ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.96 by ayin, Fri Dec 21 10:06:50 2007 UTC vs.
Revision 1.155 by root, Mon May 19 13:48:20 2008 UTC

…		…
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
9	=head1 EXAMPLE PROGRAM	9	=head2 EXAMPLE PROGRAM
10		10
		11	// a single header file is required
11	#include <ev.h>	12	#include <ev.h>
12		13
		14	// every watcher type has its own typedef'd struct
		15	// with the name ev_<type>
13	ev_io stdin_watcher;	16	ev_io stdin_watcher;
14	ev_timer timeout_watcher;	17	ev_timer timeout_watcher;
15		18
		19	// all watcher callbacks have a similar signature
16	/* called when data readable on stdin */	20	// this callback is called when data is readable on stdin
17	static void	21	static void
18	stdin_cb (EV_P_ struct ev_io *w, int revents)	22	stdin_cb (EV_P_ struct ev_io *w, int revents)
19	{	23	{
20	/* puts ("stdin ready"); */	24	puts ("stdin ready");
21	ev_io_stop (EV_A_ w); /* just a syntax example */	25	// for one-shot events, one must manually stop the watcher
22	ev_unloop (EV_A_ EVUNLOOP_ALL); /* leave all loop calls */	26	// with its corresponding stop function.
		27	ev_io_stop (EV_A_ w);
		28
		29	// this causes all nested ev_loop's to stop iterating
		30	ev_unloop (EV_A_ EVUNLOOP_ALL);
23	}	31	}
24		32
		33	// another callback, this time for a time-out
25	static void	34	static void
26	timeout_cb (EV_P_ struct ev_timer *w, int revents)	35	timeout_cb (EV_P_ struct ev_timer *w, int revents)
27	{	36	{
28	/* puts ("timeout"); */	37	puts ("timeout");
29	ev_unloop (EV_A_ EVUNLOOP_ONE); /* leave one loop call */	38	// this causes the innermost ev_loop to stop iterating
		39	ev_unloop (EV_A_ EVUNLOOP_ONE);
30	}	40	}
31		41
32	int	42	int
33	main (void)	43	main (void)
34	{	44	{
		45	// use the default event loop unless you have special needs
35	struct ev_loop *loop = ev_default_loop (0);	46	struct ev_loop *loop = ev_default_loop (0);
36		47
37	/* initialise an io watcher, then start it */	48	// initialise an io watcher, then start it
		49	// this one will watch for stdin to become readable
38	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);	50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
39	ev_io_start (loop, &stdin_watcher);	51	ev_io_start (loop, &stdin_watcher);
40		52
		53	// initialise a timer watcher, then start it
41	/* simple non-repeating 5.5 second timeout */	54	// simple non-repeating 5.5 second timeout
42	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);	55	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
43	ev_timer_start (loop, &timeout_watcher);	56	ev_timer_start (loop, &timeout_watcher);
44		57
45	/* loop till timeout or data ready */	58	// now wait for events to arrive
46	ev_loop (loop, 0);	59	ev_loop (loop, 0);
47		60
		61	// unloop was called, so exit
48	return 0;	62	return 0;
49	}	63	}
50		64
51	=head1 DESCRIPTION	65	=head1 DESCRIPTION
52		66
53	The newest version of this document is also available as a html-formatted	67	The newest version of this document is also available as an html-formatted
54	web page you might find easier to navigate when reading it for the first	68	web page you might find easier to navigate when reading it for the first
55	time: L<http://cvs.schmorp.de/libev/ev.html>.	69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
56		70
57	Libev is an event loop: you register interest in certain events (such as a	71	Libev is an event loop: you register interest in certain events (such as a
58	file descriptor being readable or a timeout occurring), and it will manage	72	file descriptor being readable or a timeout occurring), and it will manage
59	these event sources and provide your program with events.	73	these event sources and provide your program with events.
60		74
…		…
65	You register interest in certain events by registering so-called I<event	79	You register interest in certain events by registering so-called I<event
66	watchers>, which are relatively small C structures you initialise with the	80	watchers>, which are relatively small C structures you initialise with the
67	details of the event, and then hand it over to libev by I<starting> the	81	details of the event, and then hand it over to libev by I<starting> the
68	watcher.	82	watcher.
69		83
70	=head1 FEATURES	84	=head2 FEATURES
71		85
72	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the	86	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
73	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms	87	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
74	for file descriptor events (C<ev_io>), the Linux C<inotify> interface	88	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
75	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers	89	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
…		…
82		96
83	It also is quite fast (see this	97	It also is quite fast (see this
84	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent	98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
85	for example).	99	for example).
86		100
87	=head1 CONVENTIONS	101	=head2 CONVENTIONS
88		102
89	Libev is very configurable. In this manual the default configuration will	103	Libev is very configurable. In this manual the default (and most common)
90	be described, which supports multiple event loops. For more info about	104	configuration will be described, which supports multiple event loops. For
91	various configuration options please have a look at B<EMBED> section in	105	more info about various configuration options please have a look at
92	this manual. If libev was configured without support for multiple event	106	B<EMBED> section in this manual. If libev was configured without support
93	loops, then all functions taking an initial argument of name C<loop>	107	for multiple event loops, then all functions taking an initial argument of
94	(which is always of type C<struct ev_loop *>) will not have this argument.	108	name C<loop> (which is always of type C<struct ev_loop *>) will not have
		109	this argument.
95		110
96	=head1 TIME REPRESENTATION	111	=head2 TIME REPRESENTATION
97		112
98	Libev represents time as a single floating point number, representing the	113	Libev represents time as a single floating point number, representing the
99	(fractional) number of seconds since the (POSIX) epoch (somewhere near	114	(fractional) number of seconds since the (POSIX) epoch (somewhere near
100	the beginning of 1970, details are complicated, don't ask). This type is	115	the beginning of 1970, details are complicated, don't ask). This type is
101	called C<ev_tstamp>, which is what you should use too. It usually aliases	116	called C<ev_tstamp>, which is what you should use too. It usually aliases
…		…
115		130
116	Returns the current time as libev would use it. Please note that the	131	Returns the current time as libev would use it. Please note that the
117	C<ev_now> function is usually faster and also often returns the timestamp	132	C<ev_now> function is usually faster and also often returns the timestamp
118	you actually want to know.	133	you actually want to know.
119		134
		135	=item ev_sleep (ev_tstamp interval)
		136
		137	Sleep for the given interval: The current thread will be blocked until
		138	either it is interrupted or the given time interval has passed. Basically
		139	this is a subsecond-resolution C<sleep ()>.
		140
120	=item int ev_version_major ()	141	=item int ev_version_major ()
121		142
122	=item int ev_version_minor ()	143	=item int ev_version_minor ()
123		144
124	You can find out the major and minor ABI version numbers of the library	145	You can find out the major and minor ABI version numbers of the library
…		…
175	See the description of C<ev_embed> watchers for more info.	196	See the description of C<ev_embed> watchers for more info.
176		197
177	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	198	=item ev_set_allocator (void *(*cb)(void *ptr, long size))
178		199
179	Sets the allocation function to use (the prototype is similar - the	200	Sets the allocation function to use (the prototype is similar - the
180	semantics is identical - to the realloc C function). It is used to	201	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
181	allocate and free memory (no surprises here). If it returns zero when	202	used to allocate and free memory (no surprises here). If it returns zero
182	memory needs to be allocated, the library might abort or take some	203	when memory needs to be allocated (C<size != 0>), the library might abort
183	potentially destructive action. The default is your system realloc	204	or take some potentially destructive action.
184	function.	205
		206	Since some systems (at least OpenBSD and Darwin) fail to implement
		207	correct C<realloc> semantics, libev will use a wrapper around the system
		208	C<realloc> and C<free> functions by default.
185		209
186	You could override this function in high-availability programs to, say,	210	You could override this function in high-availability programs to, say,
187	free some memory if it cannot allocate memory, to use a special allocator,	211	free some memory if it cannot allocate memory, to use a special allocator,
188	or even to sleep a while and retry until some memory is available.	212	or even to sleep a while and retry until some memory is available.
189		213
190	Example: Replace the libev allocator with one that waits a bit and then	214	Example: Replace the libev allocator with one that waits a bit and then
191	retries).	215	retries (example requires a standards-compliant C<realloc>).
192		216
193	static void *	217	static void *
194	persistent_realloc (void *ptr, size_t size)	218	persistent_realloc (void *ptr, size_t size)
195	{	219	{
196	for (;;)	220	for (;;)
…		…
235		259
236	An event loop is described by a C<struct ev_loop *>. The library knows two	260	An event loop is described by a C<struct ev_loop *>. The library knows two
237	types of such loops, the I<default> loop, which supports signals and child	261	types of such loops, the I<default> loop, which supports signals and child
238	events, and dynamically created loops which do not.	262	events, and dynamically created loops which do not.
239		263
240	If you use threads, a common model is to run the default event loop
241	in your main thread (or in a separate thread) and for each thread you
242	create, you also create another event loop. Libev itself does no locking
243	whatsoever, so if you mix calls to the same event loop in different
244	threads, make sure you lock (this is usually a bad idea, though, even if
245	done correctly, because it's hideous and inefficient).
246
247	=over 4	264	=over 4
248		265
249	=item struct ev_loop *ev_default_loop (unsigned int flags)	266	=item struct ev_loop *ev_default_loop (unsigned int flags)
250		267
251	This will initialise the default event loop if it hasn't been initialised	268	This will initialise the default event loop if it hasn't been initialised
…		…
253	false. If it already was initialised it simply returns it (and ignores the	270	false. If it already was initialised it simply returns it (and ignores the
254	flags. If that is troubling you, check C<ev_backend ()> afterwards).	271	flags. If that is troubling you, check C<ev_backend ()> afterwards).
255		272
256	If you don't know what event loop to use, use the one returned from this	273	If you don't know what event loop to use, use the one returned from this
257	function.	274	function.
		275
		276	Note that this function is I<not> thread-safe, so if you want to use it
		277	from multiple threads, you have to lock (note also that this is unlikely,
		278	as loops cannot bes hared easily between threads anyway).
		279
		280	The default loop is the only loop that can handle C<ev_signal> and
		281	C<ev_child> watchers, and to do this, it always registers a handler
		282	for C<SIGCHLD>. If this is a problem for your app you can either
		283	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		284	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		285	C<ev_default_init>.
258		286
259	The flags argument can be used to specify special behaviour or specific	287	The flags argument can be used to specify special behaviour or specific
260	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).	288	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
261		289
262	The following flags are supported:	290	The following flags are supported:
…		…
284	enabling this flag.	312	enabling this flag.
285		313
286	This works by calling C<getpid ()> on every iteration of the loop,	314	This works by calling C<getpid ()> on every iteration of the loop,
287	and thus this might slow down your event loop if you do a lot of loop	315	and thus this might slow down your event loop if you do a lot of loop
288	iterations and little real work, but is usually not noticeable (on my	316	iterations and little real work, but is usually not noticeable (on my
289	Linux system for example, C<getpid> is actually a simple 5-insn sequence	317	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
290	without a syscall and thus I<very> fast, but my Linux system also has	318	without a syscall and thus I<very> fast, but my GNU/Linux system also has
291	C<pthread_atfork> which is even faster).	319	C<pthread_atfork> which is even faster).
292		320
293	The big advantage of this flag is that you can forget about fork (and	321	The big advantage of this flag is that you can forget about fork (and
294	forget about forgetting to tell libev about forking) when you use this	322	forget about forgetting to tell libev about forking) when you use this
295	flag.	323	flag.
