ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.174 by root, Mon Aug 18 23:23:45 2008 UTC vs.
Revision 1.191 by root, Tue Sep 30 19:45:23 2008 UTC

…		…
214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for	214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
215	recommended ones.	215	recommended ones.
216		216
217	See the description of C<ev_embed> watchers for more info.	217	See the description of C<ev_embed> watchers for more info.
218		218
219	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	219	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
220		220
221	Sets the allocation function to use (the prototype is similar - the	221	Sets the allocation function to use (the prototype is similar - the
222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is	222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
223	used to allocate and free memory (no surprises here). If it returns zero	223	used to allocate and free memory (no surprises here). If it returns zero
224	when memory needs to be allocated (C<size != 0>), the library might abort	224	when memory needs to be allocated (C<size != 0>), the library might abort
…		…
250	}	250	}
251		251
252	...	252	...
253	ev_set_allocator (persistent_realloc);	253	ev_set_allocator (persistent_realloc);
254		254
255	=item ev_set_syserr_cb (void (*cb)(const char *msg));	255	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
256		256
257	Set the callback function to call on a retryable system call error (such	257	Set the callback function to call on a retryable system call error (such
258	as failed select, poll, epoll_wait). The message is a printable string	258	as failed select, poll, epoll_wait). The message is a printable string
259	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
260	callback is set, then libev will expect it to remedy the situation, no	260	callback is set, then libev will expect it to remedy the situation, no
…		…
359	writing a server, you should C<accept ()> in a loop to accept as many	359	writing a server, you should C<accept ()> in a loop to accept as many
360	connections as possible during one iteration. You might also want to have	360	connections as possible during one iteration. You might also want to have
361	a look at C<ev_set_io_collect_interval ()> to increase the amount of	361	a look at C<ev_set_io_collect_interval ()> to increase the amount of
362	readiness notifications you get per iteration.	362	readiness notifications you get per iteration.
363		363
		364	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		365	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		366	C<exceptfds> set on that platform).
		367
364	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	368	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
365		369
366	And this is your standard poll(2) backend. It's more complicated	370	And this is your standard poll(2) backend. It's more complicated
367	than select, but handles sparse fds better and has no artificial	371	than select, but handles sparse fds better and has no artificial
368	limit on the number of fds you can use (except it will slow down	372	limit on the number of fds you can use (except it will slow down
369	considerably with a lot of inactive fds). It scales similarly to select,	373	considerably with a lot of inactive fds). It scales similarly to select,
370	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for	374	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
371	performance tips.	375	performance tips.
		376
		377	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		378	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
372		379
373	=item C<EVBACKEND_EPOLL> (value 4, Linux)	380	=item C<EVBACKEND_EPOLL> (value 4, Linux)
374		381
375	For few fds, this backend is a bit little slower than poll and select,	382	For few fds, this backend is a bit little slower than poll and select,
376	but it scales phenomenally better. While poll and select usually scale	383	but it scales phenomenally better. While poll and select usually scale
…		…
389	Please note that epoll sometimes generates spurious notifications, so you	396	Please note that epoll sometimes generates spurious notifications, so you
390	need to use non-blocking I/O or other means to avoid blocking when no data	397	need to use non-blocking I/O or other means to avoid blocking when no data
391	(or space) is available.	398	(or space) is available.
392		399
393	Best performance from this backend is achieved by not unregistering all	400	Best performance from this backend is achieved by not unregistering all
394	watchers for a file descriptor until it has been closed, if possible, i.e.	401	watchers for a file descriptor until it has been closed, if possible,
395	keep at least one watcher active per fd at all times.	402	i.e. keep at least one watcher active per fd at all times. Stopping and
		403	starting a watcher (without re-setting it) also usually doesn't cause
		404	extra overhead.
396		405
397	While nominally embeddable in other event loops, this feature is broken in	406	While nominally embeddable in other event loops, this feature is broken in
398	all kernel versions tested so far.	407	all kernel versions tested so far.
399		408
		409	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		410	C<EVBACKEND_POLL>.
		411
400	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	412	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
401		413
402	Kqueue deserves special mention, as at the time of this writing, it	414	Kqueue deserves special mention, as at the time of this writing, it was
403	was broken on all BSDs except NetBSD (usually it doesn't work reliably	415	broken on all BSDs except NetBSD (usually it doesn't work reliably with
404	with anything but sockets and pipes, except on Darwin, where of course	416	anything but sockets and pipes, except on Darwin, where of course it's
405	it's completely useless). For this reason it's not being "auto-detected"	417	completely useless). For this reason it's not being "auto-detected" unless
406	unless you explicitly specify it explicitly in the flags (i.e. using	418	you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or
407	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)	419	libev was compiled on a known-to-be-good (-enough) system like NetBSD.
408	system like NetBSD.
409		420
410	You still can embed kqueue into a normal poll or select backend and use it	421	You still can embed kqueue into a normal poll or select backend and use it
411	only for sockets (after having made sure that sockets work with kqueue on	422	only for sockets (after having made sure that sockets work with kqueue on
412	the target platform). See C<ev_embed> watchers for more info.	423	the target platform). See C<ev_embed> watchers for more info.
413		424
414	It scales in the same way as the epoll backend, but the interface to the	425	It scales in the same way as the epoll backend, but the interface to the
415	kernel is more efficient (which says nothing about its actual speed, of	426	kernel is more efficient (which says nothing about its actual speed, of
416	course). While stopping, setting and starting an I/O watcher does never	427	course). While stopping, setting and starting an I/O watcher does never
417	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	428	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
418	two event changes per incident, support for C<fork ()> is very bad and it	429	two event changes per incident. Support for C<fork ()> is very bad and it
419	drops fds silently in similarly hard-to-detect cases.	430	drops fds silently in similarly hard-to-detect cases.
420		431
421	This backend usually performs well under most conditions.	432	This backend usually performs well under most conditions.
422		433
423	While nominally embeddable in other event loops, this doesn't work	434	While nominally embeddable in other event loops, this doesn't work
424	everywhere, so you might need to test for this. And since it is broken	435	everywhere, so you might need to test for this. And since it is broken
425	almost everywhere, you should only use it when you have a lot of sockets	436	almost everywhere, you should only use it when you have a lot of sockets
426	(for which it usually works), by embedding it into another event loop	437	(for which it usually works), by embedding it into another event loop
427	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for	438	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
428	sockets.	439	using it only for sockets.
		440
		441	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		442	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		443	C<NOTE_EOF>.
429		444
430	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)	445	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
431		446
432	This is not implemented yet (and might never be, unless you send me an	447	This is not implemented yet (and might never be, unless you send me an
433	implementation). According to reports, C</dev/poll> only supports sockets	448	implementation). According to reports, C</dev/poll> only supports sockets
…		…
446	While this backend scales well, it requires one system call per active	461	While this backend scales well, it requires one system call per active
447	file descriptor per loop iteration. For small and medium numbers of file	462	file descriptor per loop iteration. For small and medium numbers of file
448	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	463	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
449	might perform better.	464	might perform better.
450		465
451	On the positive side, ignoring the spurious readiness notifications, this	466	On the positive side, with the exception of the spurious readiness
452	backend actually performed to specification in all tests and is fully	467	notifications, this backend actually performed fully to specification
453	embeddable, which is a rare feat among the OS-specific backends.	468	in all tests and is fully embeddable, which is a rare feat among the
		469	OS-specific backends.
		470
		471	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		472	C<EVBACKEND_POLL>.
454		473
455	=item C<EVBACKEND_ALL>	474	=item C<EVBACKEND_ALL>
456		475
457	Try all backends (even potentially broken ones that wouldn't be tried	476	Try all backends (even potentially broken ones that wouldn't be tried
458	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	477	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
…		…
464		483
465	If one or more of these are or'ed into the flags value, then only these	484	If one or more of these are or'ed into the flags value, then only these
466	backends will be tried (in the reverse order as listed here). If none are	485	backends will be tried (in the reverse order as listed here). If none are
467	specified, all backends in C<ev_recommended_backends ()> will be tried.	486	specified, all backends in C<ev_recommended_backends ()> will be tried.
468		487
469	The most typical usage is like this:	488	Example: This is the most typical usage.
470		489
471	if (!ev_default_loop (0))	490	if (!ev_default_loop (0))
472	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");	491	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
473		492
474	Restrict libev to the select and poll backends, and do not allow	493	Example: Restrict libev to the select and poll backends, and do not allow
475	environment settings to be taken into account:	494	environment settings to be taken into account:
476		495
477	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);	496	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
478		497
479	Use whatever libev has to offer, but make sure that kqueue is used if	498	Example: Use whatever libev has to offer, but make sure that kqueue is
480	available (warning, breaks stuff, best use only with your own private	499	used if available (warning, breaks stuff, best use only with your own
481	event loop and only if you know the OS supports your types of fds):	500	private event loop and only if you know the OS supports your types of
		501	fds):
482		502
483	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);	503	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
484		504
485	=item struct ev_loop *ev_loop_new (unsigned int flags)	505	=item struct ev_loop *ev_loop_new (unsigned int flags)
486		506
…		…
544		564
545	=item ev_loop_fork (loop)	565	=item ev_loop_fork (loop)
546		566
547	Like C<ev_default_fork>, but acts on an event loop created by	567	Like C<ev_default_fork>, but acts on an event loop created by
548	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	568	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
549	after fork, and how you do this is entirely your own problem.	569	after fork that you want to re-use in the child, and how you do this is
		570	entirely your own problem.
550		571
551	=item int ev_is_default_loop (loop)	572	=item int ev_is_default_loop (loop)
552		573
553	Returns true when the given loop actually is the default loop, false otherwise.	574	Returns true when the given loop is, in fact, the default loop, and false
		575	otherwise.
554		576
555	=item unsigned int ev_loop_count (loop)	577	=item unsigned int ev_loop_count (loop)
556		578
557	Returns the count of loop iterations for the loop, which is identical to	579	Returns the count of loop iterations for the loop, which is identical to
558	the number of times libev did poll for new events. It starts at C<0> and	580	the number of times libev did poll for new events. It starts at C<0> and
…		…
573	received events and started processing them. This timestamp does not	595	received events and started processing them. This timestamp does not
574	change as long as callbacks are being processed, and this is also the base	596	change as long as callbacks are being processed, and this is also the base
575	time used for relative timers. You can treat it as the timestamp of the	597	time used for relative timers. You can treat it as the timestamp of the
576	event occurring (or more correctly, libev finding out about it).	598	event occurring (or more correctly, libev finding out about it).
577		599
		600	=item ev_now_update (loop)
		601
		602	Establishes the current time by querying the kernel, updating the time
		603	returned by C<ev_now ()> in the progress. This is a costly operation and
		604	is usually done automatically within C<ev_loop ()>.
		605
		606	This function is rarely useful, but when some event callback runs for a
		607	very long time without entering the event loop, updating libev's idea of
		608	the current time is a good idea.
		609
		610	See also "The special problem of time updates" in the C<ev_timer> section.
		611
578	=item ev_loop (loop, int flags)	612	=item ev_loop (loop, int flags)
579		613
580	Finally, this is it, the event handler. This function usually is called	614	Finally, this is it, the event handler. This function usually is called
581	after you initialised all your watchers and you want to start handling	615	after you initialised all your watchers and you want to start handling
582	events.	616	events.
…		…
584	If the flags argument is specified as C<0>, it will not return until	618	If the flags argument is specified as C<0>, it will not return until
585	either no event watchers are active anymore or C<ev_unloop> was called.	619	either no event watchers are active anymore or C<ev_unloop> was called.
