
		3595	The problem here is that this will happen I<before> C<start_new_request>
		3596	has returned, so C<request> is not set.
		3597
		3598	Even if you pass the request by some safer means to the callback, you
		3599	might want to do something to the request after starting it, such as
		3600	canceling it, which probably isn't working so well when the callback has
		3601	already been invoked.
		3602
		3603	A common way around all these issues is to make sure that
		3604	C<start_new_request> I<always> returns before the callback is invoked. If
		3605	C<start_new_request> immediately knows the result, it can artificially
		3606	delay invoking the callback by e.g. using a C<prepare> or C<idle> watcher
		3607	for example, or more sneakily, by reusing an existing (stopped) watcher
		3608	and pushing it into the pending queue:
		3609
		3610	ev_set_cb (watcher, callback);
		3611	ev_feed_event (EV_A_ watcher, 0);
		3612
		3613	This way, C<start_new_request> can safely return before the callback is
		3614	invoked, while not delaying callback invocation too much.
		3615
		3616	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3617
		3618	Often (especially in GUI toolkits) there are places where you have
		3619	I<modal> interaction, which is most easily implemented by recursively
		3620	invoking C<ev_run>.
		3621
		3622	This brings the problem of exiting - a callback might want to finish the
		3623	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3624	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3625	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3626	other combination: In these cases, C<ev_break> will not work alone.
		3627
		3628	The solution is to maintain "break this loop" variable for each C<ev_run>
		3629	invocation, and use a loop around C<ev_run> until the condition is
		3630	triggered, using C<EVRUN_ONCE>:
		3631
		3632	// main loop
		3633	int exit_main_loop = 0;
		3634
		3635	while (!exit_main_loop)
		3636	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3637
		3638	// in a model watcher
		3639	int exit_nested_loop = 0;
		3640
		3641	while (!exit_nested_loop)
		3642	ev_run (EV_A_ EVRUN_ONCE);
		3643
		3644	To exit from any of these loops, just set the corresponding exit variable:
		3645
		3646	// exit modal loop
		3647	exit_nested_loop = 1;
		3648
		3649	// exit main program, after modal loop is finished
		3650	exit_main_loop = 1;
		3651
		3652	// exit both
		3653	exit_main_loop = exit_nested_loop = 1;
		3654
		3655	=head2 THREAD LOCKING EXAMPLE
		3656
		3657	Here is a fictitious example of how to run an event loop in a different
		3658	thread from where callbacks are being invoked and watchers are
		3659	created/added/removed.
		3660
		3661	For a real-world example, see the C<EV::Loop::Async> perl module,
		3662	which uses exactly this technique (which is suited for many high-level
		3663	languages).
		3664
		3665	The example uses a pthread mutex to protect the loop data, a condition
		3666	variable to wait for callback invocations, an async watcher to notify the
		3667	event loop thread and an unspecified mechanism to wake up the main thread.
		3668
		3669	First, you need to associate some data with the event loop:
		3670
		3671	typedef struct {
		3672	mutex_t lock; /* global loop lock */
		3673	ev_async async_w;
		3674	thread_t tid;
		3675	cond_t invoke_cv;
		3676	} userdata;
		3677
		3678	void prepare_loop (EV_P)
		3679	{
		3680	// for simplicity, we use a static userdata struct.
		3681	static userdata u;
		3682
		3683	ev_async_init (&u->async_w, async_cb);
		3684	ev_async_start (EV_A_ &u->async_w);
		3685
		3686	pthread_mutex_init (&u->lock, 0);
		3687	pthread_cond_init (&u->invoke_cv, 0);
		3688
		3689	// now associate this with the loop
		3690	ev_set_userdata (EV_A_ u);
		3691	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3692	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3693
		3694	// then create the thread running ev_run
		3695	pthread_create (&u->tid, 0, l_run, EV_A);
		3696	}
		3697
		3698	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3699	solely to wake up the event loop so it takes notice of any new watchers
		3700	that might have been added:
		3701
		3702	static void
		3703	async_cb (EV_P_ ev_async *w, int revents)
		3704	{
		3705	// just used for the side effects
		3706	}
		3707
		3708	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3709	protecting the loop data, respectively.
