ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.176 by root, Mon Sep 8 17:24:39 2008 UTC vs.
Revision 1.370 by root, Thu Jun 2 23:42:40 2011 UTC

…		…
9	=head2 EXAMPLE PROGRAM	9	=head2 EXAMPLE PROGRAM
10		10
11	// a single header file is required	11	// a single header file is required
12	#include <ev.h>	12	#include <ev.h>
13		13
		14	#include <stdio.h> // for puts
		15
14	// every watcher type has its own typedef'd struct	16	// every watcher type has its own typedef'd struct
15	// with the name ev_<type>	17	// with the name ev_TYPE
16	ev_io stdin_watcher;	18	ev_io stdin_watcher;
17	ev_timer timeout_watcher;	19	ev_timer timeout_watcher;
18		20
19	// all watcher callbacks have a similar signature	21	// all watcher callbacks have a similar signature
20	// this callback is called when data is readable on stdin	22	// this callback is called when data is readable on stdin
21	static void	23	static void
22	stdin_cb (EV_P_ struct ev_io *w, int revents)	24	stdin_cb (EV_P_ ev_io *w, int revents)
23	{	25	{
24	puts ("stdin ready");	26	puts ("stdin ready");
25	// for one-shot events, one must manually stop the watcher	27	// for one-shot events, one must manually stop the watcher
26	// with its corresponding stop function.	28	// with its corresponding stop function.
27	ev_io_stop (EV_A_ w);	29	ev_io_stop (EV_A_ w);
28		30
29	// this causes all nested ev_loop's to stop iterating	31	// this causes all nested ev_run's to stop iterating
30	ev_unloop (EV_A_ EVUNLOOP_ALL);	32	ev_break (EV_A_ EVBREAK_ALL);
31	}	33	}
32		34
33	// another callback, this time for a time-out	35	// another callback, this time for a time-out
34	static void	36	static void
35	timeout_cb (EV_P_ struct ev_timer *w, int revents)	37	timeout_cb (EV_P_ ev_timer *w, int revents)
36	{	38	{
37	puts ("timeout");	39	puts ("timeout");
38	// this causes the innermost ev_loop to stop iterating	40	// this causes the innermost ev_run to stop iterating
39	ev_unloop (EV_A_ EVUNLOOP_ONE);	41	ev_break (EV_A_ EVBREAK_ONE);
40	}	42	}
41		43
42	int	44	int
43	main (void)	45	main (void)
44	{	46	{
45	// use the default event loop unless you have special needs	47	// use the default event loop unless you have special needs
46	struct ev_loop *loop = ev_default_loop (0);	48	struct ev_loop *loop = EV_DEFAULT;
47		49
48	// initialise an io watcher, then start it	50	// initialise an io watcher, then start it
49	// this one will watch for stdin to become readable	51	// this one will watch for stdin to become readable
50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);	52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
51	ev_io_start (loop, &stdin_watcher);	53	ev_io_start (loop, &stdin_watcher);
…		…
54	// simple non-repeating 5.5 second timeout	56	// simple non-repeating 5.5 second timeout
55	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);	57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
56	ev_timer_start (loop, &timeout_watcher);	58	ev_timer_start (loop, &timeout_watcher);
57		59
58	// now wait for events to arrive	60	// now wait for events to arrive
59	ev_loop (loop, 0);	61	ev_run (loop, 0);
60		62
61	// unloop was called, so exit	63	// break was called, so exit
62	return 0;	64	return 0;
63	}	65	}
64		66
65	=head1 DESCRIPTION	67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
66		70
67	The newest version of this document is also available as an html-formatted	71	The newest version of this document is also available as an html-formatted
68	web page you might find easier to navigate when reading it for the first	72	web page you might find easier to navigate when reading it for the first
69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.	73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familiarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 WHAT TO READ WHEN IN A HURRY
		84
		85	This manual tries to be very detailed, but unfortunately, this also makes
		86	it very long. If you just want to know the basics of libev, I suggest
		87	reading L<ANATOMY OF A WATCHER>, then the L<EXAMPLE PROGRAM> above and
		88	look up the missing functions in L<GLOBAL FUNCTIONS> and the C<ev_io> and
		89	C<ev_timer> sections in L<WATCHER TYPES>.
		90
		91	=head1 ABOUT LIBEV
70		92
71	Libev is an event loop: you register interest in certain events (such as a	93	Libev is an event loop: you register interest in certain events (such as a
72	file descriptor being readable or a timeout occurring), and it will manage	94	file descriptor being readable or a timeout occurring), and it will manage
73	these event sources and provide your program with events.	95	these event sources and provide your program with events.