…		…
300	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	328	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
301		329
302	This is your standard select(2) backend. Not I<completely> standard, as	330	This is your standard select(2) backend. Not I<completely> standard, as
303	libev tries to roll its own fd_set with no limits on the number of fds,	331	libev tries to roll its own fd_set with no limits on the number of fds,
304	but if that fails, expect a fairly low limit on the number of fds when	332	but if that fails, expect a fairly low limit on the number of fds when
305	using this backend. It doesn't scale too well (O(highest_fd)), but its usually	333	using this backend. It doesn't scale too well (O(highest_fd)), but its
306	the fastest backend for a low number of fds.	334	usually the fastest backend for a low number of (low-numbered :) fds.
		335
		336	To get good performance out of this backend you need a high amount of
		337	parallelity (most of the file descriptors should be busy). If you are
		338	writing a server, you should C<accept ()> in a loop to accept as many
		339	connections as possible during one iteration. You might also want to have
		340	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		341	readiness notifications you get per iteration.
307		342
308	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	343	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
309		344
310	And this is your standard poll(2) backend. It's more complicated than	345	And this is your standard poll(2) backend. It's more complicated
311	select, but handles sparse fds better and has no artificial limit on the	346	than select, but handles sparse fds better and has no artificial
312	number of fds you can use (except it will slow down considerably with a	347	limit on the number of fds you can use (except it will slow down
313	lot of inactive fds). It scales similarly to select, i.e. O(total_fds).	348	considerably with a lot of inactive fds). It scales similarly to select,
		349	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		350	performance tips.
314		351
315	=item C<EVBACKEND_EPOLL> (value 4, Linux)	352	=item C<EVBACKEND_EPOLL> (value 4, Linux)
316		353
317	For few fds, this backend is a bit little slower than poll and select,	354	For few fds, this backend is a bit little slower than poll and select,
318	but it scales phenomenally better. While poll and select usually scale	355	but it scales phenomenally better. While poll and select usually scale
319	like O(total_fds) where n is the total number of fds (or the highest fd),	356	like O(total_fds) where n is the total number of fds (or the highest fd),
320	epoll scales either O(1) or O(active_fds). The epoll design has a number	357	epoll scales either O(1) or O(active_fds). The epoll design has a number
321	of shortcomings, such as silently dropping events in some hard-to-detect	358	of shortcomings, such as silently dropping events in some hard-to-detect
322	cases and rewiring a syscall per fd change, no fork support and bad	359	cases and requiring a syscall per fd change, no fork support and bad
323	support for dup:	360	support for dup.
324		361
325	While stopping, setting and starting an I/O watcher in the same iteration	362	While stopping, setting and starting an I/O watcher in the same iteration
326	will result in some caching, there is still a syscall per such incident	363	will result in some caching, there is still a syscall per such incident
327	(because the fd could point to a different file description now), so its	364	(because the fd could point to a different file description now), so its
328	best to avoid that. Also, C<dup ()>'ed file descriptors might not work	365	best to avoid that. Also, C<dup ()>'ed file descriptors might not work
…		…
330		367
331	Please note that epoll sometimes generates spurious notifications, so you	368	Please note that epoll sometimes generates spurious notifications, so you
332	need to use non-blocking I/O or other means to avoid blocking when no data	369	need to use non-blocking I/O or other means to avoid blocking when no data
333	(or space) is available.	370	(or space) is available.
334		371
		372	Best performance from this backend is achieved by not unregistering all
		373	watchers for a file descriptor until it has been closed, if possible, i.e.
		374	keep at least one watcher active per fd at all times.
		375
		376	While nominally embeddeble in other event loops, this feature is broken in
		377	all kernel versions tested so far.
		378
335	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	379	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
336		380
337	Kqueue deserves special mention, as at the time of this writing, it	381	Kqueue deserves special mention, as at the time of this writing, it
338	was broken on I<all> BSDs (usually it doesn't work with anything but	382	was broken on all BSDs except NetBSD (usually it doesn't work reliably
339	sockets and pipes, except on Darwin, where of course it's completely	383	with anything but sockets and pipes, except on Darwin, where of course
340	useless. On NetBSD, it seems to work for all the FD types I tested, so it
341	is used by default there). For this reason it's not being "autodetected"	384	it's completely useless). For this reason it's not being "autodetected"
342	unless you explicitly specify it explicitly in the flags (i.e. using	385	unless you explicitly specify it explicitly in the flags (i.e. using
343	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)	386	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
344	system like NetBSD.	387	system like NetBSD.
345		388
		389	You still can embed kqueue into a normal poll or select backend and use it
		390	only for sockets (after having made sure that sockets work with kqueue on
		391	the target platform). See C<ev_embed> watchers for more info.
		392
346	It scales in the same way as the epoll backend, but the interface to the	393	It scales in the same way as the epoll backend, but the interface to the
347	kernel is more efficient (which says nothing about its actual speed,	394	kernel is more efficient (which says nothing about its actual speed, of
348	of course). While stopping, setting and starting an I/O watcher does	395	course). While stopping, setting and starting an I/O watcher does never
349	never cause an extra syscall as with epoll, it still adds up to two event	396	cause an extra syscall as with C<EVBACKEND_EPOLL>, it still adds up to
350	changes per incident, support for C<fork ()> is very bad and it drops fds	397	two event changes per incident, support for C<fork ()> is very bad and it
351	silently in similarly hard-to-detetc cases.	398	drops fds silently in similarly hard-to-detect cases.
		399
		400	This backend usually performs well under most conditions.
		401
		402	While nominally embeddable in other event loops, this doesn't work
		403	everywhere, so you might need to test for this. And since it is broken
		404	almost everywhere, you should only use it when you have a lot of sockets
		405	(for which it usually works), by embedding it into another event loop
		406	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for
		407	sockets.
352		408
353	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)	409	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
354		410
355	This is not implemented yet (and might never be).	411	This is not implemented yet (and might never be, unless you send me an
		412	implementation). According to reports, C</dev/poll> only supports sockets
		413	and is not embeddable, which would limit the usefulness of this backend
		414	immensely.
356		415
357	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	416	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
358		417
359	This uses the Solaris 10 event port mechanism. As with everything on Solaris,	418	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
360	it's really slow, but it still scales very well (O(active_fds)).	419	it's really slow, but it still scales very well (O(active_fds)).
361		420
362	Please note that solaris event ports can deliver a lot of spurious	421	Please note that solaris event ports can deliver a lot of spurious
363	notifications, so you need to use non-blocking I/O or other means to avoid	422	notifications, so you need to use non-blocking I/O or other means to avoid
364	blocking when no data (or space) is available.	423	blocking when no data (or space) is available.
365		424
		425	While this backend scales well, it requires one system call per active
		426	file descriptor per loop iteration. For small and medium numbers of file
		427	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		428	might perform better.
		429
		430	On the positive side, ignoring the spurious readiness notifications, this
		431	backend actually performed to specification in all tests and is fully
		432	embeddable, which is a rare feat among the OS-specific backends.
		433
366	=item C<EVBACKEND_ALL>	434	=item C<EVBACKEND_ALL>
367		435
368	Try all backends (even potentially broken ones that wouldn't be tried	436	Try all backends (even potentially broken ones that wouldn't be tried
369	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	437	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
370	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	438	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
371		439
		440	It is definitely not recommended to use this flag.
		441
372	=back	442	=back
373		443
374	If one or more of these are ored into the flags value, then only these	444	If one or more of these are ored into the flags value, then only these
375	backends will be tried (in the reverse order as given here). If none are	445	backends will be tried (in the reverse order as listed here). If none are
376	specified, most compiled-in backend will be tried, usually in reverse	446	specified, all backends in C<ev_recommended_backends ()> will be tried.
377	order of their flag values :)
378		447
379	The most typical usage is like this:	448	The most typical usage is like this:
380		449
381	if (!ev_default_loop (0))	450	if (!ev_default_loop (0))
382	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");	451	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
…		…
396		465
397	Similar to C<ev_default_loop>, but always creates a new event loop that is	466	Similar to C<ev_default_loop>, but always creates a new event loop that is
398	always distinct from the default loop. Unlike the default loop, it cannot	467	always distinct from the default loop. Unlike the default loop, it cannot
399	handle signal and child watchers, and attempts to do so will be greeted by	468	handle signal and child watchers, and attempts to do so will be greeted by
400	undefined behaviour (or a failed assertion if assertions are enabled).	469	undefined behaviour (or a failed assertion if assertions are enabled).
		470
		471	Note that this function I<is> thread-safe, and the recommended way to use
		472	libev with threads is indeed to create one loop per thread, and using the
		473	default loop in the "main" or "initial" thread.
401		474
402	Example: Try to create a event loop that uses epoll and nothing else.	475	Example: Try to create a event loop that uses epoll and nothing else.
403		476
404	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);	477	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
405	if (!epoller)	478	if (!epoller)
…		…
429	Like C<ev_default_destroy>, but destroys an event loop created by an	502	Like C<ev_default_destroy>, but destroys an event loop created by an
430	earlier call to C<ev_loop_new>.	503	earlier call to C<ev_loop_new>.
431		504
432	=item ev_default_fork ()	505	=item ev_default_fork ()
433		506
		507	This function sets a flag that causes subsequent C<ev_loop> iterations
434	This function reinitialises the kernel state for backends that have	508	to reinitialise the kernel state for backends that have one. Despite the
435	one. Despite the name, you can call it anytime, but it makes most sense	509	name, you can call it anytime, but it makes most sense after forking, in
436	after forking, in either the parent or child process (or both, but that	510	the child process (or both child and parent, but that again makes little
437	again makes little sense).	511	sense). You I<must> call it in the child before using any of the libev
		512	functions, and it will only take effect at the next C<ev_loop> iteration.
438		513
439	You I<must> call this function in the child process after forking if and	514	On the other hand, you only need to call this function in the child
440	only if you want to use the event library in both processes. If you just	515	process if and only if you want to use the event library in the child. If
441	fork+exec, you don't have to call it.	516	you just fork+exec, you don't have to call it at all.
442		517
443	The function itself is quite fast and it's usually not a problem to call	518	The function itself is quite fast and it's usually not a problem to call
444	it just in case after a fork. To make this easy, the function will fit in	519	it just in case after a fork. To make this easy, the function will fit in
445	quite nicely into a call to C<pthread_atfork>:	520	quite nicely into a call to C<pthread_atfork>:
446		521
447	pthread_atfork (0, 0, ev_default_fork);	522	pthread_atfork (0, 0, ev_default_fork);
448		523
449	At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use
450	without calling this function, so if you force one of those backends you
451	do not need to care.
452
453	=item ev_loop_fork (loop)	524	=item ev_loop_fork (loop)
454		525
455	Like C<ev_default_fork>, but acts on an event loop created by	526	Like C<ev_default_fork>, but acts on an event loop created by
456	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	527	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
457	after fork, and how you do this is entirely your own problem.	528	after fork, and how you do this is entirely your own problem.
		529
		530	=item int ev_is_default_loop (loop)
		531
		532	Returns true when the given loop actually is the default loop, false otherwise.
458		533
459	=item unsigned int ev_loop_count (loop)	534	=item unsigned int ev_loop_count (loop)
460		535
461	Returns the count of loop iterations for the loop, which is identical to	536	Returns the count of loop iterations for the loop, which is identical to
462	the number of times libev did poll for new events. It starts at C<0> and	537	the number of times libev did poll for new events. It starts at C<0> and
…		…
507	usually a better approach for this kind of thing.	582	usually a better approach for this kind of thing.