		3710
		3711	static void
		3712	l_release (EV_P)
		3713	{
		3714	userdata *u = ev_userdata (EV_A);
		3715	pthread_mutex_unlock (&u->lock);
		3716	}
		3717
		3718	static void
		3719	l_acquire (EV_P)
		3720	{
		3721	userdata *u = ev_userdata (EV_A);
		3722	pthread_mutex_lock (&u->lock);
		3723	}
		3724
		3725	The event loop thread first acquires the mutex, and then jumps straight
		3726	into C<ev_run>:
		3727
		3728	void *
		3729	l_run (void *thr_arg)
		3730	{
		3731	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3732
		3733	l_acquire (EV_A);
		3734	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3735	ev_run (EV_A_ 0);
		3736	l_release (EV_A);
		3737
		3738	return 0;
		3739	}
		3740
		3741	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3742	signal the main thread via some unspecified mechanism (signals? pipe
		3743	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3744	have been called (in a while loop because a) spurious wakeups are possible
		3745	and b) skipping inter-thread-communication when there are no pending
		3746	watchers is very beneficial):
		3747
		3748	static void
		3749	l_invoke (EV_P)
		3750	{
		3751	userdata *u = ev_userdata (EV_A);
		3752
		3753	while (ev_pending_count (EV_A))
		3754	{
		3755	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3756	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3757	}
		3758	}
		3759
		3760	Now, whenever the main thread gets told to invoke pending watchers, it
		3761	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3762	thread to continue:
		3763
		3764	static void
		3765	real_invoke_pending (EV_P)
		3766	{
		3767	userdata *u = ev_userdata (EV_A);
		3768
		3769	pthread_mutex_lock (&u->lock);
		3770	ev_invoke_pending (EV_A);
		3771	pthread_cond_signal (&u->invoke_cv);
		3772	pthread_mutex_unlock (&u->lock);
		3773	}
		3774
		3775	Whenever you want to start/stop a watcher or do other modifications to an
		3776	event loop, you will now have to lock:
		3777
		3778	ev_timer timeout_watcher;
		3779	userdata *u = ev_userdata (EV_A);
		3780
		3781	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3782
		3783	pthread_mutex_lock (&u->lock);
		3784	ev_timer_start (EV_A_ &timeout_watcher);
		3785	ev_async_send (EV_A_ &u->async_w);
		3786	pthread_mutex_unlock (&u->lock);
		3787
		3788	Note that sending the C<ev_async> watcher is required because otherwise
		3789	an event loop currently blocking in the kernel will have no knowledge
		3790	about the newly added timer. By waking up the loop it will pick up any new
		3791	watchers in the next event loop iteration.
		3792
		3793	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3794
		3795	While the overhead of a callback that e.g. schedules a thread is small, it
		3796	is still an overhead. If you embed libev, and your main usage is with some
		3797	kind of threads or coroutines, you might want to customise libev so that
		3798	doesn't need callbacks anymore.
		3799
		3800	Imagine you have coroutines that you can switch to using a function
		3801	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3802	and that due to some magic, the currently active coroutine is stored in a
		3803	global called C<current_coro>. Then you can build your own "wait for libev
		3804	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3805	the differing C<;> conventions):
		3806
		3807	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3808	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3809
		3810	That means instead of having a C callback function, you store the
		3811	coroutine to switch to in each watcher, and instead of having libev call
		3812	your callback, you instead have it switch to that coroutine.
		3813
		3814	A coroutine might now wait for an event with a function called
		3815	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3816	matter when, or whether the watcher is active or not when this function is
		3817	called):
		3818
		3819	void
		3820	wait_for_event (ev_watcher *w)
		3821	{
		3822	ev_cb_set (w) = current_coro;
		3823	switch_to (libev_coro);
		3824	}
		3825
		3826	That basically suspends the coroutine inside C<wait_for_event> and
		3827	continues the libev coroutine, which, when appropriate, switches back to
		3828	this or any other coroutine. I am sure if you sue this your own :)
		3829
		3830	You can do similar tricks if you have, say, threads with an event queue -
		3831	instead of storing a coroutine, you store the queue object and instead of
		3832	switching to a coroutine, you push the watcher onto the queue and notify
		3833	any waiters.