74		96
…		…
84	=head2 FEATURES	106	=head2 FEATURES
85		107
86	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the	108	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
87	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms	109	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
88	for file descriptor events (C<ev_io>), the Linux C<inotify> interface	110	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
89	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers	111	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
90	with customised rescheduling (C<ev_periodic>), synchronous signals	112	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
91	(C<ev_signal>), process status change events (C<ev_child>), and event	113	timers (C<ev_timer>), absolute timers with customised rescheduling
92	watchers dealing with the event loop mechanism itself (C<ev_idle>,	114	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
93	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as	115	change events (C<ev_child>), and event watchers dealing with the event
94	file watchers (C<ev_stat>) and even limited support for fork events	116	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
95	(C<ev_fork>).	117	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		118	limited support for fork events (C<ev_fork>).
96		119
97	It also is quite fast (see this	120	It also is quite fast (see this
98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent	121	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
99	for example).	122	for example).
100		123
…		…
108	name C<loop> (which is always of type C<struct ev_loop *>) will not have	131	name C<loop> (which is always of type C<struct ev_loop *>) will not have
109	this argument.	132	this argument.
110		133
111	=head2 TIME REPRESENTATION	134	=head2 TIME REPRESENTATION
112		135
113	Libev represents time as a single floating point number, representing the	136	Libev represents time as a single floating point number, representing
114	(fractional) number of seconds since the (POSIX) epoch (somewhere near	137	the (fractional) number of seconds since the (POSIX) epoch (in practice
115	the beginning of 1970, details are complicated, don't ask). This type is	138	somewhere near the beginning of 1970, details are complicated, don't
116	called C<ev_tstamp>, which is what you should use too. It usually aliases	139	ask). This type is called C<ev_tstamp>, which is what you should use
117	to the C<double> type in C, and when you need to do any calculations on	140	too. It usually aliases to the C<double> type in C. When you need to do
118	it, you should treat it as some floating point value. Unlike the name	141	any calculations on it, you should treat it as some floating point value.
		142
119	component C<stamp> might indicate, it is also used for time differences	143	Unlike the name component C<stamp> might indicate, it is also used for
120	throughout libev.	144	time differences (e.g. delays) throughout libev.
121		145
122	=head1 ERROR HANDLING	146	=head1 ERROR HANDLING
123		147
124	Libev knows three classes of errors: operating system errors, usage errors	148	Libev knows three classes of errors: operating system errors, usage errors
125	and internal errors (bugs).	149	and internal errors (bugs).
…		…
149		173
150	=item ev_tstamp ev_time ()	174	=item ev_tstamp ev_time ()
151		175
152	Returns the current time as libev would use it. Please note that the	176	Returns the current time as libev would use it. Please note that the
153	C<ev_now> function is usually faster and also often returns the timestamp	177	C<ev_now> function is usually faster and also often returns the timestamp
154	you actually want to know.	178	you actually want to know. Also interesting is the combination of
		179	C<ev_update_now> and C<ev_now>.
155		180
156	=item ev_sleep (ev_tstamp interval)	181	=item ev_sleep (ev_tstamp interval)
157		182
158	Sleep for the given interval: The current thread will be blocked until	183	Sleep for the given interval: The current thread will be blocked until
159	either it is interrupted or the given time interval has passed. Basically	184	either it is interrupted or the given time interval has passed. Basically
…		…
176	as this indicates an incompatible change. Minor versions are usually	201	as this indicates an incompatible change. Minor versions are usually
177	compatible to older versions, so a larger minor version alone is usually	202	compatible to older versions, so a larger minor version alone is usually
178	not a problem.	203	not a problem.
179		204
180	Example: Make sure we haven't accidentally been linked against the wrong	205	Example: Make sure we haven't accidentally been linked against the wrong
181	version.	206	version (note, however, that this will not detect other ABI mismatches,
		207	such as LFS or reentrancy).