508		583
509	Here are the gory details of what C<ev_loop> does:	584	Here are the gory details of what C<ev_loop> does:
510		585
511	- Before the first iteration, call any pending watchers.	586	- Before the first iteration, call any pending watchers.
512	* If there are no active watchers (reference count is zero), return.	587	* If EVFLAG_FORKCHECK was used, check for a fork.
513	- Queue all prepare watchers and then call all outstanding watchers.	588	- If a fork was detected, queue and call all fork watchers.
		589	- Queue and call all prepare watchers.
514	- If we have been forked, recreate the kernel state.	590	- If we have been forked, recreate the kernel state.
515	- Update the kernel state with all outstanding changes.	591	- Update the kernel state with all outstanding changes.
516	- Update the "event loop time".	592	- Update the "event loop time".
517	- Calculate for how long to block.	593	- Calculate for how long to sleep or block, if at all
		594	(active idle watchers, EVLOOP_NONBLOCK or not having
		595	any active watchers at all will result in not sleeping).
		596	- Sleep if the I/O and timer collect interval say so.
518	- Block the process, waiting for any events.	597	- Block the process, waiting for any events.
519	- Queue all outstanding I/O (fd) events.	598	- Queue all outstanding I/O (fd) events.
520	- Update the "event loop time" and do time jump handling.	599	- Update the "event loop time" and do time jump handling.
521	- Queue all outstanding timers.	600	- Queue all outstanding timers.
522	- Queue all outstanding periodics.	601	- Queue all outstanding periodics.
523	- If no events are pending now, queue all idle watchers.	602	- If no events are pending now, queue all idle watchers.
524	- Queue all check watchers.	603	- Queue all check watchers.
525	- Call all queued watchers in reverse order (i.e. check watchers first).	604	- Call all queued watchers in reverse order (i.e. check watchers first).
526	Signals and child watchers are implemented as I/O watchers, and will	605	Signals and child watchers are implemented as I/O watchers, and will
527	be handled here by queueing them when their watcher gets executed.	606	be handled here by queueing them when their watcher gets executed.
528	- If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	607	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
529	were used, return, otherwise continue with step *.	608	were used, or there are no active watchers, return, otherwise
		609	continue with step *.
530		610
531	Example: Queue some jobs and then loop until no events are outsanding	611	Example: Queue some jobs and then loop until no events are outstanding
532	anymore.	612	anymore.
533		613
534	... queue jobs here, make sure they register event watchers as long	614	... queue jobs here, make sure they register event watchers as long
535	... as they still have work to do (even an idle watcher will do..)	615	... as they still have work to do (even an idle watcher will do..)
536	ev_loop (my_loop, 0);	616	ev_loop (my_loop, 0);
…		…
540		620
541	Can be used to make a call to C<ev_loop> return early (but only after it	621	Can be used to make a call to C<ev_loop> return early (but only after it
542	has processed all outstanding events). The C<how> argument must be either	622	has processed all outstanding events). The C<how> argument must be either
543	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	623	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
544	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	624	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
		625
		626	This "unloop state" will be cleared when entering C<ev_loop> again.
545		627
546	=item ev_ref (loop)	628	=item ev_ref (loop)
547		629
548	=item ev_unref (loop)	630	=item ev_unref (loop)
549		631
…		…
554	returning, ev_unref() after starting, and ev_ref() before stopping it. For	636	returning, ev_unref() after starting, and ev_ref() before stopping it. For
555	example, libev itself uses this for its internal signal pipe: It is not	637	example, libev itself uses this for its internal signal pipe: It is not
556	visible to the libev user and should not keep C<ev_loop> from exiting if	638	visible to the libev user and should not keep C<ev_loop> from exiting if
557	no event watchers registered by it are active. It is also an excellent	639	no event watchers registered by it are active. It is also an excellent
558	way to do this for generic recurring timers or from within third-party	640	way to do this for generic recurring timers or from within third-party
559	libraries. Just remember to I<unref after start> and I<ref before stop>.	641	libraries. Just remember to I<unref after start> and I<ref before stop>
		642	(but only if the watcher wasn't active before, or was active before,
		643	respectively).
560		644
561	Example: Create a signal watcher, but keep it from keeping C<ev_loop>	645	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
562	running when nothing else is active.	646	running when nothing else is active.
563		647
564	struct ev_signal exitsig;	648	struct ev_signal exitsig;
…		…
568		652
569	Example: For some weird reason, unregister the above signal handler again.	653	Example: For some weird reason, unregister the above signal handler again.
570		654
571	ev_ref (loop);	655	ev_ref (loop);
572	ev_signal_stop (loop, &exitsig);	656	ev_signal_stop (loop, &exitsig);
		657
		658	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		659
		660	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		661
		662	These advanced functions influence the time that libev will spend waiting
		663	for events. Both are by default C<0>, meaning that libev will try to
		664	invoke timer/periodic callbacks and I/O callbacks with minimum latency.
		665
		666	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		667	allows libev to delay invocation of I/O and timer/periodic callbacks to
		668	increase efficiency of loop iterations.
		669
		670	The background is that sometimes your program runs just fast enough to
		671	handle one (or very few) event(s) per loop iteration. While this makes
		672	the program responsive, it also wastes a lot of CPU time to poll for new
		673	events, especially with backends like C<select ()> which have a high
		674	overhead for the actual polling but can deliver many events at once.
		675
		676	By setting a higher I<io collect interval> you allow libev to spend more
		677	time collecting I/O events, so you can handle more events per iteration,
		678	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		679	C<ev_timer>) will be not affected. Setting this to a non-null value will
		680	introduce an additional C<ev_sleep ()> call into most loop iterations.
		681
		682	Likewise, by setting a higher I<timeout collect interval> you allow libev
		683	to spend more time collecting timeouts, at the expense of increased
		684	latency (the watcher callback will be called later). C<ev_io> watchers
		685	will not be affected. Setting this to a non-null value will not introduce
		686	any overhead in libev.
		687
		688	Many (busy) programs can usually benefit by setting the io collect
		689	interval to a value near C<0.1> or so, which is often enough for
		690	interactive servers (of course not for games), likewise for timeouts. It
		691	usually doesn't make much sense to set it to a lower value than C<0.01>,
		692	as this approsaches the timing granularity of most systems.
573		693
574	=back	694	=back
575		695
576		696
577	=head1 ANATOMY OF A WATCHER	697	=head1 ANATOMY OF A WATCHER
…		…
676		796
677	=item C<EV_FORK>	797	=item C<EV_FORK>
678		798
679	The event loop has been resumed in the child process after fork (see	799	The event loop has been resumed in the child process after fork (see
680	C<ev_fork>).	800	C<ev_fork>).
		801
		802	=item C<EV_ASYNC>
		803
		804	The given async watcher has been asynchronously notified (see C<ev_async>).
681		805
682	=item C<EV_ERROR>	806	=item C<EV_ERROR>
683		807
684	An unspecified error has occured, the watcher has been stopped. This might	808	An unspecified error has occured, the watcher has been stopped. This might
685	happen because the watcher could not be properly started because libev	809	happen because the watcher could not be properly started because libev
…		…
903	In general you can register as many read and/or write event watchers per	1027	In general you can register as many read and/or write event watchers per
904	fd as you want (as long as you don't confuse yourself). Setting all file	1028	fd as you want (as long as you don't confuse yourself). Setting all file
905	descriptors to non-blocking mode is also usually a good idea (but not	1029	descriptors to non-blocking mode is also usually a good idea (but not
906	required if you know what you are doing).	1030	required if you know what you are doing).
907		1031
908	You have to be careful with dup'ed file descriptors, though. Some backends
909	(the linux epoll backend is a notable example) cannot handle dup'ed file
910	descriptors correctly if you register interest in two or more fds pointing
911	to the same underlying file/socket/etc. description (that is, they share
912	the same underlying "file open").
913
914	If you must do this, then force the use of a known-to-be-good backend	1032	If you must do this, then force the use of a known-to-be-good backend
915	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and	1033	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
916	C<EVBACKEND_POLL>).	1034	C<EVBACKEND_POLL>).
917		1035
918	Another thing you have to watch out for is that it is quite easy to	1036	Another thing you have to watch out for is that it is quite easy to
919	receive "spurious" readyness notifications, that is your callback might	1037	receive "spurious" readiness notifications, that is your callback might
920	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1038	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
921	because there is no data. Not only are some backends known to create a	1039	because there is no data. Not only are some backends known to create a
922	lot of those (for example solaris ports), it is very easy to get into	1040	lot of those (for example solaris ports), it is very easy to get into
923	this situation even with a relatively standard program structure. Thus	1041	this situation even with a relatively standard program structure. Thus
924	it is best to always use non-blocking I/O: An extra C<read>(2) returning	1042	it is best to always use non-blocking I/O: An extra C<read>(2) returning
…		…
952	optimisations to libev.	1070	optimisations to libev.
953		1071
954	=head3 The special problem of dup'ed file descriptors	1072	=head3 The special problem of dup'ed file descriptors
955		1073
956	Some backends (e.g. epoll), cannot register events for file descriptors,	1074	Some backends (e.g. epoll), cannot register events for file descriptors,
957	but only events for the underlying file descriptions. That menas when you	1075	but only events for the underlying file descriptions. That means when you
958	have C<dup ()>'ed file descriptors and register events for them, only one	1076	have C<dup ()>'ed file descriptors or weirder constellations, and register
959	file descriptor might actually receive events.	1077	events for them, only one file descriptor might actually receive events.
960		1078
961	There is no workaorund possible except not registering events	1079	There is no workaround possible except not registering events
962	for potentially C<dup ()>'ed file descriptors or to resort to	1080	for potentially C<dup ()>'ed file descriptors, or to resort to
963	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1081	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
964		1082
965	=head3 The special problem of fork	1083	=head3 The special problem of fork
966		1084
967	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1085	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
…		…
971	To support fork in your programs, you either have to call	1089	To support fork in your programs, you either have to call
972	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,	1090	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
973	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1091	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
974	C<EVBACKEND_POLL>.	1092	C<EVBACKEND_POLL>.
975		1093
		1094	=head3 The special problem of SIGPIPE
		1095
		1096	While not really specific to libev, it is easy to forget about SIGPIPE:
		1097	when reading from a pipe whose other end has been closed, your program
		1098	gets send a SIGPIPE, which, by default, aborts your program. For most
		1099	programs this is sensible behaviour, for daemons, this is usually
		1100	undesirable.
		1101
		1102	So when you encounter spurious, unexplained daemon exits, make sure you
		1103	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1104	somewhere, as that would have given you a big clue).
		1105
976		1106
977	=head3 Watcher-Specific Functions	1107	=head3 Watcher-Specific Functions
978		1108
979	=over 4	1109	=over 4
980		1110
…		…
993	=item int events [read-only]	1123	=item int events [read-only]
994		1124
995	The events being watched.	1125	The events being watched.
996		1126
997	=back	1127	=back
		1128
		1129	=head3 Examples
998		1130
999	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well	1131	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1000	readable, but only once. Since it is likely line-buffered, you could	1132	readable, but only once. Since it is likely line-buffered, you could
1001	attempt to read a whole line in the callback.	1133	attempt to read a whole line in the callback.
1002		1134
…		…
1055	configure a timer to trigger every 10 seconds, then it will trigger at	1187	configure a timer to trigger every 10 seconds, then it will trigger at
1056	exactly 10 second intervals. If, however, your program cannot keep up with	1188	exactly 10 second intervals. If, however, your program cannot keep up with
1057	the timer (because it takes longer than those 10 seconds to do stuff) the	1189	the timer (because it takes longer than those 10 seconds to do stuff) the
1058	timer will not fire more than once per event loop iteration.	1190	timer will not fire more than once per event loop iteration.