		3834
		3835	To embed libev, see L<EMBEDDING>, but in short, it's easiest to create two
		3836	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3837
		3838	// my_ev.h
		3839	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3840	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
		3841	#include "../libev/ev.h"
		3842
		3843	// my_ev.c
		3844	#define EV_H "my_ev.h"
		3845	#include "../libev/ev.c"
		3846
		3847	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3848	F<my_ev.c> into your project. When properly specifying include paths, you
		3849	can even use F<ev.h> as header file name directly.
3370		3850
3371		3851
3372	=head1 LIBEVENT EMULATION	3852	=head1 LIBEVENT EMULATION
3373		3853
3374	Libev offers a compatibility emulation layer for libevent. It cannot	3854	Libev offers a compatibility emulation layer for libevent. It cannot
3375	emulate the internals of libevent, so here are some usage hints:	3855	emulate the internals of libevent, so here are some usage hints:
3376		3856
3377	=over 4	3857	=over 4
		3858
		3859	=item * Only the libevent-1.4.1-beta API is being emulated.
		3860
		3861	This was the newest libevent version available when libev was implemented,
		3862	and is still mostly unchanged in 2010.
3378		3863
3379	=item * Use it by including <event.h>, as usual.	3864	=item * Use it by including <event.h>, as usual.
3380		3865
3381	=item * The following members are fully supported: ev_base, ev_callback,	3866	=item * The following members are fully supported: ev_base, ev_callback,
3382	ev_arg, ev_fd, ev_res, ev_events.	3867	ev_arg, ev_fd, ev_res, ev_events.
…		…
3417	Care has been taken to keep the overhead low. The only data member the C++	3902	Care has been taken to keep the overhead low. The only data member the C++
3418	classes add (compared to plain C-style watchers) is the event loop pointer	3903	classes add (compared to plain C-style watchers) is the event loop pointer
3419	that the watcher is associated with (or no additional members at all if	3904	that the watcher is associated with (or no additional members at all if
3420	you disable C<EV_MULTIPLICITY> when embedding libev).	3905	you disable C<EV_MULTIPLICITY> when embedding libev).
3421		3906
3422	Currently, functions, and static and non-static member functions can be	3907	Currently, functions, static and non-static member functions and classes
3423	used as callbacks. Other types should be easy to add as long as they only	3908	with C<operator ()> can be used as callbacks. Other types should be easy
3424	need one additional pointer for context. If you need support for other	3909	to add as long as they only need one additional pointer for context. If
3425	types of functors please contact the author (preferably after implementing	3910	you need support for other types of functors please contact the author
3426	it).	3911	(preferably after implementing it).
3427		3912
3428	Here is a list of things available in the C<ev> namespace:	3913	Here is a list of things available in the C<ev> namespace:
3429		3914
3430	=over 4	3915	=over 4
3431		3916
…		…
3584	watchers in the constructor.	4069	watchers in the constructor.
3585		4070
3586	class myclass	4071	class myclass
3587	{	4072	{
3588	ev::io io ; void io_cb (ev::io &w, int revents);	4073	ev::io io ; void io_cb (ev::io &w, int revents);
3589	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);	4074	ev::io io2 ; void io2_cb (ev::io &w, int revents);
3590	ev::idle idle; void idle_cb (ev::idle &w, int revents);	4075	ev::idle idle; void idle_cb (ev::idle &w, int revents);
3591		4076
3592	myclass (int fd)	4077	myclass (int fd)
3593	{	4078	{
3594	io .set <myclass, &myclass::io_cb > (this);	4079	io .set <myclass, &myclass::io_cb > (this);
…		…
3645	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.	4130	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
3646		4131
3647	=item D	4132	=item D
3648		4133
3649	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	4134	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
3650	be found at L<http://proj.llucax.com.ar/wiki/evd>.	4135	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
3651		4136
3652	=item Ocaml	4137	=item Ocaml
3653		4138
3654	Erkki Seppala has written Ocaml bindings for libev, to be found at	4139	Erkki Seppala has written Ocaml bindings for libev, to be found at
3655	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.	4140	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
…		…
3703	suitable for use with C<EV_A>.	4188	suitable for use with C<EV_A>.