182		208
183	assert (("libev version mismatch",	209	assert (("libev version mismatch",
184	ev_version_major () == EV_VERSION_MAJOR	210	ev_version_major () == EV_VERSION_MAJOR
185	&& ev_version_minor () >= EV_VERSION_MINOR));	211	&& ev_version_minor () >= EV_VERSION_MINOR));
186		212
…		…
197	assert (("sorry, no epoll, no sex",	223	assert (("sorry, no epoll, no sex",
198	ev_supported_backends () & EVBACKEND_EPOLL));	224	ev_supported_backends () & EVBACKEND_EPOLL));
199		225
200	=item unsigned int ev_recommended_backends ()	226	=item unsigned int ev_recommended_backends ()
201		227
202	Return the set of all backends compiled into this binary of libev and also	228	Return the set of all backends compiled into this binary of libev and
203	recommended for this platform. This set is often smaller than the one	229	also recommended for this platform, meaning it will work for most file
		230	descriptor types. This set is often smaller than the one returned by
204	returned by C<ev_supported_backends>, as for example kqueue is broken on	231	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
205	most BSDs and will not be auto-detected unless you explicitly request it	232	and will not be auto-detected unless you explicitly request it (assuming
206	(assuming you know what you are doing). This is the set of backends that	233	you know what you are doing). This is the set of backends that libev will
207	libev will probe for if you specify no backends explicitly.	234	probe for if you specify no backends explicitly.
208		235
209	=item unsigned int ev_embeddable_backends ()	236	=item unsigned int ev_embeddable_backends ()
210		237
211	Returns the set of backends that are embeddable in other event loops. This	238	Returns the set of backends that are embeddable in other event loops. This
212	is the theoretical, all-platform, value. To find which backends	239	value is platform-specific but can include backends not available on the
213	might be supported on the current system, you would need to look at	240	current system. To find which embeddable backends might be supported on
214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for	241	the current system, you would need to look at C<ev_embeddable_backends ()
215	recommended ones.	242	& ev_supported_backends ()>, likewise for recommended ones.
216		243
217	See the description of C<ev_embed> watchers for more info.	244	See the description of C<ev_embed> watchers for more info.
218		245
219	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	246	=item ev_set_allocator (void *(*cb)(void *ptr, long size))
220		247
…		…
250	}	277	}
251		278
252	...	279	...
253	ev_set_allocator (persistent_realloc);	280	ev_set_allocator (persistent_realloc);
254		281
255	=item ev_set_syserr_cb (void (*cb)(const char *msg));	282	=item ev_set_syserr_cb (void (*cb)(const char *msg))
256		283
257	Set the callback function to call on a retryable system call error (such	284	Set the callback function to call on a retryable system call error (such
258	as failed select, poll, epoll_wait). The message is a printable string	285	as failed select, poll, epoll_wait). The message is a printable string
259	indicating the system call or subsystem causing the problem. If this	286	indicating the system call or subsystem causing the problem. If this
260	callback is set, then libev will expect it to remedy the situation, no	287	callback is set, then libev will expect it to remedy the situation, no
…		…
272	}	299	}
273		300
274	...	301	...
275	ev_set_syserr_cb (fatal_error);	302	ev_set_syserr_cb (fatal_error);
276		303
		304	=item ev_feed_signal (int signum)
		305
		306	This function can be used to "simulate" a signal receive. It is completely
		307	safe to call this function at any time, from any context, including signal
		308	handlers or random threads.
		309
		310	Its main use is to customise signal handling in your process, especially
		311	in the presence of threads. For example, you could block signals
		312	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		313	creating any loops), and in one thread, use C<sigwait> or any other
		314	mechanism to wait for signals, then "deliver" them to libev by calling
		315	C<ev_feed_signal>.
		316
277	=back	317	=back
278		318
279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	319	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
280		320
281	An event loop is described by a C<struct ev_loop *>. The library knows two	321	An event loop is described by a C<struct ev_loop *> (the C<struct> is
282	types of such loops, the I<default> loop, which supports signals and child	322	I<not> optional in this case unless libev 3 compatibility is disabled, as
283	events, and dynamically created loops which do not.	323	libev 3 had an C<ev_loop> function colliding with the struct name).