586		620
587	Please note that an explicit C<ev_unloop> is usually better than	621	Please note that an explicit C<ev_unloop> is usually better than
588	relying on all watchers to be stopped when deciding when a program has	622	relying on all watchers to be stopped when deciding when a program has
589	finished (especially in interactive programs), but having a program that	623	finished (especially in interactive programs), but having a program
590	automatically loops as long as it has to and no longer by virtue of	624	that automatically loops as long as it has to and no longer by virtue
591	relying on its watchers stopping correctly is a thing of beauty.	625	of relying on its watchers stopping correctly, that is truly a thing of
		626	beauty.
592		627
593	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	628	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
594	those events and any outstanding ones, but will not block your process in	629	those events and any already outstanding ones, but will not block your
595	case there are no events and will return after one iteration of the loop.	630	process in case there are no events and will return after one iteration of
		631	the loop.
596		632
597	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	633	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
598	necessary) and will handle those and any outstanding ones. It will block	634	necessary) and will handle those and any already outstanding ones. It
599	your process until at least one new event arrives, and will return after	635	will block your process until at least one new event arrives (which could
600	one iteration of the loop. This is useful if you are waiting for some	636	be an event internal to libev itself, so there is no guarentee that a
601	external event in conjunction with something not expressible using other	637	user-registered callback will be called), and will return after one
		638	iteration of the loop.
		639
		640	This is useful if you are waiting for some external event in conjunction
		641	with something not expressible using other libev watchers (i.e. "roll your
602	libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is	642	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
603	usually a better approach for this kind of thing.	643	usually a better approach for this kind of thing.
604		644
605	Here are the gory details of what C<ev_loop> does:	645	Here are the gory details of what C<ev_loop> does:
606		646
607	- Before the first iteration, call any pending watchers.	647	- Before the first iteration, call any pending watchers.
…		…
617	any active watchers at all will result in not sleeping).	657	any active watchers at all will result in not sleeping).
618	- Sleep if the I/O and timer collect interval say so.	658	- Sleep if the I/O and timer collect interval say so.
619	- Block the process, waiting for any events.	659	- Block the process, waiting for any events.
620	- Queue all outstanding I/O (fd) events.	660	- Queue all outstanding I/O (fd) events.
621	- Update the "event loop time" (ev_now ()), and do time jump adjustments.	661	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
622	- Queue all outstanding timers.	662	- Queue all expired timers.
623	- Queue all outstanding periodics.	663	- Queue all expired periodics.
624	- Unless any events are pending now, queue all idle watchers.	664	- Unless any events are pending now, queue all idle watchers.
625	- Queue all check watchers.	665	- Queue all check watchers.
626	- Call all queued watchers in reverse order (i.e. check watchers first).	666	- Call all queued watchers in reverse order (i.e. check watchers first).
627	Signals and child watchers are implemented as I/O watchers, and will	667	Signals and child watchers are implemented as I/O watchers, and will
628	be handled here by queueing them when their watcher gets executed.	668	be handled here by queueing them when their watcher gets executed.
…		…
651		691
652	=item ev_unref (loop)	692	=item ev_unref (loop)
653		693
654	Ref/unref can be used to add or remove a reference count on the event	694	Ref/unref can be used to add or remove a reference count on the event
655	loop: Every watcher keeps one reference, and as long as the reference	695	loop: Every watcher keeps one reference, and as long as the reference
656	count is nonzero, C<ev_loop> will not return on its own. If you have	696	count is nonzero, C<ev_loop> will not return on its own.
		697
657	a watcher you never unregister that should not keep C<ev_loop> from	698	If you have a watcher you never unregister that should not keep C<ev_loop>
658	returning, ev_unref() after starting, and ev_ref() before stopping it. For	699	from returning, call ev_unref() after starting, and ev_ref() before
		700	stopping it.
		701
659	example, libev itself uses this for its internal signal pipe: It is not	702	As an example, libev itself uses this for its internal signal pipe: It is
660	visible to the libev user and should not keep C<ev_loop> from exiting if	703	not visible to the libev user and should not keep C<ev_loop> from exiting
661	no event watchers registered by it are active. It is also an excellent	704	if no event watchers registered by it are active. It is also an excellent
662	way to do this for generic recurring timers or from within third-party	705	way to do this for generic recurring timers or from within third-party
663	libraries. Just remember to I<unref after start> and I<ref before stop>	706	libraries. Just remember to I<unref after start> and I<ref before stop>
664	(but only if the watcher wasn't active before, or was active before,	707	(but only if the watcher wasn't active before, or was active before,
665	respectively).	708	respectively).
666		709
…		…
689	Setting these to a higher value (the C<interval> I<must> be >= C<0>)	732	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
690	allows libev to delay invocation of I/O and timer/periodic callbacks	733	allows libev to delay invocation of I/O and timer/periodic callbacks
691	to increase efficiency of loop iterations (or to increase power-saving	734	to increase efficiency of loop iterations (or to increase power-saving
692	opportunities).	735	opportunities).
693		736
694	The background is that sometimes your program runs just fast enough to	737	The idea is that sometimes your program runs just fast enough to handle
695	handle one (or very few) event(s) per loop iteration. While this makes	738	one (or very few) event(s) per loop iteration. While this makes the
696	the program responsive, it also wastes a lot of CPU time to poll for new	739	program responsive, it also wastes a lot of CPU time to poll for new
697	events, especially with backends like C<select ()> which have a high	740	events, especially with backends like C<select ()> which have a high
698	overhead for the actual polling but can deliver many events at once.	741	overhead for the actual polling but can deliver many events at once.
699		742
700	By setting a higher I<io collect interval> you allow libev to spend more	743	By setting a higher I<io collect interval> you allow libev to spend more
701	time collecting I/O events, so you can handle more events per iteration,	744	time collecting I/O events, so you can handle more events per iteration,
…		…
703	C<ev_timer>) will be not affected. Setting this to a non-null value will	746	C<ev_timer>) will be not affected. Setting this to a non-null value will
704	introduce an additional C<ev_sleep ()> call into most loop iterations.	747	introduce an additional C<ev_sleep ()> call into most loop iterations.
705		748
706	Likewise, by setting a higher I<timeout collect interval> you allow libev	749	Likewise, by setting a higher I<timeout collect interval> you allow libev
707	to spend more time collecting timeouts, at the expense of increased	750	to spend more time collecting timeouts, at the expense of increased
708	latency (the watcher callback will be called later). C<ev_io> watchers	751	latency/jitter/inexactness (the watcher callback will be called
709	will not be affected. Setting this to a non-null value will not introduce	752	later). C<ev_io> watchers will not be affected. Setting this to a non-null
710	any overhead in libev.	753	value will not introduce any overhead in libev.
711		754
712	Many (busy) programs can usually benefit by setting the I/O collect	755	Many (busy) programs can usually benefit by setting the I/O collect
713	interval to a value near C<0.1> or so, which is often enough for	756	interval to a value near C<0.1> or so, which is often enough for
714	interactive servers (of course not for games), likewise for timeouts. It	757	interactive servers (of course not for games), likewise for timeouts. It
715	usually doesn't make much sense to set it to a lower value than C<0.01>,	758	usually doesn't make much sense to set it to a lower value than C<0.01>,
…		…
723	they fire on, say, one-second boundaries only.	766	they fire on, say, one-second boundaries only.
724		767
725	=item ev_loop_verify (loop)	768	=item ev_loop_verify (loop)
726		769
727	This function only does something when C<EV_VERIFY> support has been	770	This function only does something when C<EV_VERIFY> support has been
728	compiled in. It tries to go through all internal structures and checks	771	compiled in. which is the default for non-minimal builds. It tries to go
729	them for validity. If anything is found to be inconsistent, it will print	772	through all internal structures and checks them for validity. If anything
730	an error message to standard error and call C<abort ()>.	773	is found to be inconsistent, it will print an error message to standard
		774	error and call C<abort ()>.
731		775
732	This can be used to catch bugs inside libev itself: under normal	776	This can be used to catch bugs inside libev itself: under normal
733	circumstances, this function will never abort as of course libev keeps its	777	circumstances, this function will never abort as of course libev keeps its
734	data structures consistent.	778	data structures consistent.
735		779
…		…
851	happen because the watcher could not be properly started because libev	895	happen because the watcher could not be properly started because libev
852	ran out of memory, a file descriptor was found to be closed or any other	896	ran out of memory, a file descriptor was found to be closed or any other
853	problem. You best act on it by reporting the problem and somehow coping	897	problem. You best act on it by reporting the problem and somehow coping
854	with the watcher being stopped.	898	with the watcher being stopped.
855		899
856	Libev will usually signal a few "dummy" events together with an error,	900	Libev will usually signal a few "dummy" events together with an error, for
857	for example it might indicate that a fd is readable or writable, and if	901	example it might indicate that a fd is readable or writable, and if your
858	your callbacks is well-written it can just attempt the operation and cope	902	callbacks is well-written it can just attempt the operation and cope with
859	with the error from read() or write(). This will not work in multi-threaded	903	the error from read() or write(). This will not work in multi-threaded
860	programs, though, so beware.	904	programs, though, as the fd could already be closed and reused for another
		905	thing, so beware.
861		906
862	=back	907	=back
863		908
864	=head2 GENERIC WATCHER FUNCTIONS	909	=head2 GENERIC WATCHER FUNCTIONS
865		910
…		…
881	(or never started) and there are no pending events outstanding.	926	(or never started) and there are no pending events outstanding.
882		927
883	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,	928	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,
884	int revents)>.	929	int revents)>.
885		930
		931	Example: Initialise an C<ev_io> watcher in two steps.
		932
		933	ev_io w;
		934	ev_init (&w, my_cb);
		935	ev_io_set (&w, STDIN_FILENO, EV_READ);
		936
886	=item C<ev_TYPE_set> (ev_TYPE *, [args])	937	=item C<ev_TYPE_set> (ev_TYPE *, [args])
887		938
888	This macro initialises the type-specific parts of a watcher. You need to	939	This macro initialises the type-specific parts of a watcher. You need to
889	call C<ev_init> at least once before you call this macro, but you can	940	call C<ev_init> at least once before you call this macro, but you can
890	call C<ev_TYPE_set> any number of times. You must not, however, call this	941	call C<ev_TYPE_set> any number of times. You must not, however, call this
…		…
892	difference to the C<ev_init> macro).	943	difference to the C<ev_init> macro).
893		944
894	Although some watcher types do not have type-specific arguments	945	Although some watcher types do not have type-specific arguments
895	(e.g. C<ev_prepare>) you still need to call its C<set> macro.	946	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
896		947
		948	See C<ev_init>, above, for an example.
		949
897	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])	950	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
898		951
899	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro	952	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
900	calls into a single call. This is the most convenient method to initialise	953	calls into a single call. This is the most convenient method to initialise
901	a watcher. The same limitations apply, of course.	954	a watcher. The same limitations apply, of course.
902		955
		956	Example: Initialise and set an C<ev_io> watcher in one step.
		957
		958	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		959
903	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)	960	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
904		961
905	Starts (activates) the given watcher. Only active watchers will receive	962	Starts (activates) the given watcher. Only active watchers will receive
906	events. If the watcher is already active nothing will happen.	963	events. If the watcher is already active nothing will happen.
		964
		965	Example: Start the C<ev_io> watcher that is being abused as example in this
		966	whole section.
		967
		968	ev_io_start (EV_DEFAULT_UC, &w);
907		969
908	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)	970	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
909		971
910	Stops the given watcher again (if active) and clears the pending	972	Stops the given watcher again (if active) and clears the pending
911	status. It is possible that stopped watchers are pending (for example,	973	status. It is possible that stopped watchers are pending (for example,
…		…
968		1030
969	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1031	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
970		1032
971	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1033	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
972	C<loop> nor C<revents> need to be valid as long as the watcher callback	1034	C<loop> nor C<revents> need to be valid as long as the watcher callback
973	can deal with that fact.	1035	can deal with that fact, as both are simply passed through to the
		1036	callback.