1059		1191
1060	=item ev_timer_again (loop)	1192	=item ev_timer_again (loop, ev_timer *)
1061		1193
1062	This will act as if the timer timed out and restart it again if it is	1194	This will act as if the timer timed out and restart it again if it is
1063	repeating. The exact semantics are:	1195	repeating. The exact semantics are:
1064		1196
1065	If the timer is pending, its pending status is cleared.	1197	If the timer is pending, its pending status is cleared.
…		…
1100	or C<ev_timer_again> is called and determines the next timeout (if any),	1232	or C<ev_timer_again> is called and determines the next timeout (if any),
1101	which is also when any modifications are taken into account.	1233	which is also when any modifications are taken into account.
1102		1234
1103	=back	1235	=back
1104		1236
		1237	=head3 Examples
		1238
1105	Example: Create a timer that fires after 60 seconds.	1239	Example: Create a timer that fires after 60 seconds.
1106		1240
1107	static void	1241	static void
1108	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	1242	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
1109	{	1243	{
…		…
1172	In this configuration the watcher triggers an event at the wallclock time	1306	In this configuration the watcher triggers an event at the wallclock time
1173	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1307	C<at> and doesn't repeat. It will not adjust when a time jump occurs,
1174	that is, if it is to be run at January 1st 2011 then it will run when the	1308	that is, if it is to be run at January 1st 2011 then it will run when the
1175	system time reaches or surpasses this time.	1309	system time reaches or surpasses this time.
1176		1310
1177	=item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	1311	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
1178		1312
1179	In this mode the watcher will always be scheduled to time out at the next	1313	In this mode the watcher will always be scheduled to time out at the next
1180	C<at + N * interval> time (for some integer N, which can also be negative)	1314	C<at + N * interval> time (for some integer N, which can also be negative)
1181	and then repeat, regardless of any time jumps.	1315	and then repeat, regardless of any time jumps.
1182		1316
…		…
1239	Simply stops and restarts the periodic watcher again. This is only useful	1373	Simply stops and restarts the periodic watcher again. This is only useful
1240	when you changed some parameters or the reschedule callback would return	1374	when you changed some parameters or the reschedule callback would return
1241	a different time than the last time it was called (e.g. in a crond like	1375	a different time than the last time it was called (e.g. in a crond like
1242	program when the crontabs have changed).	1376	program when the crontabs have changed).
1243		1377
		1378	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1379
		1380	When active, returns the absolute time that the watcher is supposed to
		1381	trigger next.
		1382
1244	=item ev_tstamp offset [read-write]	1383	=item ev_tstamp offset [read-write]
1245		1384
1246	When repeating, this contains the offset value, otherwise this is the	1385	When repeating, this contains the offset value, otherwise this is the
1247	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	1386	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
1248		1387
…		…
1259		1398
1260	The current reschedule callback, or C<0>, if this functionality is	1399	The current reschedule callback, or C<0>, if this functionality is
1261	switched off. Can be changed any time, but changes only take effect when	1400	switched off. Can be changed any time, but changes only take effect when
1262	the periodic timer fires or C<ev_periodic_again> is being called.	1401	the periodic timer fires or C<ev_periodic_again> is being called.
1263		1402
1264	=item ev_tstamp at [read-only]
1265
1266	When active, contains the absolute time that the watcher is supposed to
1267	trigger next.
1268
1269	=back	1403	=back
		1404
		1405	=head3 Examples
1270		1406
1271	Example: Call a callback every hour, or, more precisely, whenever the	1407	Example: Call a callback every hour, or, more precisely, whenever the
1272	system clock is divisible by 3600. The callback invocation times have	1408	system clock is divisible by 3600. The callback invocation times have
1273	potentially a lot of jittering, but good long-term stability.	1409	potentially a lot of jittering, but good long-term stability.
1274		1410
…		…
1314	with the kernel (thus it coexists with your own signal handlers as long	1450	with the kernel (thus it coexists with your own signal handlers as long
1315	as you don't register any with libev). Similarly, when the last signal	1451	as you don't register any with libev). Similarly, when the last signal
1316	watcher for a signal is stopped libev will reset the signal handler to	1452	watcher for a signal is stopped libev will reset the signal handler to
1317	SIG_DFL (regardless of what it was set to before).	1453	SIG_DFL (regardless of what it was set to before).
1318		1454
		1455	If possible and supported, libev will install its handlers with
		1456	C<SA_RESTART> behaviour enabled, so syscalls should not be unduly
		1457	interrupted. If you have a problem with syscalls getting interrupted by
		1458	signals you can block all signals in an C<ev_check> watcher and unblock
		1459	them in an C<ev_prepare> watcher.
		1460
1319	=head3 Watcher-Specific Functions and Data Members	1461	=head3 Watcher-Specific Functions and Data Members
1320		1462
1321	=over 4	1463	=over 4
1322		1464
1323	=item ev_signal_init (ev_signal *, callback, int signum)	1465	=item ev_signal_init (ev_signal *, callback, int signum)
…		…
1331		1473
1332	The signal the watcher watches out for.	1474	The signal the watcher watches out for.
1333		1475
1334	=back	1476	=back
1335		1477
		1478	=head3 Examples
		1479
		1480	Example: Try to exit cleanly on SIGINT and SIGTERM.
		1481
		1482	static void
		1483	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
		1484	{
		1485	ev_unloop (loop, EVUNLOOP_ALL);
		1486	}
		1487
		1488	struct ev_signal signal_watcher;
		1489	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1490	ev_signal_start (loop, &sigint_cb);
		1491
1336		1492
1337	=head2 C<ev_child> - watch out for process status changes	1493	=head2 C<ev_child> - watch out for process status changes
1338		1494
1339	Child watchers trigger when your process receives a SIGCHLD in response to	1495	Child watchers trigger when your process receives a SIGCHLD in response to
1340	some child status changes (most typically when a child of yours dies).	1496	some child status changes (most typically when a child of yours dies). It
		1497	is permissible to install a child watcher I<after> the child has been
		1498	forked (which implies it might have already exited), as long as the event
		1499	loop isn't entered (or is continued from a watcher).
		1500
		1501	Only the default event loop is capable of handling signals, and therefore
		1502	you can only rgeister child watchers in the default event loop.
		1503
		1504	=head3 Process Interaction
		1505
		1506	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1507	initialised. This is necessary to guarantee proper behaviour even if
		1508	the first child watcher is started after the child exits. The occurance
		1509	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1510	synchronously as part of the event loop processing. Libev always reaps all
		1511	children, even ones not watched.
		1512
		1513	=head3 Overriding the Built-In Processing
		1514
		1515	Libev offers no special support for overriding the built-in child
		1516	processing, but if your application collides with libev's default child
		1517	handler, you can override it easily by installing your own handler for
		1518	C<SIGCHLD> after initialising the default loop, and making sure the
		1519	default loop never gets destroyed. You are encouraged, however, to use an
		1520	event-based approach to child reaping and thus use libev's support for
		1521	that, so other libev users can use C<ev_child> watchers freely.
1341		1522
1342	=head3 Watcher-Specific Functions and Data Members	1523	=head3 Watcher-Specific Functions and Data Members
1343		1524
1344	=over 4	1525	=over 4
1345		1526
1346	=item ev_child_init (ev_child *, callback, int pid)	1527	=item ev_child_init (ev_child *, callback, int pid, int trace)
1347		1528
1348	=item ev_child_set (ev_child *, int pid)	1529	=item ev_child_set (ev_child *, int pid, int trace)
1349		1530
1350	Configures the watcher to wait for status changes of process C<pid> (or	1531	Configures the watcher to wait for status changes of process C<pid> (or
1351	I<any> process if C<pid> is specified as C<0>). The callback can look	1532	I<any> process if C<pid> is specified as C<0>). The callback can look
1352	at the C<rstatus> member of the C<ev_child> watcher structure to see	1533	at the C<rstatus> member of the C<ev_child> watcher structure to see
1353	the status word (use the macros from C<sys/wait.h> and see your systems	1534	the status word (use the macros from C<sys/wait.h> and see your systems
1354	C<waitpid> documentation). The C<rpid> member contains the pid of the	1535	C<waitpid> documentation). The C<rpid> member contains the pid of the
1355	process causing the status change.	1536	process causing the status change. C<trace> must be either C<0> (only
		1537	activate the watcher when the process terminates) or C<1> (additionally
		1538	activate the watcher when the process is stopped or continued).
1356		1539
1357	=item int pid [read-only]	1540	=item int pid [read-only]
1358		1541
1359	The process id this watcher watches out for, or C<0>, meaning any process id.	1542	The process id this watcher watches out for, or C<0>, meaning any process id.
1360		1543
…		…
1367	The process exit/trace status caused by C<rpid> (see your systems	1550	The process exit/trace status caused by C<rpid> (see your systems
1368	C<waitpid> and C<sys/wait.h> documentation for details).	1551	C<waitpid> and C<sys/wait.h> documentation for details).
1369		1552
1370	=back	1553	=back
1371		1554
1372	Example: Try to exit cleanly on SIGINT and SIGTERM.	1555	=head3 Examples
		1556
		1557	Example: C<fork()> a new process and install a child handler to wait for
		1558	its completion.
		1559
		1560	ev_child cw;
1373		1561
1374	static void	1562	static void
1375	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)	1563	child_cb (EV_P_ struct ev_child *w, int revents)
1376	{	1564	{
1377	ev_unloop (loop, EVUNLOOP_ALL);	1565	ev_child_stop (EV_A_ w);
		1566	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1378	}	1567	}
1379		1568
1380	struct ev_signal signal_watcher;	1569	pid_t pid = fork ();
1381	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	1570
1382	ev_signal_start (loop, &sigint_cb);	1571	if (pid < 0)
		1572	// error
		1573	else if (pid == 0)
		1574	{
		1575	// the forked child executes here
		1576	exit (1);
		1577	}
		1578	else
		1579	{
		1580	ev_child_init (&cw, child_cb, pid, 0);
		1581	ev_child_start (EV_DEFAULT_ &cw);
		1582	}
1383		1583
1384		1584
1385	=head2 C<ev_stat> - did the file attributes just change?	1585	=head2 C<ev_stat> - did the file attributes just change?
1386		1586
1387	This watches a filesystem path for attribute changes. That is, it calls	1587	This watches a filesystem path for attribute changes. That is, it calls
…		…
1410	as even with OS-supported change notifications, this can be	1610	as even with OS-supported change notifications, this can be
1411	resource-intensive.	1611	resource-intensive.
1412		1612
1413	At the time of this writing, only the Linux inotify interface is	1613	At the time of this writing, only the Linux inotify interface is
1414	implemented (implementing kqueue support is left as an exercise for the	1614	implemented (implementing kqueue support is left as an exercise for the
		1615	reader, note, however, that the author sees no way of implementing ev_stat
1415	reader). Inotify will be used to give hints only and should not change the	1616	semantics with kqueue). Inotify will be used to give hints only and should
1416	semantics of C<ev_stat> watchers, which means that libev sometimes needs	1617	not change the semantics of C<ev_stat> watchers, which means that libev
1417	to fall back to regular polling again even with inotify, but changes are	1618	sometimes needs to fall back to regular polling again even with inotify,
1418	usually detected immediately, and if the file exists there will be no	1619	but changes are usually detected immediately, and if the file exists there
1419	polling.	1620	will be no polling.