		324
		325	The library knows two types of such loops, the I<default> loop, which
		326	supports child process events, and dynamically created event loops which
		327	do not.
284		328
285	=over 4	329	=over 4
286		330
287	=item struct ev_loop *ev_default_loop (unsigned int flags)	331	=item struct ev_loop *ev_default_loop (unsigned int flags)
288		332
289	This will initialise the default event loop if it hasn't been initialised	333	This returns the "default" event loop object, which is what you should
290	yet and return it. If the default loop could not be initialised, returns	334	normally use when you just need "the event loop". Event loop objects and
291	false. If it already was initialised it simply returns it (and ignores the	335	the C<flags> parameter are described in more detail in the entry for
292	flags. If that is troubling you, check C<ev_backend ()> afterwards).	336	C<ev_loop_new>.
		337
		338	If the default loop is already initialised then this function simply
		339	returns it (and ignores the flags. If that is troubling you, check
		340	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		341	flags, which should almost always be C<0>, unless the caller is also the
		342	one calling C<ev_run> or otherwise qualifies as "the main program".
293		343
294	If you don't know what event loop to use, use the one returned from this	344	If you don't know what event loop to use, use the one returned from this
295	function.	345	function (or via the C<EV_DEFAULT> macro).
296		346
297	Note that this function is I<not> thread-safe, so if you want to use it	347	Note that this function is I<not> thread-safe, so if you want to use it
298	from multiple threads, you have to lock (note also that this is unlikely,	348	from multiple threads, you have to employ some kind of mutex (note also
299	as loops cannot bes hared easily between threads anyway).	349	that this case is unlikely, as loops cannot be shared easily between
		350	threads anyway).
300		351
301	The default loop is the only loop that can handle C<ev_signal> and	352	The default loop is the only loop that can handle C<ev_child> watchers,
302	C<ev_child> watchers, and to do this, it always registers a handler	353	and to do this, it always registers a handler for C<SIGCHLD>. If this is
303	for C<SIGCHLD>. If this is a problem for your application you can either	354	a problem for your application you can either create a dynamic loop with
304	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you	355	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
305	can simply overwrite the C<SIGCHLD> signal handler I<after> calling	356	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
306	C<ev_default_init>.	357
		358	Example: This is the most typical usage.
		359
		360	if (!ev_default_loop (0))
		361	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		362
		363	Example: Restrict libev to the select and poll backends, and do not allow
		364	environment settings to be taken into account:
		365
		366	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		367
		368	=item struct ev_loop *ev_loop_new (unsigned int flags)
		369
		370	This will create and initialise a new event loop object. If the loop
		371	could not be initialised, returns false.
		372
		373	This function is thread-safe, and one common way to use libev with
		374	threads is indeed to create one loop per thread, and using the default
		375	loop in the "main" or "initial" thread.
307		376
308	The flags argument can be used to specify special behaviour or specific	377	The flags argument can be used to specify special behaviour or specific
309	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).	378	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
310		379
311	The following flags are supported:	380	The following flags are supported:
…		…
326	useful to try out specific backends to test their performance, or to work	395	useful to try out specific backends to test their performance, or to work
327	around bugs.	396	around bugs.
328		397
329	=item C<EVFLAG_FORKCHECK>	398	=item C<EVFLAG_FORKCHECK>
330		399
331	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after	400	Instead of calling C<ev_loop_fork> manually after a fork, you can also
332	a fork, you can also make libev check for a fork in each iteration by	401	make libev check for a fork in each iteration by enabling this flag.
333	enabling this flag.
334		402
335	This works by calling C<getpid ()> on every iteration of the loop,	403	This works by calling C<getpid ()> on every iteration of the loop,
336	and thus this might slow down your event loop if you do a lot of loop	404	and thus this might slow down your event loop if you do a lot of loop
337	iterations and little real work, but is usually not noticeable (on my	405	iterations and little real work, but is usually not noticeable (on my
338	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence	406	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
…		…
344	flag.	412	flag.
345		413
346	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>	414	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
347	environment variable.	415	environment variable.