3704		4189
3705	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	4190	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
3706		4191
3707	Similar to the other two macros, this gives you the value of the default	4192	Similar to the other two macros, this gives you the value of the default
3708	loop, if multiple loops are supported ("ev loop default").	4193	loop, if multiple loops are supported ("ev loop default"). The default loop
		4194	will be initialised if it isn't already initialised.
		4195
		4196	For non-multiplicity builds, these macros do nothing, so you always have
		4197	to initialise the loop somewhere.
3709		4198
3710	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>	4199	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
3711		4200
3712	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the	4201	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
3713	default loop has been initialised (C<UC> == unchecked). Their behaviour	4202	default loop has been initialised (C<UC> == unchecked). Their behaviour
…		…
3858	supported). It will also not define any of the structs usually found in	4347	supported). It will also not define any of the structs usually found in
3859	F<event.h> that are not directly supported by the libev core alone.	4348	F<event.h> that are not directly supported by the libev core alone.
3860		4349
3861	In standalone mode, libev will still try to automatically deduce the	4350	In standalone mode, libev will still try to automatically deduce the
3862	configuration, but has to be more conservative.	4351	configuration, but has to be more conservative.
		4352
		4353	=item EV_USE_FLOOR
		4354
		4355	If defined to be C<1>, libev will use the C<floor ()> function for its
		4356	periodic reschedule calculations, otherwise libev will fall back on a
		4357	portable (slower) implementation. If you enable this, you usually have to
		4358	link against libm or something equivalent. Enabling this when the C<floor>
		4359	function is not available will fail, so the safe default is to not enable
		4360	this.
3863		4361
3864	=item EV_USE_MONOTONIC	4362	=item EV_USE_MONOTONIC
3865		4363
3866	If defined to be C<1>, libev will try to detect the availability of the	4364	If defined to be C<1>, libev will try to detect the availability of the
3867	monotonic clock option at both compile time and runtime. Otherwise no	4365	monotonic clock option at both compile time and runtime. Otherwise no
…		…
4000	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.	4498	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
4001		4499
4002	=item EV_ATOMIC_T	4500	=item EV_ATOMIC_T
4003		4501
4004	Libev requires an integer type (suitable for storing C<0> or C<1>) whose	4502	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
4005	access is atomic with respect to other threads or signal contexts. No such	4503	access is atomic and serialised with respect to other threads or signal
4006	type is easily found in the C language, so you can provide your own type	4504	contexts. No such type is easily found in the C language, so you can
4007	that you know is safe for your purposes. It is used both for signal handler "locking"	4505	provide your own type that you know is safe for your purposes. It is used
4008	as well as for signal and thread safety in C<ev_async> watchers.	4506	both for signal handler "locking" as well as for signal and thread safety
		4507	in C<ev_async> watchers.
4009		4508
4010	In the absence of this define, libev will use C<sig_atomic_t volatile>	4509	In the absence of this define, libev will use C<sig_atomic_t volatile>
4011	(from F<signal.h>), which is usually good enough on most platforms.	4510	(from F<signal.h>), which is usually good enough on most platforms,
		4511	although strictly speaking using a type that also implies a memory fence
		4512	is required.