		1621
		1622	=head3 ABI Issues (Largefile Support)
		1623
		1624	Libev by default (unless the user overrides this) uses the default
		1625	compilation environment, which means that on systems with optionally
		1626	disabled large file support, you get the 32 bit version of the stat
		1627	structure. When using the library from programs that change the ABI to
		1628	use 64 bit file offsets the programs will fail. In that case you have to
		1629	compile libev with the same flags to get binary compatibility. This is
		1630	obviously the case with any flags that change the ABI, but the problem is
		1631	most noticably with ev_stat and largefile support.
		1632
		1633	=head3 Inotify
		1634
		1635	When C<inotify (7)> support has been compiled into libev (generally only
		1636	available on Linux) and present at runtime, it will be used to speed up
		1637	change detection where possible. The inotify descriptor will be created lazily
		1638	when the first C<ev_stat> watcher is being started.
		1639
		1640	Inotify presence does not change the semantics of C<ev_stat> watchers
		1641	except that changes might be detected earlier, and in some cases, to avoid
		1642	making regular C<stat> calls. Even in the presence of inotify support
		1643	there are many cases where libev has to resort to regular C<stat> polling.
		1644
		1645	(There is no support for kqueue, as apparently it cannot be used to
		1646	implement this functionality, due to the requirement of having a file
		1647	descriptor open on the object at all times).
		1648
		1649	=head3 The special problem of stat time resolution
		1650
		1651	The C<stat ()> syscall only supports full-second resolution portably, and
		1652	even on systems where the resolution is higher, many filesystems still
		1653	only support whole seconds.
		1654
		1655	That means that, if the time is the only thing that changes, you can
		1656	easily miss updates: on the first update, C<ev_stat> detects a change and
		1657	calls your callback, which does something. When there is another update
		1658	within the same second, C<ev_stat> will be unable to detect it as the stat
		1659	data does not change.
		1660
		1661	The solution to this is to delay acting on a change for slightly more
		1662	than a second (or till slightly after the next full second boundary), using
		1663	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		1664	ev_timer_again (loop, w)>).
		1665
		1666	The C<.02> offset is added to work around small timing inconsistencies
		1667	of some operating systems (where the second counter of the current time
		1668	might be be delayed. One such system is the Linux kernel, where a call to
		1669	C<gettimeofday> might return a timestamp with a full second later than
		1670	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		1671	update file times then there will be a small window where the kernel uses
		1672	the previous second to update file times but libev might already execute
		1673	the timer callback).
1420		1674
1421	=head3 Watcher-Specific Functions and Data Members	1675	=head3 Watcher-Specific Functions and Data Members
1422		1676
1423	=over 4	1677	=over 4
1424		1678
…		…
1430	C<path>. The C<interval> is a hint on how quickly a change is expected to	1684	C<path>. The C<interval> is a hint on how quickly a change is expected to
1431	be detected and should normally be specified as C<0> to let libev choose	1685	be detected and should normally be specified as C<0> to let libev choose
1432	a suitable value. The memory pointed to by C<path> must point to the same	1686	a suitable value. The memory pointed to by C<path> must point to the same
1433	path for as long as the watcher is active.	1687	path for as long as the watcher is active.
1434		1688
1435	The callback will be receive C<EV_STAT> when a change was detected,	1689	The callback will receive C<EV_STAT> when a change was detected, relative
1436	relative to the attributes at the time the watcher was started (or the	1690	to the attributes at the time the watcher was started (or the last change
1437	last change was detected).	1691	was detected).
1438		1692
1439	=item ev_stat_stat (ev_stat *)	1693	=item ev_stat_stat (loop, ev_stat *)
1440		1694
1441	Updates the stat buffer immediately with new values. If you change the	1695	Updates the stat buffer immediately with new values. If you change the
1442	watched path in your callback, you could call this fucntion to avoid	1696	watched path in your callback, you could call this function to avoid
1443	detecting this change (while introducing a race condition). Can also be	1697	detecting this change (while introducing a race condition if you are not
1444	useful simply to find out the new values.	1698	the only one changing the path). Can also be useful simply to find out the
		1699	new values.
1445		1700
1446	=item ev_statdata attr [read-only]	1701	=item ev_statdata attr [read-only]
1447		1702
1448	The most-recently detected attributes of the file. Although the type is of	1703	The most-recently detected attributes of the file. Although the type is
1449	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types	1704	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1450	suitable for your system. If the C<st_nlink> member is C<0>, then there	1705	suitable for your system, but you can only rely on the POSIX-standardised
		1706	members to be present. If the C<st_nlink> member is C<0>, then there was
1451	was some error while C<stat>ing the file.	1707	some error while C<stat>ing the file.
1452		1708
1453	=item ev_statdata prev [read-only]	1709	=item ev_statdata prev [read-only]
1454		1710
1455	The previous attributes of the file. The callback gets invoked whenever	1711	The previous attributes of the file. The callback gets invoked whenever
1456	C<prev> != C<attr>.	1712	C<prev> != C<attr>, or, more precisely, one or more of these members
		1713	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		1714	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
1457		1715
1458	=item ev_tstamp interval [read-only]	1716	=item ev_tstamp interval [read-only]
1459		1717
1460	The specified interval.	1718	The specified interval.
1461		1719
1462	=item const char *path [read-only]	1720	=item const char *path [read-only]
1463		1721
1464	The filesystem path that is being watched.	1722	The filesystem path that is being watched.
1465		1723
1466	=back	1724	=back
		1725
		1726	=head3 Examples
1467		1727
1468	Example: Watch C</etc/passwd> for attribute changes.	1728	Example: Watch C</etc/passwd> for attribute changes.
1469		1729
1470	static void	1730	static void
1471	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)	1731	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
…		…
1484	}	1744	}
1485		1745
1486	...	1746	...
1487	ev_stat passwd;	1747	ev_stat passwd;
1488		1748
1489	ev_stat_init (&passwd, passwd_cb, "/etc/passwd");	1749	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1490	ev_stat_start (loop, &passwd);	1750	ev_stat_start (loop, &passwd);
		1751
		1752	Example: Like above, but additionally use a one-second delay so we do not
		1753	miss updates (however, frequent updates will delay processing, too, so
		1754	one might do the work both on C<ev_stat> callback invocation I<and> on
		1755	C<ev_timer> callback invocation).
		1756
		1757	static ev_stat passwd;
		1758	static ev_timer timer;
		1759
		1760	static void
		1761	timer_cb (EV_P_ ev_timer *w, int revents)
		1762	{
		1763	ev_timer_stop (EV_A_ w);
		1764
		1765	/* now it's one second after the most recent passwd change */
		1766	}
		1767
		1768	static void
		1769	stat_cb (EV_P_ ev_stat *w, int revents)
		1770	{
		1771	/* reset the one-second timer */
		1772	ev_timer_again (EV_A_ &timer);
		1773	}
		1774
		1775	...
		1776	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		1777	ev_stat_start (loop, &passwd);
		1778	ev_timer_init (&timer, timer_cb, 0., 1.02);
1491		1779
1492		1780
1493	=head2 C<ev_idle> - when you've got nothing better to do...	1781	=head2 C<ev_idle> - when you've got nothing better to do...
1494		1782
1495	Idle watchers trigger events when no other events of the same or higher	1783	Idle watchers trigger events when no other events of the same or higher
…		…
1521	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	1809	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1522	believe me.	1810	believe me.
1523		1811
1524	=back	1812	=back
1525		1813
		1814	=head3 Examples
		1815
1526	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the	1816	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1527	callback, free it. Also, use no error checking, as usual.	1817	callback, free it. Also, use no error checking, as usual.
1528		1818
1529	static void	1819	static void
1530	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)	1820	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
1531	{	1821	{
1532	free (w);	1822	free (w);
1533	// now do something you wanted to do when the program has	1823	// now do something you wanted to do when the program has
1534	// no longer asnything immediate to do.	1824	// no longer anything immediate to do.
1535	}	1825	}
1536		1826
1537	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));	1827	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
1538	ev_idle_init (idle_watcher, idle_cb);	1828	ev_idle_init (idle_watcher, idle_cb);
1539	ev_idle_start (loop, idle_cb);	1829	ev_idle_start (loop, idle_cb);
…		…
1581		1871
1582	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	1872	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1583	priority, to ensure that they are being run before any other watchers	1873	priority, to ensure that they are being run before any other watchers
1584	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,	1874	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
1585	too) should not activate ("feed") events into libev. While libev fully	1875	too) should not activate ("feed") events into libev. While libev fully
1586	supports this, they will be called before other C<ev_check> watchers did	1876	supports this, they might get executed before other C<ev_check> watchers
1587	their job. As C<ev_check> watchers are often used to embed other event	1877	did their job. As C<ev_check> watchers are often used to embed other
1588	loops those other event loops might be in an unusable state until their	1878	(non-libev) event loops those other event loops might be in an unusable
1589	C<ev_check> watcher ran (always remind yourself to coexist peacefully with	1879	state until their C<ev_check> watcher ran (always remind yourself to
1590	others).	1880	coexist peacefully with others).
1591		1881
1592	=head3 Watcher-Specific Functions and Data Members	1882	=head3 Watcher-Specific Functions and Data Members
1593		1883
1594	=over 4	1884	=over 4
1595		1885
…		…
1601	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	1891	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1602	macros, but using them is utterly, utterly and completely pointless.	1892	macros, but using them is utterly, utterly and completely pointless.
1603		1893
1604	=back	1894	=back
1605		1895
		1896	=head3 Examples
		1897
1606	There are a number of principal ways to embed other event loops or modules	1898	There are a number of principal ways to embed other event loops or modules
1607	into libev. Here are some ideas on how to include libadns into libev	1899	into libev. Here are some ideas on how to include libadns into libev
1608	(there is a Perl module named C<EV::ADNS> that does this, which you could	1900	(there is a Perl module named C<EV::ADNS> that does this, which you could
1609	use for an actually working example. Another Perl module named C<EV::Glib>	1901	use as a working example. Another Perl module named C<EV::Glib> embeds a
1610	embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV	1902	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
1611	into the Glib event loop).	1903	Glib event loop).
1612		1904
1613	Method 1: Add IO watchers and a timeout watcher in a prepare handler,	1905	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1614	and in a check watcher, destroy them and call into libadns. What follows	1906	and in a check watcher, destroy them and call into libadns. What follows
1615	is pseudo-code only of course. This requires you to either use a low	1907	is pseudo-code only of course. This requires you to either use a low
1616	priority for the check watcher or use C<ev_clear_pending> explicitly, as	1908	priority for the check watcher or use C<ev_clear_pending> explicitly, as
…		…
1734	=head2 C<ev_embed> - when one backend isn't enough...	2026	=head2 C<ev_embed> - when one backend isn't enough...
1735		2027
1736	This is a rather advanced watcher type that lets you embed one event loop	2028	This is a rather advanced watcher type that lets you embed one event loop
1737	into another (currently only C<ev_io> events are supported in the embedded	2029	into another (currently only C<ev_io> events are supported in the embedded
1738	loop, other types of watchers might be handled in a delayed or incorrect	2030	loop, other types of watchers might be handled in a delayed or incorrect
1739	fashion and must not be used). (See portability notes, below).	2031	fashion and must not be used).
1740		2032
1741	There are primarily two reasons you would want that: work around bugs and	2033	There are primarily two reasons you would want that: work around bugs and
1742	prioritise I/O.	2034	prioritise I/O.
1743		2035
1744	As an example for a bug workaround, the kqueue backend might only support	2036	As an example for a bug workaround, the kqueue backend might only support
…		…
1778	portable one.	2070	portable one.