348		416
		417	=item C<EVFLAG_NOINOTIFY>
		418
		419	When this flag is specified, then libev will not attempt to use the
		420	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		421	testing, this flag can be useful to conserve inotify file descriptors, as
		422	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		423
		424	=item C<EVFLAG_SIGNALFD>
		425
		426	When this flag is specified, then libev will attempt to use the
		427	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		428	delivers signals synchronously, which makes it both faster and might make
		429	it possible to get the queued signal data. It can also simplify signal
		430	handling with threads, as long as you properly block signals in your
		431	threads that are not interested in handling them.
		432
		433	Signalfd will not be used by default as this changes your signal mask, and
		434	there are a lot of shoddy libraries and programs (glib's threadpool for
		435	example) that can't properly initialise their signal masks.
		436
		437	=item C<EVFLAG_NOSIGMASK>
		438
		439	When this flag is specified, then libev will avoid to modify the signal
		440	mask. Specifically, this means you ahve to make sure signals are unblocked
		441	when you want to receive them.
		442
		443	This behaviour is useful when you want to do your own signal handling, or
		444	want to handle signals only in specific threads and want to avoid libev
		445	unblocking the signals.
		446
		447	It's also required by POSIX in a threaded program, as libev calls
		448	C<sigprocmask>, whose behaviour is officially unspecified.
		449
		450	This flag's behaviour will become the default in future versions of libev.
		451
349	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	452	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
350		453
351	This is your standard select(2) backend. Not I<completely> standard, as	454	This is your standard select(2) backend. Not I<completely> standard, as
352	libev tries to roll its own fd_set with no limits on the number of fds,	455	libev tries to roll its own fd_set with no limits on the number of fds,
353	but if that fails, expect a fairly low limit on the number of fds when	456	but if that fails, expect a fairly low limit on the number of fds when
…		…
359	writing a server, you should C<accept ()> in a loop to accept as many	462	writing a server, you should C<accept ()> in a loop to accept as many
360	connections as possible during one iteration. You might also want to have	463	connections as possible during one iteration. You might also want to have
361	a look at C<ev_set_io_collect_interval ()> to increase the amount of	464	a look at C<ev_set_io_collect_interval ()> to increase the amount of
362	readiness notifications you get per iteration.	465	readiness notifications you get per iteration.
363		466
		467	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		468	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		469	C<exceptfds> set on that platform).
		470
364	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	471	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
365		472
366	And this is your standard poll(2) backend. It's more complicated	473	And this is your standard poll(2) backend. It's more complicated
367	than select, but handles sparse fds better and has no artificial	474	than select, but handles sparse fds better and has no artificial
368	limit on the number of fds you can use (except it will slow down	475	limit on the number of fds you can use (except it will slow down
369	considerably with a lot of inactive fds). It scales similarly to select,	476	considerably with a lot of inactive fds). It scales similarly to select,
370	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for	477	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
371	performance tips.	478	performance tips.
372		479
		480	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		481	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
		482
373	=item C<EVBACKEND_EPOLL> (value 4, Linux)	483	=item C<EVBACKEND_EPOLL> (value 4, Linux)
374		484
		485	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		486	kernels).
		487
375	For few fds, this backend is a bit little slower than poll and select,	488	For few fds, this backend is a bit little slower than poll and select, but
376	but it scales phenomenally better. While poll and select usually scale	489	it scales phenomenally better. While poll and select usually scale like
377	like O(total_fds) where n is the total number of fds (or the highest fd),	490	O(total_fds) where total_fds is the total number of fds (or the highest
378	epoll scales either O(1) or O(active_fds). The epoll design has a number	491	fd), epoll scales either O(1) or O(active_fds).
379	of shortcomings, such as silently dropping events in some hard-to-detect	492
380	cases and requiring a system call per fd change, no fork support and bad	493	The epoll mechanism deserves honorable mention as the most misdesigned
381	support for dup.	494	of the more advanced event mechanisms: mere annoyances include silently
		495	dropping file descriptors, requiring a system call per change per file
		496	descriptor (and unnecessary guessing of parameters), problems with dup,
		497	returning before the timeout value, resulting in additional iterations
		498	(and only giving 5ms accuracy while select on the same platform gives
		499	0.1ms) and so on. The biggest issue is fork races, however - if a program
		500	forks then I<both> parent and child process have to recreate the epoll
		501	set, which can take considerable time (one syscall per file descriptor)
		502	and is of course hard to detect.