4012		4513
4013	=item EV_H (h)	4514	=item EV_H (h)
4014		4515
4015	The name of the F<ev.h> header file used to include it. The default if	4516	The name of the F<ev.h> header file used to include it. The default if
4016	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be	4517	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
…		…
4040	will have the C<struct ev_loop *> as first argument, and you can create	4541	will have the C<struct ev_loop *> as first argument, and you can create
4041	additional independent event loops. Otherwise there will be no support	4542	additional independent event loops. Otherwise there will be no support
4042	for multiple event loops and there is no first event loop pointer	4543	for multiple event loops and there is no first event loop pointer
4043	argument. Instead, all functions act on the single default loop.	4544	argument. Instead, all functions act on the single default loop.
4044		4545
		4546	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4547	default loop when multiplicity is switched off - you always have to
		4548	initialise the loop manually in this case.
		4549
4045	=item EV_MINPRI	4550	=item EV_MINPRI
4046		4551
4047	=item EV_MAXPRI	4552	=item EV_MAXPRI
4048		4553
4049	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to	4554	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
…		…
4146		4651
4147	With an intelligent-enough linker (gcc+binutils are intelligent enough	4652	With an intelligent-enough linker (gcc+binutils are intelligent enough
4148	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by	4653	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
4149	your program might be left out as well - a binary starting a timer and an	4654	your program might be left out as well - a binary starting a timer and an
4150	I/O watcher then might come out at only 5Kb.	4655	I/O watcher then might come out at only 5Kb.
		4656
		4657	=item EV_API_STATIC
		4658
		4659	If this symbol is defined (by default it is not), then all identifiers
		4660	will have static linkage. This means that libev will not export any
		4661	identifiers, and you cannot link against libev anymore. This can be useful
		4662	when you embed libev, only want to use libev functions in a single file,
		4663	and do not want its identifiers to be visible.
		4664
		4665	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4666	wants to use libev.
4151		4667
4152	=item EV_AVOID_STDIO	4668	=item EV_AVOID_STDIO
4153		4669
4154	If this is set to C<1> at compiletime, then libev will avoid using stdio	4670	If this is set to C<1> at compiletime, then libev will avoid using stdio
4155	functions (printf, scanf, perror etc.). This will increase the code size	4671	functions (printf, scanf, perror etc.). This will increase the code size
…		…
4299	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4815	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
4300		4816
4301	#include "ev_cpp.h"	4817	#include "ev_cpp.h"
4302	#include "ev.c"	4818	#include "ev.c"
4303		4819
4304	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES	4820	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
4305		4821
4306	=head2 THREADS AND COROUTINES	4822	=head2 THREADS AND COROUTINES
4307		4823
4308	=head3 THREADS	4824	=head3 THREADS
4309		4825
…		…
4360	default loop and triggering an C<ev_async> watcher from the default loop	4876	default loop and triggering an C<ev_async> watcher from the default loop
4361	watcher callback into the event loop interested in the signal.	4877	watcher callback into the event loop interested in the signal.
4362		4878
4363	=back	4879	=back
4364		4880
4365	=head4 THREAD LOCKING EXAMPLE	4881	See also L<THREAD LOCKING EXAMPLE>.
4366
4367	Here is a fictitious example of how to run an event loop in a different
4368	thread than where callbacks are being invoked and watchers are
4369	created/added/removed.
4370
4371	For a real-world example, see the C<EV::Loop::Async> perl module,
4372	which uses exactly this technique (which is suited for many high-level
4373	languages).
4374
4375	The example uses a pthread mutex to protect the loop data, a condition
4376	variable to wait for callback invocations, an async watcher to notify the
4377	event loop thread and an unspecified mechanism to wake up the main thread.
4378
4379	First, you need to associate some data with the event loop:
4380
4381	typedef struct {
4382	mutex_t lock; /* global loop lock */
4383	ev_async async_w;
4384	thread_t tid;
4385	cond_t invoke_cv;
4386	} userdata;
4387
4388	void prepare_loop (EV_P)
4389	{
4390	// for simplicity, we use a static userdata struct.