974		1037
975	=item int ev_clear_pending (loop, ev_TYPE *watcher)	1038	=item int ev_clear_pending (loop, ev_TYPE *watcher)
976		1039
977	If the watcher is pending, this function returns clears its pending status	1040	If the watcher is pending, this function clears its pending status and
978	and returns its C<revents> bitset (as if its callback was invoked). If the	1041	returns its C<revents> bitset (as if its callback was invoked). If the
979	watcher isn't pending it does nothing and returns C<0>.	1042	watcher isn't pending it does nothing and returns C<0>.
980		1043
		1044	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1045	callback to be invoked, which can be accomplished with this function.
		1046
981	=back	1047	=back
982		1048
983		1049
984	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1050	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
985		1051
986	Each watcher has, by default, a member C<void *data> that you can change	1052	Each watcher has, by default, a member C<void *data> that you can change
987	and read at any time, libev will completely ignore it. This can be used	1053	and read at any time: libev will completely ignore it. This can be used
988	to associate arbitrary data with your watcher. If you need more data and	1054	to associate arbitrary data with your watcher. If you need more data and
989	don't want to allocate memory and store a pointer to it in that data	1055	don't want to allocate memory and store a pointer to it in that data
990	member, you can also "subclass" the watcher type and provide your own	1056	member, you can also "subclass" the watcher type and provide your own
991	data:	1057	data:
992		1058
…		…
994	{	1060	{
995	struct ev_io io;	1061	struct ev_io io;
996	int otherfd;	1062	int otherfd;
997	void *somedata;	1063	void *somedata;
998	struct whatever *mostinteresting;	1064	struct whatever *mostinteresting;
999	}	1065	};
		1066
		1067	...
		1068	struct my_io w;
		1069	ev_io_init (&w.io, my_cb, fd, EV_READ);
1000		1070
1001	And since your callback will be called with a pointer to the watcher, you	1071	And since your callback will be called with a pointer to the watcher, you
1002	can cast it back to your own type:	1072	can cast it back to your own type:
1003		1073
1004	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1074	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
…		…
1008	}	1078	}
1009		1079
1010	More interesting and less C-conformant ways of casting your callback type	1080	More interesting and less C-conformant ways of casting your callback type
1011	instead have been omitted.	1081	instead have been omitted.
1012		1082
1013	Another common scenario is having some data structure with multiple	1083	Another common scenario is to use some data structure with multiple
1014	watchers:	1084	embedded watchers:
1015		1085
1016	struct my_biggy	1086	struct my_biggy
1017	{	1087	{
1018	int some_data;	1088	int some_data;
1019	ev_timer t1;	1089	ev_timer t1;
1020	ev_timer t2;	1090	ev_timer t2;
1021	}	1091	}
1022		1092
1023	In this case getting the pointer to C<my_biggy> is a bit more complicated,	1093	In this case getting the pointer to C<my_biggy> is a bit more
1024	you need to use C<offsetof>:	1094	complicated: Either you store the address of your C<my_biggy> struct
		1095	in the C<data> member of the watcher (for woozies), or you need to use
		1096	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1097	programmers):
1025		1098
1026	#include <stddef.h>	1099	#include <stddef.h>
1027		1100
1028	static void	1101	static void
1029	t1_cb (EV_P_ struct ev_timer *w, int revents)	1102	t1_cb (EV_P_ struct ev_timer *w, int revents)
…		…
1069	In general you can register as many read and/or write event watchers per	1142	In general you can register as many read and/or write event watchers per
1070	fd as you want (as long as you don't confuse yourself). Setting all file	1143	fd as you want (as long as you don't confuse yourself). Setting all file
1071	descriptors to non-blocking mode is also usually a good idea (but not	1144	descriptors to non-blocking mode is also usually a good idea (but not
1072	required if you know what you are doing).	1145	required if you know what you are doing).
1073		1146
1074	If you must do this, then force the use of a known-to-be-good backend	1147	If you cannot use non-blocking mode, then force the use of a
1075	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and	1148	known-to-be-good backend (at the time of this writing, this includes only
1076	C<EVBACKEND_POLL>).	1149	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
1077		1150
1078	Another thing you have to watch out for is that it is quite easy to	1151	Another thing you have to watch out for is that it is quite easy to
1079	receive "spurious" readiness notifications, that is your callback might	1152	receive "spurious" readiness notifications, that is your callback might
1080	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1153	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1081	because there is no data. Not only are some backends known to create a	1154	because there is no data. Not only are some backends known to create a
1082	lot of those (for example Solaris ports), it is very easy to get into	1155	lot of those (for example Solaris ports), it is very easy to get into
1083	this situation even with a relatively standard program structure. Thus	1156	this situation even with a relatively standard program structure. Thus
1084	it is best to always use non-blocking I/O: An extra C<read>(2) returning	1157	it is best to always use non-blocking I/O: An extra C<read>(2) returning
1085	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1158	C<EAGAIN> is far preferable to a program hanging until some data arrives.
1086		1159
1087	If you cannot run the fd in non-blocking mode (for example you should not	1160	If you cannot run the fd in non-blocking mode (for example you should
1088	play around with an Xlib connection), then you have to separately re-test	1161	not play around with an Xlib connection), then you have to separately
1089	whether a file descriptor is really ready with a known-to-be good interface	1162	re-test whether a file descriptor is really ready with a known-to-be good
1090	such as poll (fortunately in our Xlib example, Xlib already does this on	1163	interface such as poll (fortunately in our Xlib example, Xlib already
1091	its own, so its quite safe to use).	1164	does this on its own, so its quite safe to use). Some people additionally
		1165	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1166	indefinitely.
		1167
		1168	But really, best use non-blocking mode.
1092		1169
1093	=head3 The special problem of disappearing file descriptors	1170	=head3 The special problem of disappearing file descriptors
1094		1171
1095	Some backends (e.g. kqueue, epoll) need to be told about closing a file	1172	Some backends (e.g. kqueue, epoll) need to be told about closing a file
1096	descriptor (either by calling C<close> explicitly or by any other means,	1173	descriptor (either due to calling C<close> explicitly or any other means,
1097	such as C<dup>). The reason is that you register interest in some file	1174	such as C<dup2>). The reason is that you register interest in some file
1098	descriptor, but when it goes away, the operating system will silently drop	1175	descriptor, but when it goes away, the operating system will silently drop
1099	this interest. If another file descriptor with the same number then is	1176	this interest. If another file descriptor with the same number then is
1100	registered with libev, there is no efficient way to see that this is, in	1177	registered with libev, there is no efficient way to see that this is, in
1101	fact, a different file descriptor.	1178	fact, a different file descriptor.
1102		1179
…		…
1133	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1210	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
1134	C<EVBACKEND_POLL>.	1211	C<EVBACKEND_POLL>.
1135		1212
1136	=head3 The special problem of SIGPIPE	1213	=head3 The special problem of SIGPIPE
1137		1214
1138	While not really specific to libev, it is easy to forget about SIGPIPE:	1215	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1139	when writing to a pipe whose other end has been closed, your program gets	1216	when writing to a pipe whose other end has been closed, your program gets
1140	send a SIGPIPE, which, by default, aborts your program. For most programs	1217	sent a SIGPIPE, which, by default, aborts your program. For most programs
1141	this is sensible behaviour, for daemons, this is usually undesirable.	1218	this is sensible behaviour, for daemons, this is usually undesirable.
1142		1219
1143	So when you encounter spurious, unexplained daemon exits, make sure you	1220	So when you encounter spurious, unexplained daemon exits, make sure you
1144	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon	1221	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1145	somewhere, as that would have given you a big clue).	1222	somewhere, as that would have given you a big clue).
…		…
1152	=item ev_io_init (ev_io *, callback, int fd, int events)	1229	=item ev_io_init (ev_io *, callback, int fd, int events)
1153		1230
1154	=item ev_io_set (ev_io *, int fd, int events)	1231	=item ev_io_set (ev_io *, int fd, int events)
1155		1232
1156	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to	1233	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1157	receive events for and events is either C<EV_READ>, C<EV_WRITE> or	1234	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
1158	C<EV_READ | EV_WRITE> to receive the given events.	1235	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
1159		1236
1160	=item int fd [read-only]	1237	=item int fd [read-only]
1161		1238
1162	The file descriptor being watched.	1239	The file descriptor being watched.
1163		1240
…		…
1175		1252
1176	static void	1253	static void
1177	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1254	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1178	{	1255	{
1179	ev_io_stop (loop, w);	1256	ev_io_stop (loop, w);
1180	.. read from stdin here (or from w->fd) and haqndle any I/O errors	1257	.. read from stdin here (or from w->fd) and handle any I/O errors
1181	}	1258	}
1182		1259
1183	...	1260	...
1184	struct ev_loop *loop = ev_default_init (0);	1261	struct ev_loop *loop = ev_default_init (0);
1185	struct ev_io stdin_readable;	1262	struct ev_io stdin_readable;
…		…
1193	Timer watchers are simple relative timers that generate an event after a	1270	Timer watchers are simple relative timers that generate an event after a
1194	given time, and optionally repeating in regular intervals after that.	1271	given time, and optionally repeating in regular intervals after that.
1195		1272
1196	The timers are based on real time, that is, if you register an event that	1273	The timers are based on real time, that is, if you register an event that
1197	times out after an hour and you reset your system clock to January last	1274	times out after an hour and you reset your system clock to January last
1198	year, it will still time out after (roughly) and hour. "Roughly" because	1275	year, it will still time out after (roughly) one hour. "Roughly" because
1199	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1276	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1200	monotonic clock option helps a lot here).	1277	monotonic clock option helps a lot here).
		1278
		1279	The callback is guaranteed to be invoked only I<after> its timeout has
		1280	passed, but if multiple timers become ready during the same loop iteration
		1281	then order of execution is undefined.
		1282
		1283	=head3 The special problem of time updates
		1284
		1285	Establishing the current time is a costly operation (it usually takes at
		1286	least two system calls): EV therefore updates its idea of the current
		1287	time only before and after C<ev_loop> collects new events, which causes a
		1288	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1289	lots of events in one iteration.
1201		1290
1202	The relative timeouts are calculated relative to the C<ev_now ()>	1291	The relative timeouts are calculated relative to the C<ev_now ()>
1203	time. This is usually the right thing as this timestamp refers to the time	1292	time. This is usually the right thing as this timestamp refers to the time
1204	of the event triggering whatever timeout you are modifying/starting. If	1293	of the event triggering whatever timeout you are modifying/starting. If
1205	you suspect event processing to be delayed and you I<need> to base the timeout	1294	you suspect event processing to be delayed and you I<need> to base the
1206	on the current time, use something like this to adjust for this:	1295	timeout on the current time, use something like this to adjust for this:
1207		1296
1208	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1297	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1209		1298
1210	The callback is guaranteed to be invoked only after its timeout has passed,	1299	If the event loop is suspended for a long time, you can also force an
1211	but if multiple timers become ready during the same loop iteration then	1300	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1212	order of execution is undefined.	1301	()>.
1213		1302
1214	=head3 Watcher-Specific Functions and Data Members	1303	=head3 Watcher-Specific Functions and Data Members
1215		1304
1216	=over 4	1305	=over 4
1217		1306
…		…
1266	ev_timer_again (loop, timer);	1355	ev_timer_again (loop, timer);
1267		1356
1268	This is more slightly efficient then stopping/starting the timer each time	1357	This is more slightly efficient then stopping/starting the timer each time
1269	you want to modify its timeout value.	1358	you want to modify its timeout value.