		503
		504	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
		505	but of course I<doesn't>, and epoll just loves to report events for
		506	totally I<different> file descriptors (even already closed ones, so
		507	one cannot even remove them from the set) than registered in the set
		508	(especially on SMP systems). Libev tries to counter these spurious
		509	notifications by employing an additional generation counter and comparing
		510	that against the events to filter out spurious ones, recreating the set
		511	when required. Epoll also errornously rounds down timeouts, but gives you
		512	no way to know when and by how much, so sometimes you have to busy-wait
		513	because epoll returns immediately despite a nonzero timeout. And last
		514	not least, it also refuses to work with some file descriptors which work
		515	perfectly fine with C<select> (files, many character devices...).
		516
		517	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		518	cobbled together in a hurry, no thought to design or interaction with
		519	others. Oh, the pain, will it ever stop...
382		520
383	While stopping, setting and starting an I/O watcher in the same iteration	521	While stopping, setting and starting an I/O watcher in the same iteration
384	will result in some caching, there is still a system call per such incident	522	will result in some caching, there is still a system call per such
385	(because the fd could point to a different file description now), so its	523	incident (because the same I<file descriptor> could point to a different
386	best to avoid that. Also, C<dup ()>'ed file descriptors might not work	524	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
387	very well if you register events for both fds.	525	file descriptors might not work very well if you register events for both
388		526	file descriptors.
389	Please note that epoll sometimes generates spurious notifications, so you
390	need to use non-blocking I/O or other means to avoid blocking when no data
391	(or space) is available.
392		527
393	Best performance from this backend is achieved by not unregistering all	528	Best performance from this backend is achieved by not unregistering all
394	watchers for a file descriptor until it has been closed, if possible, i.e.	529	watchers for a file descriptor until it has been closed, if possible,
395	keep at least one watcher active per fd at all times.	530	i.e. keep at least one watcher active per fd at all times. Stopping and
		531	starting a watcher (without re-setting it) also usually doesn't cause
		532	extra overhead. A fork can both result in spurious notifications as well
		533	as in libev having to destroy and recreate the epoll object, which can
		534	take considerable time and thus should be avoided.
		535
		536	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		537	faster than epoll for maybe up to a hundred file descriptors, depending on
		538	the usage. So sad.
396		539
397	While nominally embeddable in other event loops, this feature is broken in	540	While nominally embeddable in other event loops, this feature is broken in
398	all kernel versions tested so far.	541	all kernel versions tested so far.
		542
		543	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		544	C<EVBACKEND_POLL>.
399		545
400	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	546	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
401		547
402	Kqueue deserves special mention, as at the time of this writing, it	548	Kqueue deserves special mention, as at the time of this writing, it
403	was broken on all BSDs except NetBSD (usually it doesn't work reliably	549	was broken on all BSDs except NetBSD (usually it doesn't work reliably
404	with anything but sockets and pipes, except on Darwin, where of course	550	with anything but sockets and pipes, except on Darwin, where of course
405	it's completely useless). For this reason it's not being "auto-detected"	551	it's completely useless). Unlike epoll, however, whose brokenness
		552	is by design, these kqueue bugs can (and eventually will) be fixed
		553	without API changes to existing programs. For this reason it's not being
406	unless you explicitly specify it explicitly in the flags (i.e. using	554	"auto-detected" unless you explicitly specify it in the flags (i.e. using
407	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)	555	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
408	system like NetBSD.	556	system like NetBSD.
409		557
410	You still can embed kqueue into a normal poll or select backend and use it	558	You still can embed kqueue into a normal poll or select backend and use it
411	only for sockets (after having made sure that sockets work with kqueue on	559	only for sockets (after having made sure that sockets work with kqueue on
…		…
413		561
414	It scales in the same way as the epoll backend, but the interface to the	562	It scales in the same way as the epoll backend, but the interface to the
415	kernel is more efficient (which says nothing about its actual speed, of	563	kernel is more efficient (which says nothing about its actual speed, of
416	course). While stopping, setting and starting an I/O watcher does never	564	course). While stopping, setting and starting an I/O watcher does never
417	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	565	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
418	two event changes per incident, support for C<fork ()> is very bad and it	566	two event changes per incident. Support for C<fork ()> is very bad (but
419	drops fds silently in similarly hard-to-detect cases.	567	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		568	cases
420		569
421	This backend usually performs well under most conditions.	570	This backend usually performs well under most conditions.