4391	static userdata u;
4392
4393	ev_async_init (&u->async_w, async_cb);
4394	ev_async_start (EV_A_ &u->async_w);
4395
4396	pthread_mutex_init (&u->lock, 0);
4397	pthread_cond_init (&u->invoke_cv, 0);
4398
4399	// now associate this with the loop
4400	ev_set_userdata (EV_A_ u);
4401	ev_set_invoke_pending_cb (EV_A_ l_invoke);
4402	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
4403
4404	// then create the thread running ev_loop
4405	pthread_create (&u->tid, 0, l_run, EV_A);
4406	}
4407
4408	The callback for the C<ev_async> watcher does nothing: the watcher is used
4409	solely to wake up the event loop so it takes notice of any new watchers
4410	that might have been added:
4411
4412	static void
4413	async_cb (EV_P_ ev_async *w, int revents)
4414	{
4415	// just used for the side effects
4416	}
4417
4418	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
4419	protecting the loop data, respectively.
4420
4421	static void
4422	l_release (EV_P)
4423	{
4424	userdata *u = ev_userdata (EV_A);
4425	pthread_mutex_unlock (&u->lock);
4426	}
4427
4428	static void
4429	l_acquire (EV_P)
4430	{
4431	userdata *u = ev_userdata (EV_A);
4432	pthread_mutex_lock (&u->lock);
4433	}
4434
4435	The event loop thread first acquires the mutex, and then jumps straight
4436	into C<ev_run>:
4437
4438	void *
4439	l_run (void *thr_arg)
4440	{
4441	struct ev_loop *loop = (struct ev_loop *)thr_arg;
4442
4443	l_acquire (EV_A);
4444	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
4445	ev_run (EV_A_ 0);
4446	l_release (EV_A);
4447
4448	return 0;
4449	}
4450
4451	Instead of invoking all pending watchers, the C<l_invoke> callback will
4452	signal the main thread via some unspecified mechanism (signals? pipe
4453	writes? C<Async::Interrupt>?) and then waits until all pending watchers
4454	have been called (in a while loop because a) spurious wakeups are possible
4455	and b) skipping inter-thread-communication when there are no pending
4456	watchers is very beneficial):
4457
4458	static void
4459	l_invoke (EV_P)
4460	{
4461	userdata *u = ev_userdata (EV_A);
4462
4463	while (ev_pending_count (EV_A))
4464	{
4465	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
4466	pthread_cond_wait (&u->invoke_cv, &u->lock);
4467	}
4468	}
4469
4470	Now, whenever the main thread gets told to invoke pending watchers, it
4471	will grab the lock, call C<ev_invoke_pending> and then signal the loop
4472	thread to continue:
4473
4474	static void
4475	real_invoke_pending (EV_P)
4476	{
4477	userdata *u = ev_userdata (EV_A);
4478
4479	pthread_mutex_lock (&u->lock);
4480	ev_invoke_pending (EV_A);
4481	pthread_cond_signal (&u->invoke_cv);
4482	pthread_mutex_unlock (&u->lock);
4483	}
4484
4485	Whenever you want to start/stop a watcher or do other modifications to an
4486	event loop, you will now have to lock:
4487
4488	ev_timer timeout_watcher;
4489	userdata *u = ev_userdata (EV_A);
4490
4491	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
4492
4493	pthread_mutex_lock (&u->lock);
4494	ev_timer_start (EV_A_ &timeout_watcher);
4495	ev_async_send (EV_A_ &u->async_w);
4496	pthread_mutex_unlock (&u->lock);
4497
4498	Note that sending the C<ev_async> watcher is required because otherwise
4499	an event loop currently blocking in the kernel will have no knowledge
4500	about the newly added timer. By waking up the loop it will pick up any new
4501	watchers in the next event loop iteration.