1779		2071
1780	So when you want to use this feature you will always have to be prepared	2072	So when you want to use this feature you will always have to be prepared
1781	that you cannot get an embeddable loop. The recommended way to get around	2073	that you cannot get an embeddable loop. The recommended way to get around
1782	this is to have a separate variables for your embeddable loop, try to	2074	this is to have a separate variables for your embeddable loop, try to
1783	create it, and if that fails, use the normal loop for everything:	2075	create it, and if that fails, use the normal loop for everything.
		2076
		2077	=head3 Watcher-Specific Functions and Data Members
		2078
		2079	=over 4
		2080
		2081	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2082
		2083	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2084
		2085	Configures the watcher to embed the given loop, which must be
		2086	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2087	invoked automatically, otherwise it is the responsibility of the callback
		2088	to invoke it (it will continue to be called until the sweep has been done,
		2089	if you do not want thta, you need to temporarily stop the embed watcher).
		2090
		2091	=item ev_embed_sweep (loop, ev_embed *)
		2092
		2093	Make a single, non-blocking sweep over the embedded loop. This works
		2094	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2095	apropriate way for embedded loops.
		2096
		2097	=item struct ev_loop *other [read-only]
		2098
		2099	The embedded event loop.
		2100
		2101	=back
		2102
		2103	=head3 Examples
		2104
		2105	Example: Try to get an embeddable event loop and embed it into the default
		2106	event loop. If that is not possible, use the default loop. The default
		2107	loop is stored in C<loop_hi>, while the mebeddable loop is stored in
		2108	C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
		2109	used).
1784		2110
1785	struct ev_loop *loop_hi = ev_default_init (0);	2111	struct ev_loop *loop_hi = ev_default_init (0);
1786	struct ev_loop *loop_lo = 0;	2112	struct ev_loop *loop_lo = 0;
1787	struct ev_embed embed;	2113	struct ev_embed embed;
1788		2114
…		…
1799	ev_embed_start (loop_hi, &embed);	2125	ev_embed_start (loop_hi, &embed);
1800	}	2126	}
1801	else	2127	else
1802	loop_lo = loop_hi;	2128	loop_lo = loop_hi;
1803		2129
1804	=head2 Portability notes	2130	Example: Check if kqueue is available but not recommended and create
		2131	a kqueue backend for use with sockets (which usually work with any
		2132	kqueue implementation). Store the kqueue/socket-only event loop in
		2133	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
1805		2134
1806	Kqueue is nominally embeddable, but this is broken on all BSDs that I	2135	struct ev_loop *loop = ev_default_init (0);
1807	tried, in various ways. Usually the embedded event loop will simply never	2136	struct ev_loop *loop_socket = 0;
1808	receive events, sometimes it will only trigger a few times, sometimes in a	2137	struct ev_embed embed;
1809	loop. Epoll is also nominally embeddable, but many Linux kernel versions	2138
1810	will always eport the epoll fd as ready, even when no events are pending.	2139	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2140	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2141	{
		2142	ev_embed_init (&embed, 0, loop_socket);
		2143	ev_embed_start (loop, &embed);
		2144	}
1811		2145
1812	While libev allows embedding these backends (they are contained in	2146	if (!loop_socket)
1813	C<ev_embeddable_backends ()>), take extreme care that it will actually	2147	loop_socket = loop;
1814	work.
1815		2148
1816	When in doubt, create a dynamic event loop forced to use sockets (this	2149	// now use loop_socket for all sockets, and loop for everything else
1817	usually works) and possibly another thread and a pipe or so to report to
1818	your main event loop.
1819
1820	=head3 Watcher-Specific Functions and Data Members
1821
1822	=over 4
1823
1824	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1825
1826	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
1827
1828	Configures the watcher to embed the given loop, which must be
1829	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1830	invoked automatically, otherwise it is the responsibility of the callback
1831	to invoke it (it will continue to be called until the sweep has been done,
1832	if you do not want thta, you need to temporarily stop the embed watcher).
1833
1834	=item ev_embed_sweep (loop, ev_embed *)
1835
1836	Make a single, non-blocking sweep over the embedded loop. This works
1837	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
1838	apropriate way for embedded loops.
1839
1840	=item struct ev_loop *other [read-only]
1841
1842	The embedded event loop.
1843
1844	=back
1845		2150
1846		2151
1847	=head2 C<ev_fork> - the audacity to resume the event loop after a fork	2152	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
1848		2153
1849	Fork watchers are called when a C<fork ()> was detected (usually because	2154	Fork watchers are called when a C<fork ()> was detected (usually because
…		…
1865	believe me.	2170	believe me.
1866		2171
1867	=back	2172	=back
1868		2173
1869		2174
		2175	=head2 C<ev_async> - how to wake up another event loop
		2176
		2177	In general, you cannot use an C<ev_loop> from multiple threads or other
		2178	asynchronous sources such as signal handlers (as opposed to multiple event
		2179	loops - those are of course safe to use in different threads).
		2180
		2181	Sometimes, however, you need to wake up another event loop you do not
		2182	control, for example because it belongs to another thread. This is what
		2183	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2184	can signal it by calling C<ev_async_send>, which is thread- and signal
		2185	safe.
		2186
		2187	This functionality is very similar to C<ev_signal> watchers, as signals,
		2188	too, are asynchronous in nature, and signals, too, will be compressed
		2189	(i.e. the number of callback invocations may be less than the number of
		2190	C<ev_async_sent> calls).
		2191
		2192	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2193	just the default loop.
		2194
		2195	=head3 Queueing
		2196
		2197	C<ev_async> does not support queueing of data in any way. The reason
		2198	is that the author does not know of a simple (or any) algorithm for a
		2199	multiple-writer-single-reader queue that works in all cases and doesn't
		2200	need elaborate support such as pthreads.
		2201
		2202	That means that if you want to queue data, you have to provide your own
		2203	queue. But at least I can tell you would implement locking around your
		2204	queue:
		2205
		2206	=over 4
		2207
		2208	=item queueing from a signal handler context
		2209
		2210	To implement race-free queueing, you simply add to the queue in the signal
		2211	handler but you block the signal handler in the watcher callback. Here is an example that does that for
		2212	some fictitiuous SIGUSR1 handler:
		2213
		2214	static ev_async mysig;
		2215
		2216	static void
		2217	sigusr1_handler (void)
		2218	{
		2219	sometype data;
		2220
		2221	// no locking etc.
		2222	queue_put (data);
		2223	ev_async_send (EV_DEFAULT_ &mysig);
		2224	}
		2225
		2226	static void
		2227	mysig_cb (EV_P_ ev_async *w, int revents)
		2228	{
		2229	sometype data;
		2230	sigset_t block, prev;
		2231
		2232	sigemptyset (&block);
		2233	sigaddset (&block, SIGUSR1);
		2234	sigprocmask (SIG_BLOCK, &block, &prev);
		2235
		2236	while (queue_get (&data))
		2237	process (data);
		2238
		2239	if (sigismember (&prev, SIGUSR1)
		2240	sigprocmask (SIG_UNBLOCK, &block, 0);
		2241	}
		2242
		2243	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2244	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2245	either...).
		2246
		2247	=item queueing from a thread context
		2248
		2249	The strategy for threads is different, as you cannot (easily) block
		2250	threads but you can easily preempt them, so to queue safely you need to
		2251	employ a traditional mutex lock, such as in this pthread example:
		2252
		2253	static ev_async mysig;
		2254	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2255
		2256	static void
		2257	otherthread (void)
		2258	{
		2259	// only need to lock the actual queueing operation
		2260	pthread_mutex_lock (&mymutex);
		2261	queue_put (data);
		2262	pthread_mutex_unlock (&mymutex);
		2263
		2264	ev_async_send (EV_DEFAULT_ &mysig);
		2265	}
		2266
		2267	static void
		2268	mysig_cb (EV_P_ ev_async *w, int revents)
		2269	{
		2270	pthread_mutex_lock (&mymutex);
		2271
		2272	while (queue_get (&data))
		2273	process (data);
		2274
		2275	pthread_mutex_unlock (&mymutex);
		2276	}
		2277
		2278	=back
		2279
		2280
		2281	=head3 Watcher-Specific Functions and Data Members
		2282
		2283	=over 4
		2284
		2285	=item ev_async_init (ev_async *, callback)
		2286
		2287	Initialises and configures the async watcher - it has no parameters of any
		2288	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
		2289	believe me.
		2290
		2291	=item ev_async_send (loop, ev_async *)
		2292
		2293	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2294	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2295	C<ev_feed_event>, this call is safe to do in other threads, signal or
		2296	similar contexts (see the dicusssion of C<EV_ATOMIC_T> in the embedding
		2297	section below on what exactly this means).
		2298
		2299	This call incurs the overhead of a syscall only once per loop iteration,
		2300	so while the overhead might be noticable, it doesn't apply to repeated
		2301	calls to C<ev_async_send>.
		2302
		2303	=item bool = ev_async_pending (ev_async *)
		2304
		2305	Returns a non-zero value when C<ev_async_send> has been called on the
		2306	watcher but the event has not yet been processed (or even noted) by the
		2307	event loop.
		2308
		2309	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2310	the loop iterates next and checks for the watcher to have become active,
		2311	it will reset the flag again. C<ev_async_pending> can be used to very
		2312	quickly check wether invoking the loop might be a good idea.
		2313
		2314	Not that this does I<not> check wether the watcher itself is pending, only
		2315	wether it has been requested to make this watcher pending.
		2316
		2317	=back
		2318
		2319
1870	=head1 OTHER FUNCTIONS	2320	=head1 OTHER FUNCTIONS
1871		2321
1872	There are some other functions of possible interest. Described. Here. Now.	2322	There are some other functions of possible interest. Described. Here. Now.
1873		2323
1874	=over 4	2324	=over 4
…		…
1942		2392
1943	=item * Priorities are not currently supported. Initialising priorities	2393	=item * Priorities are not currently supported. Initialising priorities
1944	will fail and all watchers will have the same priority, even though there	2394	will fail and all watchers will have the same priority, even though there
1945	is an ev_pri field.	2395	is an ev_pri field.
1946		2396
		2397	=item * In libevent, the last base created gets the signals, in libev, the
		2398	first base created (== the default loop) gets the signals.
		2399
1947	=item * Other members are not supported.	2400	=item * Other members are not supported.
1948		2401
1949	=item * The libev emulation is I<not> ABI compatible to libevent, you need	2402	=item * The libev emulation is I<not> ABI compatible to libevent, you need
1950	to use the libev header file and library.	2403	to use the libev header file and library.
1951		2404
…		…
2101	Example: Define a class with an IO and idle watcher, start one of them in	2554	Example: Define a class with an IO and idle watcher, start one of them in
2102	the constructor.	2555	the constructor.
2103		2556
2104	class myclass	2557	class myclass
2105	{	2558	{
2106	ev_io io; void io_cb (ev::io &w, int revents);	2559	ev::io io; void io_cb (ev::io &w, int revents);
2107	ev_idle idle void idle_cb (ev::idle &w, int revents);	2560	ev:idle idle void idle_cb (ev::idle &w, int revents);
2108		2561
2109	myclass ();	2562	myclass (int fd)
2110	}
2111
2112	myclass::myclass (int fd)
2113	{	2563	{
2114	io .set <myclass, &myclass::io_cb > (this);	2564	io .set <myclass, &myclass::io_cb > (this);
2115	idle.set <myclass, &myclass::idle_cb> (this);	2565	idle.set <myclass, &myclass::idle_cb> (this);
2116		2566
2117	io.start (fd, ev::READ);	2567	io.start (fd, ev::READ);
		2568	}
2118	}	2569	};
		2570
		2571
		2572	=head1 OTHER LANGUAGE BINDINGS
		2573
		2574	Libev does not offer other language bindings itself, but bindings for a
		2575	numbe rof languages exist in the form of third-party packages. If you know
		2576	any interesting language binding in addition to the ones listed here, drop
		2577	me a note.