422		571
423	While nominally embeddable in other event loops, this doesn't work	572	While nominally embeddable in other event loops, this doesn't work
424	everywhere, so you might need to test for this. And since it is broken	573	everywhere, so you might need to test for this. And since it is broken
425	almost everywhere, you should only use it when you have a lot of sockets	574	almost everywhere, you should only use it when you have a lot of sockets
426	(for which it usually works), by embedding it into another event loop	575	(for which it usually works), by embedding it into another event loop
427	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for	576	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
428	sockets.	577	also broken on OS X)) and, did I mention it, using it only for sockets.
		578
		579	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		580	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		581	C<NOTE_EOF>.
429		582
430	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)	583	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
431		584
432	This is not implemented yet (and might never be, unless you send me an	585	This is not implemented yet (and might never be, unless you send me an
433	implementation). According to reports, C</dev/poll> only supports sockets	586	implementation). According to reports, C</dev/poll> only supports sockets
…		…
437	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	590	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
438		591
439	This uses the Solaris 10 event port mechanism. As with everything on Solaris,	592	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
440	it's really slow, but it still scales very well (O(active_fds)).	593	it's really slow, but it still scales very well (O(active_fds)).
441		594
442	Please note that Solaris event ports can deliver a lot of spurious
443	notifications, so you need to use non-blocking I/O or other means to avoid
444	blocking when no data (or space) is available.
445
446	While this backend scales well, it requires one system call per active	595	While this backend scales well, it requires one system call per active
447	file descriptor per loop iteration. For small and medium numbers of file	596	file descriptor per loop iteration. For small and medium numbers of file
448	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend	597	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
449	might perform better.	598	might perform better.
450		599
451	On the positive side, ignoring the spurious readiness notifications, this	600	On the positive side, this backend actually performed fully to
452	backend actually performed to specification in all tests and is fully	601	specification in all tests and is fully embeddable, which is a rare feat
453	embeddable, which is a rare feat among the OS-specific backends.	602	among the OS-specific backends (I vastly prefer correctness over speed
		603	hacks).
		604
		605	On the negative side, the interface is I<bizarre> - so bizarre that
		606	even sun itself gets it wrong in their code examples: The event polling
		607	function sometimes returning events to the caller even though an error
		608	occurred, but with no indication whether it has done so or not (yes, it's
		609	even documented that way) - deadly for edge-triggered interfaces where
		610	you absolutely have to know whether an event occurred or not because you
		611	have to re-arm the watcher.
		612
		613	Fortunately libev seems to be able to work around these idiocies.
		614
		615	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		616	C<EVBACKEND_POLL>.
454		617
455	=item C<EVBACKEND_ALL>	618	=item C<EVBACKEND_ALL>
456		619
457	Try all backends (even potentially broken ones that wouldn't be tried	620	Try all backends (even potentially broken ones that wouldn't be tried
458	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	621	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
459	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	622	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
460		623
461	It is definitely not recommended to use this flag.	624	It is definitely not recommended to use this flag, use whatever
		625	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		626	at all.
		627
		628	=item C<EVBACKEND_MASK>
		629
		630	Not a backend at all, but a mask to select all backend bits from a
		631	C<flags> value, in case you want to mask out any backends from a flags
		632	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
462		633
463	=back	634	=back
464		635
465	If one or more of these are or'ed into the flags value, then only these	636	If one or more of the backend flags are or'ed into the flags value,
466	backends will be tried (in the reverse order as listed here). If none are	637	then only these backends will be tried (in the reverse order as listed
467	specified, all backends in C<ev_recommended_backends ()> will be tried.	638	here). If none are specified, all backends in C<ev_recommended_backends
468		639	()> will be tried.