1270		1359
		1360	Note, however, that it is often even more efficient to remember the
		1361	time of the last activity and let the timer time-out naturally. In the
		1362	callback, you then check whether the time-out is real, or, if there was
		1363	some activity, you reschedule the watcher to time-out in "last_activity +
		1364	timeout - ev_now ()" seconds.
		1365
1271	=item ev_tstamp repeat [read-write]	1366	=item ev_tstamp repeat [read-write]
1272		1367
1273	The current C<repeat> value. Will be used each time the watcher times out	1368	The current C<repeat> value. Will be used each time the watcher times out
1274	or C<ev_timer_again> is called and determines the next timeout (if any),	1369	or C<ev_timer_again> is called, and determines the next timeout (if any),
1275	which is also when any modifications are taken into account.	1370	which is also when any modifications are taken into account.
1276		1371
1277	=back	1372	=back
1278		1373
1279	=head3 Examples	1374	=head3 Examples
…		…
1323	to trigger the event (unlike an C<ev_timer>, which would still trigger	1418	to trigger the event (unlike an C<ev_timer>, which would still trigger
1324	roughly 10 seconds later as it uses a relative timeout).	1419	roughly 10 seconds later as it uses a relative timeout).
1325		1420
1326	C<ev_periodic>s can also be used to implement vastly more complex timers,	1421	C<ev_periodic>s can also be used to implement vastly more complex timers,
1327	such as triggering an event on each "midnight, local time", or other	1422	such as triggering an event on each "midnight, local time", or other
1328	complicated, rules.	1423	complicated rules.
1329		1424
1330	As with timers, the callback is guaranteed to be invoked only when the	1425	As with timers, the callback is guaranteed to be invoked only when the
1331	time (C<at>) has passed, but if multiple periodic timers become ready	1426	time (C<at>) has passed, but if multiple periodic timers become ready
1332	during the same loop iteration then order of execution is undefined.	1427	during the same loop iteration, then order of execution is undefined.
1333		1428
1334	=head3 Watcher-Specific Functions and Data Members	1429	=head3 Watcher-Specific Functions and Data Members
1335		1430
1336	=over 4	1431	=over 4
1337		1432
1338	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1433	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
1339		1434
1340	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1435	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
1341		1436
1342	Lots of arguments, lets sort it out... There are basically three modes of	1437	Lots of arguments, lets sort it out... There are basically three modes of
1343	operation, and we will explain them from simplest to complex:	1438	operation, and we will explain them from simplest to most complex:
1344		1439
1345	=over 4	1440	=over 4
1346		1441
1347	=item * absolute timer (at = time, interval = reschedule_cb = 0)	1442	=item * absolute timer (at = time, interval = reschedule_cb = 0)
1348		1443
1349	In this configuration the watcher triggers an event after the wall clock	1444	In this configuration the watcher triggers an event after the wall clock
1350	time C<at> has passed and doesn't repeat. It will not adjust when a time	1445	time C<at> has passed. It will not repeat and will not adjust when a time
1351	jump occurs, that is, if it is to be run at January 1st 2011 then it will	1446	jump occurs, that is, if it is to be run at January 1st 2011 then it will
1352	run when the system time reaches or surpasses this time.	1447	only run when the system clock reaches or surpasses this time.
1353		1448
1354	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	1449	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1355		1450
1356	In this mode the watcher will always be scheduled to time out at the next	1451	In this mode the watcher will always be scheduled to time out at the next
1357	C<at + N * interval> time (for some integer N, which can also be negative)	1452	C<at + N * interval> time (for some integer N, which can also be negative)
1358	and then repeat, regardless of any time jumps.	1453	and then repeat, regardless of any time jumps.
1359		1454
1360	This can be used to create timers that do not drift with respect to system	1455	This can be used to create timers that do not drift with respect to the
1361	time, for example, here is a C<ev_periodic> that triggers each hour, on	1456	system clock, for example, here is a C<ev_periodic> that triggers each
1362	the hour:	1457	hour, on the hour:
1363		1458
1364	ev_periodic_set (&periodic, 0., 3600., 0);	1459	ev_periodic_set (&periodic, 0., 3600., 0);
1365		1460
1366	This doesn't mean there will always be 3600 seconds in between triggers,	1461	This doesn't mean there will always be 3600 seconds in between triggers,
1367	but only that the callback will be called when the system time shows a	1462	but only that the callback will be called when the system time shows a
…		…
1454	=back	1549	=back
1455		1550
1456	=head3 Examples	1551	=head3 Examples
1457		1552
1458	Example: Call a callback every hour, or, more precisely, whenever the	1553	Example: Call a callback every hour, or, more precisely, whenever the
1459	system clock is divisible by 3600. The callback invocation times have	1554	system time is divisible by 3600. The callback invocation times have
1460	potentially a lot of jitter, but good long-term stability.	1555	potentially a lot of jitter, but good long-term stability.
1461		1556
1462	static void	1557	static void
1463	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1558	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
1464	{	1559	{
…		…
1474	#include <math.h>	1569	#include <math.h>
1475		1570
1476	static ev_tstamp	1571	static ev_tstamp
1477	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)	1572	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
1478	{	1573	{
1479	return fmod (now, 3600.) + 3600.;	1574	return now + (3600. - fmod (now, 3600.));
1480	}	1575	}
1481		1576
1482	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);	1577	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1483		1578
1484	Example: Call a callback every hour, starting now:	1579	Example: Call a callback every hour, starting now:
…		…
1494	Signal watchers will trigger an event when the process receives a specific	1589	Signal watchers will trigger an event when the process receives a specific
1495	signal one or more times. Even though signals are very asynchronous, libev	1590	signal one or more times. Even though signals are very asynchronous, libev
1496	will try it's best to deliver signals synchronously, i.e. as part of the	1591	will try it's best to deliver signals synchronously, i.e. as part of the
1497	normal event processing, like any other event.	1592	normal event processing, like any other event.
1498		1593
		1594	If you want signals asynchronously, just use C<sigaction> as you would
		1595	do without libev and forget about sharing the signal. You can even use
		1596	C<ev_async> from a signal handler to synchronously wake up an event loop.
		1597
1499	You can configure as many watchers as you like per signal. Only when the	1598	You can configure as many watchers as you like per signal. Only when the
1500	first watcher gets started will libev actually register a signal watcher	1599	first watcher gets started will libev actually register a signal handler
1501	with the kernel (thus it coexists with your own signal handlers as long	1600	with the kernel (thus it coexists with your own signal handlers as long as
1502	as you don't register any with libev). Similarly, when the last signal	1601	you don't register any with libev for the same signal). Similarly, when
1503	watcher for a signal is stopped libev will reset the signal handler to	1602	the last signal watcher for a signal is stopped, libev will reset the
1504	SIG_DFL (regardless of what it was set to before).	1603	signal handler to SIG_DFL (regardless of what it was set to before).
1505		1604
1506	If possible and supported, libev will install its handlers with	1605	If possible and supported, libev will install its handlers with
1507	C<SA_RESTART> behaviour enabled, so system calls should not be unduly	1606	C<SA_RESTART> behaviour enabled, so system calls should not be unduly
1508	interrupted. If you have a problem with system calls getting interrupted by	1607	interrupted. If you have a problem with system calls getting interrupted by
1509	signals you can block all signals in an C<ev_check> watcher and unblock	1608	signals you can block all signals in an C<ev_check> watcher and unblock
…		…
1526		1625
1527	=back	1626	=back
1528		1627
1529	=head3 Examples	1628	=head3 Examples
1530		1629
1531	Example: Try to exit cleanly on SIGINT and SIGTERM.	1630	Example: Try to exit cleanly on SIGINT.
1532		1631
1533	static void	1632	static void
1534	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)	1633	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
1535	{	1634	{
1536	ev_unloop (loop, EVUNLOOP_ALL);	1635	ev_unloop (loop, EVUNLOOP_ALL);
1537	}	1636	}
1538		1637
1539	struct ev_signal signal_watcher;	1638	struct ev_signal signal_watcher;
1540	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	1639	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1541	ev_signal_start (loop, &sigint_cb);	1640	ev_signal_start (loop, &signal_watcher);
1542		1641
1543		1642
1544	=head2 C<ev_child> - watch out for process status changes	1643	=head2 C<ev_child> - watch out for process status changes
1545		1644
1546	Child watchers trigger when your process receives a SIGCHLD in response to	1645	Child watchers trigger when your process receives a SIGCHLD in response to
1547	some child status changes (most typically when a child of yours dies). It	1646	some child status changes (most typically when a child of yours dies or
1548	is permissible to install a child watcher I<after> the child has been	1647	exits). It is permissible to install a child watcher I<after> the child
1549	forked (which implies it might have already exited), as long as the event	1648	has been forked (which implies it might have already exited), as long
1550	loop isn't entered (or is continued from a watcher).	1649	as the event loop isn't entered (or is continued from a watcher), i.e.,
		1650	forking and then immediately registering a watcher for the child is fine,
		1651	but forking and registering a watcher a few event loop iterations later is
		1652	not.
1551		1653
1552	Only the default event loop is capable of handling signals, and therefore	1654	Only the default event loop is capable of handling signals, and therefore
1553	you can only register child watchers in the default event loop.	1655	you can only register child watchers in the default event loop.
1554		1656
1555	=head3 Process Interaction	1657	=head3 Process Interaction
…		…
1653	the stat buffer having unspecified contents.	1755	the stat buffer having unspecified contents.
1654		1756
1655	The path I<should> be absolute and I<must not> end in a slash. If it is	1757	The path I<should> be absolute and I<must not> end in a slash. If it is
1656	relative and your working directory changes, the behaviour is undefined.	1758	relative and your working directory changes, the behaviour is undefined.
1657		1759
1658	Since there is no standard to do this, the portable implementation simply	1760	Since there is no standard kernel interface to do this, the portable
1659	calls C<stat (2)> regularly on the path to see if it changed somehow. You	1761	implementation simply calls C<stat (2)> regularly on the path to see if
1660	can specify a recommended polling interval for this case. If you specify	1762	it changed somehow. You can specify a recommended polling interval for
1661	a polling interval of C<0> (highly recommended!) then a I<suitable,	1763	this case. If you specify a polling interval of C<0> (highly recommended!)
1662	unspecified default> value will be used (which you can expect to be around	1764	then a I<suitable, unspecified default> value will be used (which
1663	five seconds, although this might change dynamically). Libev will also	1765	you can expect to be around five seconds, although this might change
1664	impose a minimum interval which is currently around C<0.1>, but thats	1766	dynamically). Libev will also impose a minimum interval which is currently
1665	usually overkill.	1767	around C<0.1>, but thats usually overkill.
1666		1768
1667	This watcher type is not meant for massive numbers of stat watchers,	1769	This watcher type is not meant for massive numbers of stat watchers,
1668	as even with OS-supported change notifications, this can be	1770	as even with OS-supported change notifications, this can be
1669	resource-intensive.	1771	resource-intensive.
1670		1772
1671	At the time of this writing, only the Linux inotify interface is	1773	At the time of this writing, the only OS-specific interface implemented
1672	implemented (implementing kqueue support is left as an exercise for the	1774	is the Linux inotify interface (implementing kqueue support is left as
1673	reader, note, however, that the author sees no way of implementing ev_stat	1775	an exercise for the reader. Note, however, that the author sees no way
1674	semantics with kqueue). Inotify will be used to give hints only and should	1776	of implementing C<ev_stat> semantics with kqueue).