		2578
		2579	=over 4
		2580
		2581	=item Perl
		2582
		2583	The EV module implements the full libev API and is actually used to test
		2584	libev. EV is developed together with libev. Apart from the EV core module,
		2585	there are additional modules that implement libev-compatible interfaces
		2586	to C<libadns> (C<EV::ADNS>), C<Net::SNMP> (C<Net::SNMP::EV>) and the
		2587	C<libglib> event core (C<Glib::EV> and C<EV::Glib>).
		2588
		2589	It can be found and installed via CPAN, its homepage is found at
		2590	L<http://software.schmorp.de/pkg/EV>.
		2591
		2592	=item Ruby
		2593
		2594	Tony Arcieri has written a ruby extension that offers access to a subset
		2595	of the libev API and adds filehandle abstractions, asynchronous DNS and
		2596	more on top of it. It can be found via gem servers. Its homepage is at
		2597	L<http://rev.rubyforge.org/>.
		2598
		2599	=item D
		2600
		2601	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		2602	be found at L<http://git.llucax.com.ar/?p=software/ev.d.git;a=summary>.
		2603
		2604	=back
2119		2605
2120		2606
2121	=head1 MACRO MAGIC	2607	=head1 MACRO MAGIC
2122		2608
2123	Libev can be compiled with a variety of options, the most fundamantal	2609	Libev can be compiled with a variety of options, the most fundamantal
…		…
2159		2645
2160	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	2646	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
2161		2647
2162	Similar to the other two macros, this gives you the value of the default	2648	Similar to the other two macros, this gives you the value of the default
2163	loop, if multiple loops are supported ("ev loop default").	2649	loop, if multiple loops are supported ("ev loop default").
		2650
		2651	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		2652
		2653	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		2654	default loop has been initialised (C<UC> == unchecked). Their behaviour
		2655	is undefined when the default loop has not been initialised by a previous
		2656	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		2657
		2658	It is often prudent to use C<EV_DEFAULT> when initialising the first
		2659	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
2164		2660
2165	=back	2661	=back
2166		2662
2167	Example: Declare and initialise a check watcher, utilising the above	2663	Example: Declare and initialise a check watcher, utilising the above
2168	macros so it will work regardless of whether multiple loops are supported	2664	macros so it will work regardless of whether multiple loops are supported
…		…
2264		2760
2265	libev.m4	2761	libev.m4
2266		2762
2267	=head2 PREPROCESSOR SYMBOLS/MACROS	2763	=head2 PREPROCESSOR SYMBOLS/MACROS
2268		2764
2269	Libev can be configured via a variety of preprocessor symbols you have to define	2765	Libev can be configured via a variety of preprocessor symbols you have to
2270	before including any of its files. The default is not to build for multiplicity	2766	define before including any of its files. The default in the absense of
2271	and only include the select backend.	2767	autoconf is noted for every option.
2272		2768
2273	=over 4	2769	=over 4
2274		2770
2275	=item EV_STANDALONE	2771	=item EV_STANDALONE
2276		2772
…		…
2297	runtime if successful). Otherwise no use of the realtime clock option will	2793	runtime if successful). Otherwise no use of the realtime clock option will
2298	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	2794	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
2299	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the	2795	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
2300	note about libraries in the description of C<EV_USE_MONOTONIC>, though.	2796	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
2301		2797
		2798	=item EV_USE_NANOSLEEP
		2799
		2800	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		2801	and will use it for delays. Otherwise it will use C<select ()>.
		2802
		2803	=item EV_USE_EVENTFD
		2804
		2805	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		2806	available and will probe for kernel support at runtime. This will improve
		2807	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		2808	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		2809	2.7 or newer, otherwise disabled.
		2810
2302	=item EV_USE_SELECT	2811	=item EV_USE_SELECT
2303		2812
2304	If undefined or defined to be C<1>, libev will compile in support for the	2813	If undefined or defined to be C<1>, libev will compile in support for the
2305	C<select>(2) backend. No attempt at autodetection will be done: if no	2814	C<select>(2) backend. No attempt at autodetection will be done: if no
2306	other method takes over, select will be it. Otherwise the select backend	2815	other method takes over, select will be it. Otherwise the select backend
…		…
2324	be used is the winsock select). This means that it will call	2833	be used is the winsock select). This means that it will call
2325	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,	2834	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2326	it is assumed that all these functions actually work on fds, even	2835	it is assumed that all these functions actually work on fds, even
2327	on win32. Should not be defined on non-win32 platforms.	2836	on win32. Should not be defined on non-win32 platforms.
2328		2837
		2838	=item EV_FD_TO_WIN32_HANDLE
		2839
		2840	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		2841	file descriptors to socket handles. When not defining this symbol (the
		2842	default), then libev will call C<_get_osfhandle>, which is usually
		2843	correct. In some cases, programs use their own file descriptor management,
		2844	in which case they can provide this function to map fds to socket handles.
		2845
2329	=item EV_USE_POLL	2846	=item EV_USE_POLL
2330		2847
2331	If defined to be C<1>, libev will compile in support for the C<poll>(2)	2848	If defined to be C<1>, libev will compile in support for the C<poll>(2)
2332	backend. Otherwise it will be enabled on non-win32 platforms. It	2849	backend. Otherwise it will be enabled on non-win32 platforms. It
2333	takes precedence over select.	2850	takes precedence over select.
2334		2851
2335	=item EV_USE_EPOLL	2852	=item EV_USE_EPOLL
2336		2853
2337	If defined to be C<1>, libev will compile in support for the Linux	2854	If defined to be C<1>, libev will compile in support for the Linux
2338	C<epoll>(7) backend. Its availability will be detected at runtime,	2855	C<epoll>(7) backend. Its availability will be detected at runtime,
2339	otherwise another method will be used as fallback. This is the	2856	otherwise another method will be used as fallback. This is the preferred
2340	preferred backend for GNU/Linux systems.	2857	backend for GNU/Linux systems. If undefined, it will be enabled if the
		2858	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2341		2859
2342	=item EV_USE_KQUEUE	2860	=item EV_USE_KQUEUE
2343		2861
2344	If defined to be C<1>, libev will compile in support for the BSD style	2862	If defined to be C<1>, libev will compile in support for the BSD style
2345	C<kqueue>(2) backend. Its actual availability will be detected at runtime,	2863	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
…		…
2364		2882
2365	=item EV_USE_INOTIFY	2883	=item EV_USE_INOTIFY
2366		2884
2367	If defined to be C<1>, libev will compile in support for the Linux inotify	2885	If defined to be C<1>, libev will compile in support for the Linux inotify
2368	interface to speed up C<ev_stat> watchers. Its actual availability will	2886	interface to speed up C<ev_stat> watchers. Its actual availability will
2369	be detected at runtime.	2887	be detected at runtime. If undefined, it will be enabled if the headers
		2888	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		2889
		2890	=item EV_ATOMIC_T
		2891
		2892	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		2893	access is atomic with respect to other threads or signal contexts. No such
		2894	type is easily found in the C language, so you can provide your own type
		2895	that you know is safe for your purposes. It is used both for signal handler "locking"
		2896	as well as for signal and thread safety in C<ev_async> watchers.
		2897
		2898	In the absense of this define, libev will use C<sig_atomic_t volatile>
		2899	(from F<signal.h>), which is usually good enough on most platforms.
2370		2900
2371	=item EV_H	2901	=item EV_H
2372		2902
2373	The name of the F<ev.h> header file used to include it. The default if	2903	The name of the F<ev.h> header file used to include it. The default if
2374	undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This	2904	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2375	can be used to virtually rename the F<ev.h> header file in case of conflicts.	2905	used to virtually rename the F<ev.h> header file in case of conflicts.
2376		2906
2377	=item EV_CONFIG_H	2907	=item EV_CONFIG_H
2378		2908
2379	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override	2909	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2380	F<ev.c>'s idea of where to find the F<config.h> file, similarly to	2910	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2381	C<EV_H>, above.	2911	C<EV_H>, above.
2382		2912
2383	=item EV_EVENT_H	2913	=item EV_EVENT_H
2384		2914
2385	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea	2915	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2386	of how the F<event.h> header can be found.	2916	of how the F<event.h> header can be found, the default is C<"event.h">.
2387		2917
2388	=item EV_PROTOTYPES	2918	=item EV_PROTOTYPES
2389		2919
2390	If defined to be C<0>, then F<ev.h> will not define any function	2920	If defined to be C<0>, then F<ev.h> will not define any function
2391	prototypes, but still define all the structs and other symbols. This is	2921	prototypes, but still define all the structs and other symbols. This is
…		…
2442	=item EV_FORK_ENABLE	2972	=item EV_FORK_ENABLE
2443		2973
2444	If undefined or defined to be C<1>, then fork watchers are supported. If	2974	If undefined or defined to be C<1>, then fork watchers are supported. If
2445	defined to be C<0>, then they are not.	2975	defined to be C<0>, then they are not.
2446		2976
		2977	=item EV_ASYNC_ENABLE
		2978
		2979	If undefined or defined to be C<1>, then async watchers are supported. If
		2980	defined to be C<0>, then they are not.
		2981
2447	=item EV_MINIMAL	2982	=item EV_MINIMAL
2448		2983
2449	If you need to shave off some kilobytes of code at the expense of some	2984	If you need to shave off some kilobytes of code at the expense of some
2450	speed, define this symbol to C<1>. Currently only used for gcc to override	2985	speed, define this symbol to C<1>. Currently this is used to override some
2451	some inlining decisions, saves roughly 30% codesize of amd64.	2986	inlining decisions, saves roughly 30% codesize of amd64. It also selects a
		2987	much smaller 2-heap for timer management over the default 4-heap.
2452		2988
2453	=item EV_PID_HASHSIZE	2989	=item EV_PID_HASHSIZE
2454		2990
2455	C<ev_child> watchers use a small hash table to distribute workload by	2991	C<ev_child> watchers use a small hash table to distribute workload by
2456	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more	2992	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
2457	than enough. If you need to manage thousands of children you might want to	2993	than enough. If you need to manage thousands of children you might want to
2458	increase this value (I<must> be a power of two).	2994	increase this value (I<must> be a power of two).
2459		2995
2460	=item EV_INOTIFY_HASHSIZE	2996	=item EV_INOTIFY_HASHSIZE
2461		2997
2462	C<ev_staz> watchers use a small hash table to distribute workload by	2998	C<ev_stat> watchers use a small hash table to distribute workload by
2463	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),	2999	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
2464	usually more than enough. If you need to manage thousands of C<ev_stat>	3000	usually more than enough. If you need to manage thousands of C<ev_stat>
2465	watchers you might want to increase this value (I<must> be a power of	3001	watchers you might want to increase this value (I<must> be a power of
2466	two).	3002	two).
		3003
		3004	=item EV_USE_4HEAP
		3005
		3006	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3007	timer and periodics heap, libev uses a 4-heap when this symbol is defined
		3008	to C<1>. The 4-heap uses more complicated (longer) code but has
		3009	noticably faster performance with many (thousands) of watchers.
		3010
		3011	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3012	(disabled).