4502		4882
4503	=head3 COROUTINES	4883	=head3 COROUTINES
4504		4884
4505	Libev is very accommodating to coroutines ("cooperative threads"):	4885	Libev is very accommodating to coroutines ("cooperative threads"):
4506	libev fully supports nesting calls to its functions from different	4886	libev fully supports nesting calls to its functions from different
…		…
4671	requires, and its I/O model is fundamentally incompatible with the POSIX	5051	requires, and its I/O model is fundamentally incompatible with the POSIX
4672	model. Libev still offers limited functionality on this platform in	5052	model. Libev still offers limited functionality on this platform in
4673	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	5053	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
4674	descriptors. This only applies when using Win32 natively, not when using	5054	descriptors. This only applies when using Win32 natively, not when using
4675	e.g. cygwin. Actually, it only applies to the microsofts own compilers,	5055	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
4676	as every compielr comes with a slightly differently broken/incompatible	5056	as every compiler comes with a slightly differently broken/incompatible
4677	environment.	5057	environment.
4678		5058
4679	Lifting these limitations would basically require the full	5059	Lifting these limitations would basically require the full
4680	re-implementation of the I/O system. If you are into this kind of thing,	5060	re-implementation of the I/O system. If you are into this kind of thing,
4681	then note that glib does exactly that for you in a very portable way (note	5061	then note that glib does exactly that for you in a very portable way (note
…		…
4814		5194
4815	The type C<double> is used to represent timestamps. It is required to	5195	The type C<double> is used to represent timestamps. It is required to
4816	have at least 51 bits of mantissa (and 9 bits of exponent), which is	5196	have at least 51 bits of mantissa (and 9 bits of exponent), which is
4817	good enough for at least into the year 4000 with millisecond accuracy	5197	good enough for at least into the year 4000 with millisecond accuracy
4818	(the design goal for libev). This requirement is overfulfilled by	5198	(the design goal for libev). This requirement is overfulfilled by
4819	implementations using IEEE 754, which is basically all existing ones. With	5199	implementations using IEEE 754, which is basically all existing ones.
		5200
4820	IEEE 754 doubles, you get microsecond accuracy until at least 2200.	5201	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5202	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5203	is either obsolete or somebody patched it to use C<long double> or
		5204	something like that, just kidding).
4821		5205
4822	=back	5206	=back
4823		5207
4824	If you know of other additional requirements drop me a note.	5208	If you know of other additional requirements drop me a note.
4825		5209
…		…
4887	=item Processing ev_async_send: O(number_of_async_watchers)	5271	=item Processing ev_async_send: O(number_of_async_watchers)
4888		5272
4889	=item Processing signals: O(max_signal_number)	5273	=item Processing signals: O(max_signal_number)
4890		5274
4891	Sending involves a system call I<iff> there were no other C<ev_async_send>	5275	Sending involves a system call I<iff> there were no other C<ev_async_send>
4892	calls in the current loop iteration. Checking for async and signal events	5276	calls in the current loop iteration and the loop is currently
		5277	blocked. Checking for async and signal events involves iterating over all
4893	involves iterating over all running async watchers or all signal numbers.	5278	running async watchers or all signal numbers.
4894		5279
4895	=back	5280	=back
4896		5281
4897		5282
4898	=head1 PORTING FROM LIBEV 3.X TO 4.X	5283	=head1 PORTING FROM LIBEV 3.X TO 4.X
…		…
5015	The physical time that is observed. It is apparently strictly monotonic :)	5400	The physical time that is observed. It is apparently strictly monotonic :)
5016		5401
5017	=item wall-clock time	5402	=item wall-clock time
5018		5403
5019	The time and date as shown on clocks. Unlike real time, it can actually	5404	The time and date as shown on clocks. Unlike real time, it can actually
5020	be wrong and jump forwards and backwards, e.g. when the you adjust your	5405	be wrong and jump forwards and backwards, e.g. when you adjust your
5021	clock.	5406	clock.
5022		5407
5023	=item watcher	5408	=item watcher
5024		5409
5025	A data structure that describes interest in certain events. Watchers need	5410	A data structure that describes interest in certain events. Watchers need
…		…
5028	=back	5413	=back
5029		5414
5030	=head1 AUTHOR	5415	=head1 AUTHOR
5031		5416
5032	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael	5417	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
5033	Magnusson and Emanuele Giaquinta.	5418	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
5034		5419

Diff Legend

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