1675	not change the semantics of C<ev_stat> watchers, which means that libev
1676	sometimes needs to fall back to regular polling again even with inotify,
1677	but changes are usually detected immediately, and if the file exists there
1678	will be no polling.
1679		1777
1680	=head3 ABI Issues (Largefile Support)	1778	=head3 ABI Issues (Largefile Support)
1681		1779
1682	Libev by default (unless the user overrides this) uses the default	1780	Libev by default (unless the user overrides this) uses the default
1683	compilation environment, which means that on systems with large file	1781	compilation environment, which means that on systems with large file
…		…
1692	file interfaces available by default (as e.g. FreeBSD does) and not	1790	file interfaces available by default (as e.g. FreeBSD does) and not
1693	optional. Libev cannot simply switch on large file support because it has	1791	optional. Libev cannot simply switch on large file support because it has
1694	to exchange stat structures with application programs compiled using the	1792	to exchange stat structures with application programs compiled using the
1695	default compilation environment.	1793	default compilation environment.
1696		1794
1697	=head3 Inotify	1795	=head3 Inotify and Kqueue
1698		1796
1699	When C<inotify (7)> support has been compiled into libev (generally only	1797	When C<inotify (7)> support has been compiled into libev (generally only
1700	available on Linux) and present at runtime, it will be used to speed up	1798	available with Linux) and present at runtime, it will be used to speed up
1701	change detection where possible. The inotify descriptor will be created lazily	1799	change detection where possible. The inotify descriptor will be created lazily
1702	when the first C<ev_stat> watcher is being started.	1800	when the first C<ev_stat> watcher is being started.
1703		1801
1704	Inotify presence does not change the semantics of C<ev_stat> watchers	1802	Inotify presence does not change the semantics of C<ev_stat> watchers
1705	except that changes might be detected earlier, and in some cases, to avoid	1803	except that changes might be detected earlier, and in some cases, to avoid
1706	making regular C<stat> calls. Even in the presence of inotify support	1804	making regular C<stat> calls. Even in the presence of inotify support
1707	there are many cases where libev has to resort to regular C<stat> polling.	1805	there are many cases where libev has to resort to regular C<stat> polling,
		1806	but as long as the path exists, libev usually gets away without polling.
1708		1807
1709	(There is no support for kqueue, as apparently it cannot be used to	1808	There is no support for kqueue, as apparently it cannot be used to
1710	implement this functionality, due to the requirement of having a file	1809	implement this functionality, due to the requirement of having a file
1711	descriptor open on the object at all times).	1810	descriptor open on the object at all times, and detecting renames, unlinks
		1811	etc. is difficult.
1712		1812
1713	=head3 The special problem of stat time resolution	1813	=head3 The special problem of stat time resolution
1714		1814
1715	The C<stat ()> system call only supports full-second resolution portably, and	1815	The C<stat ()> system call only supports full-second resolution portably, and
1716	even on systems where the resolution is higher, many file systems still	1816	even on systems where the resolution is higher, most file systems still
1717	only support whole seconds.	1817	only support whole seconds.
1718		1818
1719	That means that, if the time is the only thing that changes, you can	1819	That means that, if the time is the only thing that changes, you can
1720	easily miss updates: on the first update, C<ev_stat> detects a change and	1820	easily miss updates: on the first update, C<ev_stat> detects a change and
1721	calls your callback, which does something. When there is another update	1821	calls your callback, which does something. When there is another update
1722	within the same second, C<ev_stat> will be unable to detect it as the stat	1822	within the same second, C<ev_stat> will be unable to detect unless the
1723	data does not change.	1823	stat data does change in other ways (e.g. file size).
1724		1824
1725	The solution to this is to delay acting on a change for slightly more	1825	The solution to this is to delay acting on a change for slightly more
1726	than a second (or till slightly after the next full second boundary), using	1826	than a second (or till slightly after the next full second boundary), using
1727	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);	1827	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
1728	ev_timer_again (loop, w)>).	1828	ev_timer_again (loop, w)>).
…		…
1748	C<path>. The C<interval> is a hint on how quickly a change is expected to	1848	C<path>. The C<interval> is a hint on how quickly a change is expected to
1749	be detected and should normally be specified as C<0> to let libev choose	1849	be detected and should normally be specified as C<0> to let libev choose
1750	a suitable value. The memory pointed to by C<path> must point to the same	1850	a suitable value. The memory pointed to by C<path> must point to the same
1751	path for as long as the watcher is active.	1851	path for as long as the watcher is active.
1752		1852
1753	The callback will receive C<EV_STAT> when a change was detected, relative	1853	The callback will receive an C<EV_STAT> event when a change was detected,
1754	to the attributes at the time the watcher was started (or the last change	1854	relative to the attributes at the time the watcher was started (or the
1755	was detected).	1855	last change was detected).
1756		1856
1757	=item ev_stat_stat (loop, ev_stat *)	1857	=item ev_stat_stat (loop, ev_stat *)
1758		1858
1759	Updates the stat buffer immediately with new values. If you change the	1859	Updates the stat buffer immediately with new values. If you change the
1760	watched path in your callback, you could call this function to avoid	1860	watched path in your callback, you could call this function to avoid
…		…
1843		1943
1844		1944
1845	=head2 C<ev_idle> - when you've got nothing better to do...	1945	=head2 C<ev_idle> - when you've got nothing better to do...
1846		1946
1847	Idle watchers trigger events when no other events of the same or higher	1947	Idle watchers trigger events when no other events of the same or higher
1848	priority are pending (prepare, check and other idle watchers do not	1948	priority are pending (prepare, check and other idle watchers do not count
1849	count).	1949	as receiving "events").
1850		1950
1851	That is, as long as your process is busy handling sockets or timeouts	1951	That is, as long as your process is busy handling sockets or timeouts
1852	(or even signals, imagine) of the same or higher priority it will not be	1952	(or even signals, imagine) of the same or higher priority it will not be
1853	triggered. But when your process is idle (or only lower-priority watchers	1953	triggered. But when your process is idle (or only lower-priority watchers
1854	are pending), the idle watchers are being called once per event loop	1954	are pending), the idle watchers are being called once per event loop
…		…
1893	ev_idle_start (loop, idle_cb);	1993	ev_idle_start (loop, idle_cb);
1894		1994
1895		1995
1896	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	1996	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1897		1997
1898	Prepare and check watchers are usually (but not always) used in tandem:	1998	Prepare and check watchers are usually (but not always) used in pairs:
1899	prepare watchers get invoked before the process blocks and check watchers	1999	prepare watchers get invoked before the process blocks and check watchers
1900	afterwards.	2000	afterwards.
1901		2001
1902	You I<must not> call C<ev_loop> or similar functions that enter	2002	You I<must not> call C<ev_loop> or similar functions that enter
1903	the current event loop from either C<ev_prepare> or C<ev_check>	2003	the current event loop from either C<ev_prepare> or C<ev_check>
…		…
1906	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,	2006	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1907	C<ev_check> so if you have one watcher of each kind they will always be	2007	C<ev_check> so if you have one watcher of each kind they will always be
1908	called in pairs bracketing the blocking call.	2008	called in pairs bracketing the blocking call.
1909		2009
1910	Their main purpose is to integrate other event mechanisms into libev and	2010	Their main purpose is to integrate other event mechanisms into libev and
1911	their use is somewhat advanced. This could be used, for example, to track	2011	their use is somewhat advanced. They could be used, for example, to track
1912	variable changes, implement your own watchers, integrate net-snmp or a	2012	variable changes, implement your own watchers, integrate net-snmp or a
1913	coroutine library and lots more. They are also occasionally useful if	2013	coroutine library and lots more. They are also occasionally useful if
1914	you cache some data and want to flush it before blocking (for example,	2014	you cache some data and want to flush it before blocking (for example,
1915	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>	2015	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1916	watcher).	2016	watcher).
1917		2017
1918	This is done by examining in each prepare call which file descriptors need	2018	This is done by examining in each prepare call which file descriptors
1919	to be watched by the other library, registering C<ev_io> watchers for	2019	need to be watched by the other library, registering C<ev_io> watchers
1920	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2020	for them and starting an C<ev_timer> watcher for any timeouts (many
1921	provide just this functionality). Then, in the check watcher you check for	2021	libraries provide exactly this functionality). Then, in the check watcher,
1922	any events that occurred (by checking the pending status of all watchers	2022	you check for any events that occurred (by checking the pending status
1923	and stopping them) and call back into the library. The I/O and timer	2023	of all watchers and stopping them) and call back into the library. The
1924	callbacks will never actually be called (but must be valid nevertheless,	2024	I/O and timer callbacks will never actually be called (but must be valid
1925	because you never know, you know?).	2025	nevertheless, because you never know, you know?).
1926		2026
1927	As another example, the Perl Coro module uses these hooks to integrate	2027	As another example, the Perl Coro module uses these hooks to integrate
1928	coroutines into libev programs, by yielding to other active coroutines	2028	coroutines into libev programs, by yielding to other active coroutines
1929	during each prepare and only letting the process block if no coroutines	2029	during each prepare and only letting the process block if no coroutines
1930	are ready to run (it's actually more complicated: it only runs coroutines	2030	are ready to run (it's actually more complicated: it only runs coroutines
…		…
1933	loop from blocking if lower-priority coroutines are active, thus mapping	2033	loop from blocking if lower-priority coroutines are active, thus mapping
1934	low-priority coroutines to idle/background tasks).	2034	low-priority coroutines to idle/background tasks).
1935		2035
1936	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	2036	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1937	priority, to ensure that they are being run before any other watchers	2037	priority, to ensure that they are being run before any other watchers
		2038	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2039
1938	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,	2040	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
1939	too) should not activate ("feed") events into libev. While libev fully	2041	activate ("feed") events into libev. While libev fully supports this, they
1940	supports this, they might get executed before other C<ev_check> watchers	2042	might get executed before other C<ev_check> watchers did their job. As
1941	did their job. As C<ev_check> watchers are often used to embed other	2043	C<ev_check> watchers are often used to embed other (non-libev) event
1942	(non-libev) event loops those other event loops might be in an unusable	2044	loops those other event loops might be in an unusable state until their
1943	state until their C<ev_check> watcher ran (always remind yourself to	2045	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
1944	coexist peacefully with others).	2046	others).
1945		2047
1946	=head3 Watcher-Specific Functions and Data Members	2048	=head3 Watcher-Specific Functions and Data Members
1947		2049
1948	=over 4	2050	=over 4
1949		2051
…		…
1951		2053
1952	=item ev_check_init (ev_check *, callback)	2054	=item ev_check_init (ev_check *, callback)
1953		2055
1954	Initialises and configures the prepare or check watcher - they have no	2056	Initialises and configures the prepare or check watcher - they have no
1955	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2057	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1956	macros, but using them is utterly, utterly and completely pointless.	2058	macros, but using them is utterly, utterly, utterly and completely
		2059	pointless.
1957		2060
1958	=back	2061	=back
1959		2062
1960	=head3 Examples	2063	=head3 Examples
1961		2064
…		…
2054	}	2157	}
2055		2158
2056	// do not ever call adns_afterpoll	2159	// do not ever call adns_afterpoll
2057		2160
2058	Method 4: Do not use a prepare or check watcher because the module you	2161	Method 4: Do not use a prepare or check watcher because the module you
2059	want to embed is too inflexible to support it. Instead, you can override	2162	want to embed is not flexible enough to support it. Instead, you can
2060	their poll function. The drawback with this solution is that the main	2163	override their poll function. The drawback with this solution is that the
2061	loop is now no longer controllable by EV. The C<Glib::EV> module does	2164	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
2062	this.	2165	this approach, effectively embedding EV as a client into the horrible
		2166	libglib event loop.