		3013
		3014	=item EV_HEAP_CACHE_AT
		3015
		3016	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3017	timer and periodics heap, libev can cache the timestamp (I<at>) within
		3018	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3019	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3020	but avoids random read accesses on heap changes. This improves performance
		3021	noticably with with many (hundreds) of watchers.
		3022
		3023	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3024	(disabled).
2467		3025
2468	=item EV_COMMON	3026	=item EV_COMMON
2469		3027
2470	By default, all watchers have a C<void *data> member. By redefining	3028	By default, all watchers have a C<void *data> member. By redefining
2471	this macro to a something else you can include more and other types of	3029	this macro to a something else you can include more and other types of
…		…
2545		3103
2546	#include "ev_cpp.h"	3104	#include "ev_cpp.h"
2547	#include "ev.c"	3105	#include "ev.c"
2548		3106
2549		3107
		3108	=head1 THREADS AND COROUTINES
		3109
		3110	=head2 THREADS
		3111
		3112	Libev itself is completely threadsafe, but it uses no locking. This
		3113	means that you can use as many loops as you want in parallel, as long as
		3114	only one thread ever calls into one libev function with the same loop
		3115	parameter.
		3116
		3117	Or put differently: calls with different loop parameters can be done in
		3118	parallel from multiple threads, calls with the same loop parameter must be
		3119	done serially (but can be done from different threads, as long as only one
		3120	thread ever is inside a call at any point in time, e.g. by using a mutex
		3121	per loop).
		3122
		3123	If you want to know which design is best for your problem, then I cannot
		3124	help you but by giving some generic advice:
		3125
		3126	=over 4
		3127
		3128	=item * most applications have a main thread: use the default libev loop
		3129	in that thread, or create a seperate thread running only the default loop.
		3130
		3131	This helps integrating other libraries or software modules that use libev
		3132	themselves and don't care/know about threading.
		3133
		3134	=item * one loop per thread is usually a good model.
		3135
		3136	Doing this is almost never wrong, sometimes a better-performance model
		3137	exists, but it is always a good start.
		3138
		3139	=item * other models exist, such as the leader/follower pattern, where one
		3140	loop is handed through multiple threads in a kind of round-robbin fashion.
		3141
		3142	Chosing a model is hard - look around, learn, know that usually you cna do
		3143	better than you currently do :-)
		3144
		3145	=item * often you need to talk to some other thread which blocks in the
		3146	event loop - C<ev_async> watchers can be used to wake them up from other
		3147	threads safely (or from signal contexts...).
		3148
		3149	=back
		3150
		3151	=head2 COROUTINES
		3152
		3153	Libev is much more accomodating to coroutines ("cooperative threads"):
		3154	libev fully supports nesting calls to it's functions from different
		3155	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3156	different coroutines and switch freely between both coroutines running the
		3157	loop, as long as you don't confuse yourself). The only exception is that
		3158	you must not do this from C<ev_periodic> reschedule callbacks.
		3159
		3160	Care has been invested into making sure that libev does not keep local
		3161	state inside C<ev_loop>, and other calls do not usually allow coroutine
		3162	switches.
		3163
		3164
2550	=head1 COMPLEXITIES	3165	=head1 COMPLEXITIES
2551		3166
2552	In this section the complexities of (many of) the algorithms used inside	3167	In this section the complexities of (many of) the algorithms used inside
2553	libev will be explained. For complexity discussions about backends see the	3168	libev will be explained. For complexity discussions about backends see the
2554	documentation for C<ev_default_init>.	3169	documentation for C<ev_default_init>.
…		…
2563		3178
2564	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	3179	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2565		3180
2566	This means that, when you have a watcher that triggers in one hour and	3181	This means that, when you have a watcher that triggers in one hour and
2567	there are 100 watchers that would trigger before that then inserting will	3182	there are 100 watchers that would trigger before that then inserting will
2568	have to skip those 100 watchers.	3183	have to skip roughly seven (C<ld 100>) of these watchers.
2569		3184
2570	=item Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)	3185	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2571		3186
2572	That means that for changing a timer costs less than removing/adding them	3187	That means that changing a timer costs less than removing/adding them
2573	as only the relative motion in the event queue has to be paid for.	3188	as only the relative motion in the event queue has to be paid for.
2574		3189
2575	=item Starting io/check/prepare/idle/signal/child watchers: O(1)	3190	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
2576		3191
2577	These just add the watcher into an array or at the head of a list.	3192	These just add the watcher into an array or at the head of a list.
		3193
2578	=item Stopping check/prepare/idle watchers: O(1)	3194	=item Stopping check/prepare/idle/fork/async watchers: O(1)
2579		3195
2580	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	3196	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2581		3197
2582	These watchers are stored in lists then need to be walked to find the	3198	These watchers are stored in lists then need to be walked to find the
2583	correct watcher to remove. The lists are usually short (you don't usually	3199	correct watcher to remove. The lists are usually short (you don't usually
2584	have many watchers waiting for the same fd or signal).	3200	have many watchers waiting for the same fd or signal).
2585		3201
2586	=item Finding the next timer per loop iteration: O(1)	3202	=item Finding the next timer in each loop iteration: O(1)
		3203
		3204	By virtue of using a binary or 4-heap, the next timer is always found at a
		3205	fixed position in the storage array.
2587		3206
2588	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)	3207	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2589		3208
2590	A change means an I/O watcher gets started or stopped, which requires	3209	A change means an I/O watcher gets started or stopped, which requires
2591	libev to recalculate its status (and possibly tell the kernel).	3210	libev to recalculate its status (and possibly tell the kernel, depending
		3211	on backend and wether C<ev_io_set> was used).
2592		3212
2593	=item Activating one watcher: O(1)	3213	=item Activating one watcher (putting it into the pending state): O(1)
2594		3214
2595	=item Priority handling: O(number_of_priorities)	3215	=item Priority handling: O(number_of_priorities)
2596		3216
2597	Priorities are implemented by allocating some space for each	3217	Priorities are implemented by allocating some space for each
2598	priority. When doing priority-based operations, libev usually has to	3218	priority. When doing priority-based operations, libev usually has to
2599	linearly search all the priorities.	3219	linearly search all the priorities, but starting/stopping and activating
		3220	watchers becomes O(1) w.r.t. priority handling.
		3221
		3222	=item Sending an ev_async: O(1)
		3223
		3224	=item Processing ev_async_send: O(number_of_async_watchers)
		3225
		3226	=item Processing signals: O(max_signal_number)
		3227
		3228	Sending involves a syscall I<iff> there were no other C<ev_async_send>
		3229	calls in the current loop iteration. Checking for async and signal events
		3230	involves iterating over all running async watchers or all signal numbers.
2600		3231
2601	=back	3232	=back
2602		3233
2603		3234
		3235	=head1 Win32 platform limitations and workarounds
		3236
		3237	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3238	requires, and its I/O model is fundamentally incompatible with the POSIX
		3239	model. Libev still offers limited functionality on this platform in
		3240	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3241	descriptors. This only applies when using Win32 natively, not when using
		3242	e.g. cygwin.
		3243
		3244	Lifting these limitations would basically require the full
		3245	re-implementation of the I/O system. If you are into these kinds of
		3246	things, then note that glib does exactly that for you in a very portable
		3247	way (note also that glib is the slowest event library known to man).
		3248
		3249	There is no supported compilation method available on windows except
		3250	embedding it into other applications.
		3251
		3252	Due to the many, low, and arbitrary limits on the win32 platform and
		3253	the abysmal performance of winsockets, using a large number of sockets
		3254	is not recommended (and not reasonable). If your program needs to use
		3255	more than a hundred or so sockets, then likely it needs to use a totally
		3256	different implementation for windows, as libev offers the POSIX readiness
		3257	notification model, which cannot be implemented efficiently on windows
		3258	(microsoft monopoly games).
		3259
		3260	=over 4
		3261
		3262	=item The winsocket select function
		3263
		3264	The winsocket C<select> function doesn't follow POSIX in that it requires
		3265	socket I<handles> and not socket I<file descriptors>. This makes select
		3266	very inefficient, and also requires a mapping from file descriptors
		3267	to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
		3268	C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
		3269	symbols for more info.
		3270
		3271	The configuration for a "naked" win32 using the microsoft runtime
		3272	libraries and raw winsocket select is:
		3273
		3274	#define EV_USE_SELECT 1
		3275	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3276
		3277	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3278	complexity in the O(n²) range when using win32.
		3279
		3280	=item Limited number of file descriptors
		3281
		3282	Windows has numerous arbitrary (and low) limits on things.
		3283
		3284	Early versions of winsocket's select only supported waiting for a maximum
		3285	of C<64> handles (probably owning to the fact that all windows kernels
		3286	can only wait for C<64> things at the same time internally; microsoft
		3287	recommends spawning a chain of threads and wait for 63 handles and the
		3288	previous thread in each. Great).
		3289
		3290	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3291	to some high number (e.g. C<2048>) before compiling the winsocket select
		3292	call (which might be in libev or elsewhere, for example, perl does its own
		3293	select emulation on windows).
		3294
		3295	Another limit is the number of file descriptors in the microsoft runtime
		3296	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3297	or something like this inside microsoft). You can increase this by calling
		3298	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3299	arbitrary limit), but is broken in many versions of the microsoft runtime
		3300	libraries.
		3301
		3302	This might get you to about C<512> or C<2048> sockets (depending on
		3303	windows version and/or the phase of the moon). To get more, you need to
		3304	wrap all I/O functions and provide your own fd management, but the cost of
		3305	calling select (O(n²)) will likely make this unworkable.
		3306
		3307	=back
		3308
		3309
		3310	=head1 PORTABILITY REQUIREMENTS
		3311
		3312	In addition to a working ISO-C implementation, libev relies on a few
		3313	additional extensions:
		3314
		3315	=over 4
		3316
		3317	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3318
		3319	The type C<sig_atomic_t volatile> (or whatever is defined as
		3320	C<EV_ATOMIC_T>) must be atomic w.r.t. accesses from different
		3321	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3322	believed to be sufficiently portable.
		3323
		3324	=item C<sigprocmask> must work in a threaded environment
		3325
		3326	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3327	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3328	pthread implementations will either allow C<sigprocmask> in the "main
		3329	thread" or will block signals process-wide, both behaviours would
		3330	be compatible with libev. Interaction between C<sigprocmask> and
		3331	C<pthread_sigmask> could complicate things, however.
		3332
		3333	The most portable way to handle signals is to block signals in all threads
		3334	except the initial one, and run the default loop in the initial thread as
		3335	well.
		3336
		3337	=item C<long> must be large enough for common memory allocation sizes
		3338
		3339	To improve portability and simplify using libev, libev uses C<long>
		3340	internally instead of C<size_t> when allocating its data structures. On
		3341	non-POSIX systems (Microsoft...) this might be unexpectedly low, but
		3342	is still at least 31 bits everywhere, which is enough for hundreds of
		3343	millions of watchers.
		3344
		3345	=item C<double> must hold a time value in seconds with enough accuracy
		3346
		3347	The type C<double> is used to represent timestamps. It is required to
		3348	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3349	enough for at least into the year 4000. This requirement is fulfilled by
		3350	implementations implementing IEEE 754 (basically all existing ones).
		3351
		3352	=back
		3353
		3354	If you know of other additional requirements drop me a note.
		3355
		3356
2604	=head1 AUTHOR	3357	=head1 AUTHOR
2605		3358
2606	Marc Lehmann <libev@schmorp.de>.	3359	Marc Lehmann <libev@schmorp.de>.
2607		3360

Diff Legend

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