2063		2167
2064	static gint	2168	static gint
2065	event_poll_func (GPollFD *fds, guint nfds, gint timeout)	2169	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2066	{	2170	{
2067	int got_events = 0;	2171	int got_events = 0;
…		…
2098	prioritise I/O.	2202	prioritise I/O.
2099		2203
2100	As an example for a bug workaround, the kqueue backend might only support	2204	As an example for a bug workaround, the kqueue backend might only support
2101	sockets on some platform, so it is unusable as generic backend, but you	2205	sockets on some platform, so it is unusable as generic backend, but you
2102	still want to make use of it because you have many sockets and it scales	2206	still want to make use of it because you have many sockets and it scales
2103	so nicely. In this case, you would create a kqueue-based loop and embed it	2207	so nicely. In this case, you would create a kqueue-based loop and embed
2104	into your default loop (which might use e.g. poll). Overall operation will	2208	it into your default loop (which might use e.g. poll). Overall operation
2105	be a bit slower because first libev has to poll and then call kevent, but	2209	will be a bit slower because first libev has to call C<poll> and then
2106	at least you can use both at what they are best.	2210	C<kevent>, but at least you can use both mechanisms for what they are
		2211	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
2107		2212
2108	As for prioritising I/O: rarely you have the case where some fds have	2213	As for prioritising I/O: under rare circumstances you have the case where
2109	to be watched and handled very quickly (with low latency), and even	2214	some fds have to be watched and handled very quickly (with low latency),
2110	priorities and idle watchers might have too much overhead. In this case	2215	and even priorities and idle watchers might have too much overhead. In
2111	you would put all the high priority stuff in one loop and all the rest in	2216	this case you would put all the high priority stuff in one loop and all
2112	a second one, and embed the second one in the first.	2217	the rest in a second one, and embed the second one in the first.
2113		2218
2114	As long as the watcher is active, the callback will be invoked every time	2219	As long as the watcher is active, the callback will be invoked every time
2115	there might be events pending in the embedded loop. The callback must then	2220	there might be events pending in the embedded loop. The callback must then
2116	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	2221	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
2117	their callbacks (you could also start an idle watcher to give the embedded	2222	their callbacks (you could also start an idle watcher to give the embedded
…		…
2125	interested in that.	2230	interested in that.
2126		2231
2127	Also, there have not currently been made special provisions for forking:	2232	Also, there have not currently been made special provisions for forking:
2128	when you fork, you not only have to call C<ev_loop_fork> on both loops,	2233	when you fork, you not only have to call C<ev_loop_fork> on both loops,
2129	but you will also have to stop and restart any C<ev_embed> watchers	2234	but you will also have to stop and restart any C<ev_embed> watchers
2130	yourself.	2235	yourself - but you can use a fork watcher to handle this automatically,
		2236	and future versions of libev might do just that.
2131		2237
2132	Unfortunately, not all backends are embeddable, only the ones returned by	2238	Unfortunately, not all backends are embeddable: only the ones returned by
2133	C<ev_embeddable_backends> are, which, unfortunately, does not include any	2239	C<ev_embeddable_backends> are, which, unfortunately, does not include any
2134	portable one.	2240	portable one.
2135		2241
2136	So when you want to use this feature you will always have to be prepared	2242	So when you want to use this feature you will always have to be prepared
2137	that you cannot get an embeddable loop. The recommended way to get around	2243	that you cannot get an embeddable loop. The recommended way to get around
2138	this is to have a separate variables for your embeddable loop, try to	2244	this is to have a separate variables for your embeddable loop, try to
2139	create it, and if that fails, use the normal loop for everything.	2245	create it, and if that fails, use the normal loop for everything.
		2246
		2247	=head3 C<ev_embed> and fork
		2248
		2249	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2250	automatically be applied to the embedded loop as well, so no special
		2251	fork handling is required in that case. When the watcher is not running,
		2252	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2253	as applicable.
2140		2254
2141	=head3 Watcher-Specific Functions and Data Members	2255	=head3 Watcher-Specific Functions and Data Members
2142		2256
2143	=over 4	2257	=over 4
2144		2258
…		…
2262	is that the author does not know of a simple (or any) algorithm for a	2376	is that the author does not know of a simple (or any) algorithm for a
2263	multiple-writer-single-reader queue that works in all cases and doesn't	2377	multiple-writer-single-reader queue that works in all cases and doesn't
2264	need elaborate support such as pthreads.	2378	need elaborate support such as pthreads.
2265		2379
2266	That means that if you want to queue data, you have to provide your own	2380	That means that if you want to queue data, you have to provide your own
2267	queue. But at least I can tell you would implement locking around your	2381	queue. But at least I can tell you how to implement locking around your
2268	queue:	2382	queue:
2269		2383
2270	=over 4	2384	=over 4
2271		2385
2272	=item queueing from a signal handler context	2386	=item queueing from a signal handler context
2273		2387
2274	To implement race-free queueing, you simply add to the queue in the signal	2388	To implement race-free queueing, you simply add to the queue in the signal
2275	handler but you block the signal handler in the watcher callback. Here is an example that does that for	2389	handler but you block the signal handler in the watcher callback. Here is
2276	some fictitious SIGUSR1 handler:	2390	an example that does that for some fictitious SIGUSR1 handler:
2277		2391
2278	static ev_async mysig;	2392	static ev_async mysig;
2279		2393
2280	static void	2394	static void
2281	sigusr1_handler (void)	2395	sigusr1_handler (void)
…		…
2348		2462
2349	=item ev_async_init (ev_async *, callback)	2463	=item ev_async_init (ev_async *, callback)
2350		2464
2351	Initialises and configures the async watcher - it has no parameters of any	2465	Initialises and configures the async watcher - it has no parameters of any
2352	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,	2466	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
2353	believe me.	2467	trust me.
2354		2468
2355	=item ev_async_send (loop, ev_async *)	2469	=item ev_async_send (loop, ev_async *)
2356		2470
2357	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	2471	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2358	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	2472	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2359	C<ev_feed_event>, this call is safe to do in other threads, signal or	2473	C<ev_feed_event>, this call is safe to do from other threads, signal or
2360	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	2474	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
2361	section below on what exactly this means).	2475	section below on what exactly this means).
2362		2476
2363	This call incurs the overhead of a system call only once per loop iteration,	2477	This call incurs the overhead of a system call only once per loop iteration,
2364	so while the overhead might be noticeable, it doesn't apply to repeated	2478	so while the overhead might be noticeable, it doesn't apply to repeated
…		…
2570		2684
2571	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.	2685	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2572		2686
2573	See the method-C<set> above for more details.	2687	See the method-C<set> above for more details.
2574		2688
2575	Example:	2689	Example: Use a plain function as callback.
2576		2690
2577	static void io_cb (ev::io &w, int revents) { }	2691	static void io_cb (ev::io &w, int revents) { }
2578	iow.set <io_cb> ();	2692	iow.set <io_cb> ();
2579		2693
2580	=item w->set (struct ev_loop *)	2694	=item w->set (struct ev_loop *)
…		…
2618	Example: Define a class with an IO and idle watcher, start one of them in	2732	Example: Define a class with an IO and idle watcher, start one of them in
2619	the constructor.	2733	the constructor.
2620		2734
2621	class myclass	2735	class myclass
2622	{	2736	{
2623	ev::io io; void io_cb (ev::io &w, int revents);	2737	ev::io io ; void io_cb (ev::io &w, int revents);
2624	ev:idle idle void idle_cb (ev::idle &w, int revents);	2738	ev::idle idle; void idle_cb (ev::idle &w, int revents);
2625		2739
2626	myclass (int fd)	2740	myclass (int fd)
2627	{	2741	{
2628	io .set <myclass, &myclass::io_cb > (this);	2742	io .set <myclass, &myclass::io_cb > (this);
2629	idle.set <myclass, &myclass::idle_cb> (this);	2743	idle.set <myclass, &myclass::idle_cb> (this);
…		…
2645	=item Perl	2759	=item Perl
2646		2760
2647	The EV module implements the full libev API and is actually used to test	2761	The EV module implements the full libev API and is actually used to test
2648	libev. EV is developed together with libev. Apart from the EV core module,	2762	libev. EV is developed together with libev. Apart from the EV core module,
2649	there are additional modules that implement libev-compatible interfaces	2763	there are additional modules that implement libev-compatible interfaces
2650	to C<libadns> (C<EV::ADNS>), C<Net::SNMP> (C<Net::SNMP::EV>) and the	2764	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
2651	C<libglib> event core (C<Glib::EV> and C<EV::Glib>).	2765	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		2766	and C<EV::Glib>).
2652		2767
2653	It can be found and installed via CPAN, its homepage is at	2768	It can be found and installed via CPAN, its homepage is at
2654	L<http://software.schmorp.de/pkg/EV>.	2769	L<http://software.schmorp.de/pkg/EV>.
2655		2770
2656	=item Python	2771	=item Python
…		…
2835		2950
2836	=head2 PREPROCESSOR SYMBOLS/MACROS	2951	=head2 PREPROCESSOR SYMBOLS/MACROS
2837		2952
2838	Libev can be configured via a variety of preprocessor symbols you have to	2953	Libev can be configured via a variety of preprocessor symbols you have to
2839	define before including any of its files. The default in the absence of	2954	define before including any of its files. The default in the absence of
2840	autoconf is noted for every option.	2955	autoconf is documented for every option.
2841		2956
2842	=over 4	2957	=over 4
2843		2958
2844	=item EV_STANDALONE	2959	=item EV_STANDALONE
2845		2960
…		…
3015	When doing priority-based operations, libev usually has to linearly search	3130	When doing priority-based operations, libev usually has to linearly search
3016	all the priorities, so having many of them (hundreds) uses a lot of space	3131	all the priorities, so having many of them (hundreds) uses a lot of space
3017	and time, so using the defaults of five priorities (-2 .. +2) is usually	3132	and time, so using the defaults of five priorities (-2 .. +2) is usually
3018	fine.	3133	fine.
3019		3134
3020	If your embedding application does not need any priorities, defining these both to	3135	If your embedding application does not need any priorities, defining these
3021	C<0> will save some memory and CPU.	3136	both to C<0> will save some memory and CPU.
3022		3137
3023	=item EV_PERIODIC_ENABLE	3138	=item EV_PERIODIC_ENABLE
3024		3139
3025	If undefined or defined to be C<1>, then periodic timers are supported. If	3140	If undefined or defined to be C<1>, then periodic timers are supported. If
3026	defined to be C<0>, then they are not. Disabling them saves a few kB of	3141	defined to be C<0>, then they are not. Disabling them saves a few kB of
…		…
3033	code.	3148	code.
3034		3149
3035	=item EV_EMBED_ENABLE	3150	=item EV_EMBED_ENABLE
3036		3151
3037	If undefined or defined to be C<1>, then embed watchers are supported. If	3152	If undefined or defined to be C<1>, then embed watchers are supported. If
3038	defined to be C<0>, then they are not.	3153	defined to be C<0>, then they are not. Embed watchers rely on most other
		3154	watcher types, which therefore must not be disabled.
3039		3155
3040	=item EV_STAT_ENABLE	3156	=item EV_STAT_ENABLE
3041		3157
3042	If undefined or defined to be C<1>, then stat watchers are supported. If	3158	If undefined or defined to be C<1>, then stat watchers are supported. If
3043	defined to be C<0>, then they are not.	3159	defined to be C<0>, then they are not.
…		…
3075	two).	3191	two).
3076		3192
3077	=item EV_USE_4HEAP	3193	=item EV_USE_4HEAP
3078		3194
3079	Heaps are not very cache-efficient. To improve the cache-efficiency of the	3195	Heaps are not very cache-efficient. To improve the cache-efficiency of the
3080	timer and periodics heap, libev uses a 4-heap when this symbol is defined	3196	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
3081	to C<1>. The 4-heap uses more complicated (longer) code but has	3197	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
3082	noticeably faster performance with many (thousands) of watchers.	3198	faster performance with many (thousands) of watchers.
3083		3199
3084	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>	3200	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3085	(disabled).	3201	(disabled).
3086		3202
3087	=item EV_HEAP_CACHE_AT	3203	=item EV_HEAP_CACHE_AT
3088		3204
3089	Heaps are not very cache-efficient. To improve the cache-efficiency of the	3205	Heaps are not very cache-efficient. To improve the cache-efficiency of the
3090	timer and periodics heap, libev can cache the timestamp (I<at>) within	3206	timer and periodics heaps, libev can cache the timestamp (I<at>) within
3091	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),	3207	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
3092	which uses 8-12 bytes more per watcher and a few hundred bytes more code,	3208	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
3093	but avoids random read accesses on heap changes. This improves performance	3209	but avoids random read accesses on heap changes. This improves performance
3094	noticeably with with many (hundreds) of watchers.	3210	noticeably with many (hundreds) of watchers.
3095		3211
3096	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>	3212	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
3097	(disabled).	3213	(disabled).
3098		3214
3099	=item EV_VERIFY	3215	=item EV_VERIFY
…		…
3105	called once per loop, which can slow down libev. If set to C<3>, then the	3221	called once per loop, which can slow down libev. If set to C<3>, then the
3106	verification code will be called very frequently, which will slow down	3222	verification code will be called very frequently, which will slow down
3107	libev considerably.	3223	libev considerably.
3108		3224
3109	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be	3225	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
3110	C<0.>	3226	C<0>.
3111		3227
3112	=item EV_COMMON	3228	=item EV_COMMON
3113		3229
3114	By default, all watchers have a C<void *data> member. By redefining	3230	By default, all watchers have a C<void *data> member. By redefining
3115	this macro to a something else you can include more and other types of	3231	this macro to a something else you can include more and other types of
…		…
3132	and the way callbacks are invoked and set. Must expand to a struct member	3248	and the way callbacks are invoked and set. Must expand to a struct member
3133	definition and a statement, respectively. See the F<ev.h> header file for	3249	definition and a statement, respectively. See the F<ev.h> header file for
3134	their default definitions. One possible use for overriding these is to	3250	their default definitions. One possible use for overriding these is to
3135	avoid the C<struct ev_loop *> as first argument in all cases, or to use	3251	avoid the C<struct ev_loop *> as first argument in all cases, or to use
3136	method calls instead of plain function calls in C++.	3252	method calls instead of plain function calls in C++.
		3253
		3254	=back
3137		3255
3138	=head2 EXPORTED API SYMBOLS	3256	=head2 EXPORTED API SYMBOLS
3139		3257
3140	If you need to re-export the API (e.g. via a DLL) and you need a list of	3258	If you need to re-export the API (e.g. via a DLL) and you need a list of
3141	exported symbols, you can use the provided F<Symbol.*> files which list	3259	exported symbols, you can use the provided F<Symbol.*> files which list
…		…
3188	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	3306	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3189		3307
3190	#include "ev_cpp.h"	3308	#include "ev_cpp.h"
3191	#include "ev.c"	3309	#include "ev.c"
3192		3310
		3311	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
3193		3312
3194	=head1 THREADS AND COROUTINES	3313	=head2 THREADS AND COROUTINES
3195		3314
3196	=head2 THREADS	3315	=head3 THREADS
3197		3316
3198	Libev itself is completely thread-safe, but it uses no locking. This	3317	All libev functions are reentrant and thread-safe unless explicitly
		3318	documented otherwise, but libev implements no locking itself. This means
3199	means that you can use as many loops as you want in parallel, as long as	3319	that you can use as many loops as you want in parallel, as long as there
3200	only one thread ever calls into one libev function with the same loop	3320	are no concurrent calls into any libev function with the same loop
3201	parameter.	3321	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3322	of course): libev guarantees that different event loops share no data
		3323	structures that need any locking.
3202		3324
3203	Or put differently: calls with different loop parameters can be done in	3325	Or to put it differently: calls with different loop parameters can be done
3204	parallel from multiple threads, calls with the same loop parameter must be	3326	concurrently from multiple threads, calls with the same loop parameter
3205	done serially (but can be done from different threads, as long as only one	3327	must be done serially (but can be done from different threads, as long as
3206	thread ever is inside a call at any point in time, e.g. by using a mutex	3328	only one thread ever is inside a call at any point in time, e.g. by using
3207	per loop).	3329	a mutex per loop).
		3330
		3331	Specifically to support threads (and signal handlers), libev implements
		3332	so-called C<ev_async> watchers, which allow some limited form of
		3333	concurrency on the same event loop, namely waking it up "from the
		3334	outside".
3208		3335
3209	If you want to know which design (one loop, locking, or multiple loops	3336	If you want to know which design (one loop, locking, or multiple loops
3210	without or something else still) is best for your problem, then I cannot	3337	without or something else still) is best for your problem, then I cannot
3211	help you. I can give some generic advice however:	3338	help you, but here is some generic advice:
3212		3339
3213	=over 4	3340	=over 4
3214		3341
3215	=item * most applications have a main thread: use the default libev loop	3342	=item * most applications have a main thread: use the default libev loop
3216	in that thread, or create a separate thread running only the default loop.	3343	in that thread, or create a separate thread running only the default loop.
…		…
3228		3355
3229	Choosing a model is hard - look around, learn, know that usually you can do	3356	Choosing a model is hard - look around, learn, know that usually you can do
3230	better than you currently do :-)	3357	better than you currently do :-)
3231		3358
3232	=item * often you need to talk to some other thread which blocks in the	3359	=item * often you need to talk to some other thread which blocks in the
		3360	event loop.
		3361
3233	event loop - C<ev_async> watchers can be used to wake them up from other	3362	C<ev_async> watchers can be used to wake them up from other threads safely
3234	threads safely (or from signal contexts...).	3363	(or from signal contexts...).
		3364
		3365	An example use would be to communicate signals or other events that only
		3366	work in the default loop by registering the signal watcher with the
		3367	default loop and triggering an C<ev_async> watcher from the default loop
		3368	watcher callback into the event loop interested in the signal.
3235		3369
3236	=back	3370	=back
3237		3371
3238	=head2 COROUTINES	3372	=head3 COROUTINES
3239		3373
3240	Libev is much more accommodating to coroutines ("cooperative threads"):	3374	Libev is very accommodating to coroutines ("cooperative threads"):
3241	libev fully supports nesting calls to it's functions from different	3375	libev fully supports nesting calls to its functions from different
3242	coroutines (e.g. you can call C<ev_loop> on the same loop from two	3376	coroutines (e.g. you can call C<ev_loop> on the same loop from two
3243	different coroutines and switch freely between both coroutines running the	3377	different coroutines, and switch freely between both coroutines running the
3244	loop, as long as you don't confuse yourself). The only exception is that	3378	loop, as long as you don't confuse yourself). The only exception is that
3245	you must not do this from C<ev_periodic> reschedule callbacks.	3379	you must not do this from C<ev_periodic> reschedule callbacks.
3246		3380
3247	Care has been invested into making sure that libev does not keep local	3381	Care has been taken to ensure that libev does not keep local state inside
3248	state inside C<ev_loop>, and other calls do not usually allow coroutine	3382	C<ev_loop>, and other calls do not usually allow for coroutine switches as
3249	switches.	3383	they do not clal any callbacks.
3250		3384
		3385	=head2 COMPILER WARNINGS
3251		3386
3252	=head1 COMPLEXITIES	3387	Depending on your compiler and compiler settings, you might get no or a
		3388	lot of warnings when compiling libev code. Some people are apparently
		3389	scared by this.
3253		3390
3254	In this section the complexities of (many of) the algorithms used inside	3391	However, these are unavoidable for many reasons. For one, each compiler
3255	libev will be explained. For complexity discussions about backends see the	3392	has different warnings, and each user has different tastes regarding
3256	documentation for C<ev_default_init>.	3393	warning options. "Warn-free" code therefore cannot be a goal except when
		3394	targeting a specific compiler and compiler-version.
3257		3395
3258	All of the following are about amortised time: If an array needs to be	3396	Another reason is that some compiler warnings require elaborate
3259	extended, libev needs to realloc and move the whole array, but this	3397	workarounds, or other changes to the code that make it less clear and less
3260	happens asymptotically never with higher number of elements, so O(1) might	3398	maintainable.
3261	mean it might do a lengthy realloc operation in rare cases, but on average
3262	it is much faster and asymptotically approaches constant time.
3263		3399
3264	=over 4	3400	And of course, some compiler warnings are just plain stupid, or simply
		3401	wrong (because they don't actually warn about the condition their message
		3402	seems to warn about). For example, certain older gcc versions had some
		3403	warnings that resulted an extreme number of false positives. These have
		3404	been fixed, but some people still insist on making code warn-free with
		3405	such buggy versions.
3265		3406
3266	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	3407	While libev is written to generate as few warnings as possible,
		3408	"warn-free" code is not a goal, and it is recommended not to build libev
		3409	with any compiler warnings enabled unless you are prepared to cope with
		3410	them (e.g. by ignoring them). Remember that warnings are just that:
		3411	warnings, not errors, or proof of bugs.
3267		3412
3268	This means that, when you have a watcher that triggers in one hour and
3269	there are 100 watchers that would trigger before that then inserting will
3270	have to skip roughly seven (C<ld 100>) of these watchers.
3271		3413
3272	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)	3414	=head2 VALGRIND
3273		3415
3274	That means that changing a timer costs less than removing/adding them	3416	Valgrind has a special section here because it is a popular tool that is
3275	as only the relative motion in the event queue has to be paid for.	3417	highly useful. Unfortunately, valgrind reports are very hard to interpret.
3276		3418
3277	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)	3419	If you think you found a bug (memory leak, uninitialised data access etc.)
		3420	in libev, then check twice: If valgrind reports something like:
3278		3421
3279	These just add the watcher into an array or at the head of a list.	3422	==2274== definitely lost: 0 bytes in 0 blocks.
		3423	==2274== possibly lost: 0 bytes in 0 blocks.
		3424	==2274== still reachable: 256 bytes in 1 blocks.
3280		3425
3281	=item Stopping check/prepare/idle/fork/async watchers: O(1)	3426	Then there is no memory leak, just as memory accounted to global variables
		3427	is not a memleak - the memory is still being refernced, and didn't leak.
3282		3428
3283	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	3429	Similarly, under some circumstances, valgrind might report kernel bugs
		3430	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3431	although an acceptable workaround has been found here), or it might be
		3432	confused.
3284		3433
3285	These watchers are stored in lists then need to be walked to find the	3434	Keep in mind that valgrind is a very good tool, but only a tool. Don't
3286	correct watcher to remove. The lists are usually short (you don't usually	3435	make it into some kind of religion.
3287	have many watchers waiting for the same fd or signal).
3288		3436
3289	=item Finding the next timer in each loop iteration: O(1)	3437	If you are unsure about something, feel free to contact the mailing list
		3438	with the full valgrind report and an explanation on why you think this
		3439	is a bug in libev (best check the archives, too :). However, don't be
		3440	annoyed when you get a brisk "this is no bug" answer and take the chance
		3441	of learning how to interpret valgrind properly.
3290		3442
3291	By virtue of using a binary or 4-heap, the next timer is always found at a	3443	If you need, for some reason, empty reports from valgrind for your project
3292	fixed position in the storage array.	3444	I suggest using suppression lists.
3293		3445
3294	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3295		3446
3296	A change means an I/O watcher gets started or stopped, which requires	3447	=head1 PORTABILITY NOTES
3297	libev to recalculate its status (and possibly tell the kernel, depending
3298	on backend and whether C<ev_io_set> was used).
3299		3448
3300	=item Activating one watcher (putting it into the pending state): O(1)
3301
3302	=item Priority handling: O(number_of_priorities)
3303
3304	Priorities are implemented by allocating some space for each
3305	priority. When doing priority-based operations, libev usually has to
3306	linearly search all the priorities, but starting/stopping and activating
3307	watchers becomes O(1) w.r.t. priority handling.
3308
3309	=item Sending an ev_async: O(1)
3310
3311	=item Processing ev_async_send: O(number_of_async_watchers)
3312
3313	=item Processing signals: O(max_signal_number)
3314
3315	Sending involves a system call I<iff> there were no other C<ev_async_send>
3316	calls in the current loop iteration. Checking for async and signal events
3317	involves iterating over all running async watchers or all signal numbers.
3318
3319	=back
3320
3321
3322	=head1 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS	3449	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
3323		3450
3324	Win32 doesn't support any of the standards (e.g. POSIX) that libev	3451	Win32 doesn't support any of the standards (e.g. POSIX) that libev
3325	requires, and its I/O model is fundamentally incompatible with the POSIX	3452	requires, and its I/O model is fundamentally incompatible with the POSIX
3326	model. Libev still offers limited functionality on this platform in	3453	model. Libev still offers limited functionality on this platform in
3327	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	3454	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
…		…
3338		3465
3339	Not a libev limitation but worth mentioning: windows apparently doesn't	3466	Not a libev limitation but worth mentioning: windows apparently doesn't
3340	accept large writes: instead of resulting in a partial write, windows will	3467	accept large writes: instead of resulting in a partial write, windows will
3341	either accept everything or return C<ENOBUFS> if the buffer is too large,	3468	either accept everything or return C<ENOBUFS> if the buffer is too large,
3342	so make sure you only write small amounts into your sockets (less than a	3469	so make sure you only write small amounts into your sockets (less than a
3343	megabyte seems safe, but thsi apparently depends on the amount of memory	3470	megabyte seems safe, but this apparently depends on the amount of memory
3344	available).	3471	available).
3345		3472
3346	Due to the many, low, and arbitrary limits on the win32 platform and	3473	Due to the many, low, and arbitrary limits on the win32 platform and
3347	the abysmal performance of winsockets, using a large number of sockets	3474	the abysmal performance of winsockets, using a large number of sockets
3348	is not recommended (and not reasonable). If your program needs to use	3475	is not recommended (and not reasonable). If your program needs to use
…		…
3359	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */	3486	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3360		3487
3361	#include "ev.h"	3488	#include "ev.h"
3362		3489
3363	And compile the following F<evwrap.c> file into your project (make sure	3490	And compile the following F<evwrap.c> file into your project (make sure
3364	you do I<not> compile the F<ev.c> or any other embedded soruce files!):	3491	you do I<not> compile the F<ev.c> or any other embedded source files!):
3365		3492
3366	#include "evwrap.h"	3493	#include "evwrap.h"
3367	#include "ev.c"	3494	#include "ev.c"
3368		3495
3369	=over 4	3496	=over 4
…		…
3414	wrap all I/O functions and provide your own fd management, but the cost of	3541	wrap all I/O functions and provide your own fd management, but the cost of
3415	calling select (O(n²)) will likely make this unworkable.	3542	calling select (O(n²)) will likely make this unworkable.
3416		3543
3417	=back	3544	=back
3418		3545
3419
3420	=head1 PORTABILITY REQUIREMENTS	3546	=head2 PORTABILITY REQUIREMENTS
3421		3547
3422	In addition to a working ISO-C implementation, libev relies on a few	3548	In addition to a working ISO-C implementation and of course the
3423	additional extensions:	3549	backend-specific APIs, libev relies on a few additional extensions:
3424		3550
3425	=over 4	3551	=over 4
3426		3552
3427	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible	3553	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
3428	calling conventions regardless of C<ev_watcher_type *>.	3554	calling conventions regardless of C<ev_watcher_type *>.
…		…
3434	calls them using an C<ev_watcher *> internally.	3560	calls them using an C<ev_watcher *> internally.
3435		3561
3436	=item C<sig_atomic_t volatile> must be thread-atomic as well	3562	=item C<sig_atomic_t volatile> must be thread-atomic as well
3437		3563
3438	The type C<sig_atomic_t volatile> (or whatever is defined as	3564	The type C<sig_atomic_t volatile> (or whatever is defined as
3439	C<EV_ATOMIC_T>) must be atomic w.r.t. accesses from different	3565	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
3440	threads. This is not part of the specification for C<sig_atomic_t>, but is	3566	threads. This is not part of the specification for C<sig_atomic_t>, but is
3441	believed to be sufficiently portable.	3567	believed to be sufficiently portable.
3442		3568
3443	=item C<sigprocmask> must work in a threaded environment	3569	=item C<sigprocmask> must work in a threaded environment
3444		3570
…		…
3453	except the initial one, and run the default loop in the initial thread as	3579	except the initial one, and run the default loop in the initial thread as
3454	well.	3580	well.
3455		3581
3456	=item C<long> must be large enough for common memory allocation sizes	3582	=item C<long> must be large enough for common memory allocation sizes
3457		3583
3458	To improve portability and simplify using libev, libev uses C<long>	3584	To improve portability and simplify its API, libev uses C<long> internally
3459	internally instead of C<size_t> when allocating its data structures. On	3585	instead of C<size_t> when allocating its data structures. On non-POSIX
3460	non-POSIX systems (Microsoft...) this might be unexpectedly low, but	3586	systems (Microsoft...) this might be unexpectedly low, but is still at
3461	is still at least 31 bits everywhere, which is enough for hundreds of	3587	least 31 bits everywhere, which is enough for hundreds of millions of
3462	millions of watchers.	3588	watchers.
3463		3589
3464	=item C<double> must hold a time value in seconds with enough accuracy	3590	=item C<double> must hold a time value in seconds with enough accuracy
3465		3591
3466	The type C<double> is used to represent timestamps. It is required to	3592	The type C<double> is used to represent timestamps. It is required to
3467	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	3593	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
…		…
3471	=back	3597	=back
3472		3598
3473	If you know of other additional requirements drop me a note.	3599	If you know of other additional requirements drop me a note.
3474		3600
3475		3601
3476	=head1 COMPILER WARNINGS	3602	=head1 ALGORITHMIC COMPLEXITIES
3477		3603
3478	Depending on your compiler and compiler settings, you might get no or a	3604	In this section the complexities of (many of) the algorithms used inside
3479	lot of warnings when compiling libev code. Some people are apparently	3605	libev will be documented. For complexity discussions about backends see
3480	scared by this.	3606	the documentation for C<ev_default_init>.
3481		3607
3482	However, these are unavoidable for many reasons. For one, each compiler	3608	All of the following are about amortised time: If an array needs to be
3483	has different warnings, and each user has different tastes regarding	3609	extended, libev needs to realloc and move the whole array, but this
3484	warning options. "Warn-free" code therefore cannot be a goal except when	3610	happens asymptotically rarer with higher number of elements, so O(1) might
3485	targeting a specific compiler and compiler-version.	3611	mean that libev does a lengthy realloc operation in rare cases, but on
		3612	average it is much faster and asymptotically approaches constant time.
3486		3613
3487	Another reason is that some compiler warnings require elaborate	3614	=over 4
3488	workarounds, or other changes to the code that make it less clear and less
3489	maintainable.
3490		3615
3491	And of course, some compiler warnings are just plain stupid, or simply	3616	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
3492	wrong (because they don't actually warn about the condition their message
3493	seems to warn about).
3494		3617
3495	While libev is written to generate as few warnings as possible,	3618	This means that, when you have a watcher that triggers in one hour and
3496	"warn-free" code is not a goal, and it is recommended not to build libev	3619	there are 100 watchers that would trigger before that, then inserting will
3497	with any compiler warnings enabled unless you are prepared to cope with	3620	have to skip roughly seven (C<ld 100>) of these watchers.
3498	them (e.g. by ignoring them). Remember that warnings are just that:
3499	warnings, not errors, or proof of bugs.
3500		3621
		3622	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
3501		3623
3502	=head1 VALGRIND	3624	That means that changing a timer costs less than removing/adding them,
		3625	as only the relative motion in the event queue has to be paid for.
3503		3626
3504	Valgrind has a special section here because it is a popular tool that is	3627	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
3505	highly useful, but valgrind reports are very hard to interpret.
3506		3628
3507	If you think you found a bug (memory leak, uninitialised data access etc.)	3629	These just add the watcher into an array or at the head of a list.
3508	in libev, then check twice: If valgrind reports something like:
3509		3630
3510	==2274== definitely lost: 0 bytes in 0 blocks.	3631	=item Stopping check/prepare/idle/fork/async watchers: O(1)
3511	==2274== possibly lost: 0 bytes in 0 blocks.
3512	==2274== still reachable: 256 bytes in 1 blocks.
3513		3632
3514	Then there is no memory leak. Similarly, under some circumstances,	3633	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
3515	valgrind might report kernel bugs as if it were a bug in libev, or it
3516	might be confused (it is a very good tool, but only a tool).
3517		3634
3518	If you are unsure about something, feel free to contact the mailing list	3635	These watchers are stored in lists, so they need to be walked to find the
3519	with the full valgrind report and an explanation on why you think this is	3636	correct watcher to remove. The lists are usually short (you don't usually
3520	a bug in libev. However, don't be annoyed when you get a brisk "this is	3637	have many watchers waiting for the same fd or signal: one is typical, two
3521	no bug" answer and take the chance of learning how to interpret valgrind	3638	is rare).
3522	properly.
3523		3639
3524	If you need, for some reason, empty reports from valgrind for your project	3640	=item Finding the next timer in each loop iteration: O(1)
3525	I suggest using suppression lists.	3641
		3642	By virtue of using a binary or 4-heap, the next timer is always found at a
		3643	fixed position in the storage array.
		3644
		3645	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3646
		3647	A change means an I/O watcher gets started or stopped, which requires
		3648	libev to recalculate its status (and possibly tell the kernel, depending
		3649	on backend and whether C<ev_io_set> was used).
		3650
		3651	=item Activating one watcher (putting it into the pending state): O(1)
		3652
		3653	=item Priority handling: O(number_of_priorities)
		3654
		3655	Priorities are implemented by allocating some space for each
		3656	priority. When doing priority-based operations, libev usually has to
		3657	linearly search all the priorities, but starting/stopping and activating
		3658	watchers becomes O(1) with respect to priority handling.
		3659
		3660	=item Sending an ev_async: O(1)
		3661
		3662	=item Processing ev_async_send: O(number_of_async_watchers)
		3663
		3664	=item Processing signals: O(max_signal_number)
		3665
		3666	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3667	calls in the current loop iteration. Checking for async and signal events
		3668	involves iterating over all running async watchers or all signal numbers.
		3669
		3670	=back
3526		3671
3527		3672
3528	=head1 AUTHOR	3673	=head1 AUTHOR
3529		3674
3530	Marc Lehmann <libev@schmorp.de>.	3675	Marc Lehmann <libev@schmorp.de>.

Diff Legend

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