469	The most typical usage is like this:
470
471	if (!ev_default_loop (0))
472	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
473
474	Restrict libev to the select and poll backends, and do not allow
475	environment settings to be taken into account:
476
477	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
478
479	Use whatever libev has to offer, but make sure that kqueue is used if
480	available (warning, breaks stuff, best use only with your own private
481	event loop and only if you know the OS supports your types of fds):
482
483	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
484
485	=item struct ev_loop *ev_loop_new (unsigned int flags)
486
487	Similar to C<ev_default_loop>, but always creates a new event loop that is
488	always distinct from the default loop. Unlike the default loop, it cannot
489	handle signal and child watchers, and attempts to do so will be greeted by
490	undefined behaviour (or a failed assertion if assertions are enabled).
491
492	Note that this function I<is> thread-safe, and the recommended way to use
493	libev with threads is indeed to create one loop per thread, and using the
494	default loop in the "main" or "initial" thread.
495		640
496	Example: Try to create a event loop that uses epoll and nothing else.	641	Example: Try to create a event loop that uses epoll and nothing else.
497		642
498	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);	643	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
499	if (!epoller)	644	if (!epoller)
500	fatal ("no epoll found here, maybe it hides under your chair");	645	fatal ("no epoll found here, maybe it hides under your chair");
501		646
		647	Example: Use whatever libev has to offer, but make sure that kqueue is
		648	used if available.
		649
		650	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		651
502	=item ev_default_destroy ()	652	=item ev_loop_destroy (loop)
503		653
504	Destroys the default loop again (frees all memory and kernel state	654	Destroys an event loop object (frees all memory and kernel state
505	etc.). None of the active event watchers will be stopped in the normal	655	etc.). None of the active event watchers will be stopped in the normal
506	sense, so e.g. C<ev_is_active> might still return true. It is your	656	sense, so e.g. C<ev_is_active> might still return true. It is your
507	responsibility to either stop all watchers cleanly yourself I<before>	657	responsibility to either stop all watchers cleanly yourself I<before>
508	calling this function, or cope with the fact afterwards (which is usually	658	calling this function, or cope with the fact afterwards (which is usually
509	the easiest thing, you can just ignore the watchers and/or C<free ()> them	659	the easiest thing, you can just ignore the watchers and/or C<free ()> them
510	for example).	660	for example).
511		661
512	Note that certain global state, such as signal state, will not be freed by	662	Note that certain global state, such as signal state (and installed signal
513	this function, and related watchers (such as signal and child watchers)	663	handlers), will not be freed by this function, and related watchers (such
514	would need to be stopped manually.	664	as signal and child watchers) would need to be stopped manually.
515		665
516	In general it is not advisable to call this function except in the	666	This function is normally used on loop objects allocated by
517	rare occasion where you really need to free e.g. the signal handling	667	C<ev_loop_new>, but it can also be used on the default loop returned by
		668	C<ev_default_loop>, in which case it is not thread-safe.
		669
		670	Note that it is not advisable to call this function on the default loop
		671	except in the rare occasion where you really need to free its resources.
518	pipe fds. If you need dynamically allocated loops it is better to use	672	If you need dynamically allocated loops it is better to use C<ev_loop_new>
519	C<ev_loop_new> and C<ev_loop_destroy>).	673	and C<ev_loop_destroy>.
520		674
521	=item ev_loop_destroy (loop)	675	=item ev_loop_fork (loop)
522		676
523	Like C<ev_default_destroy>, but destroys an event loop created by an
524	earlier call to C<ev_loop_new>.
525
526	=item ev_default_fork ()
527
528	This function sets a flag that causes subsequent C<ev_loop> iterations	677	This function sets a flag that causes subsequent C<ev_run> iterations to
529	to reinitialise the kernel state for backends that have one. Despite the	678	reinitialise the kernel state for backends that have one. Despite the
530	name, you can call it anytime, but it makes most sense after forking, in	679	name, you can call it anytime, but it makes most sense after forking, in
531	the child process (or both child and parent, but that again makes little	680	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
532	sense). You I<must> call it in the child before using any of the libev	681	child before resuming or calling C<ev_run>.
533	functions, and it will only take effect at the next C<ev_loop> iteration.	682
		683	Again, you I<have> to call it on I<any> loop that you want to re-use after
		684	a fork, I<even if you do not plan to use the loop in the parent>. This is
		685	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		686	during fork.
534		687
535	On the other hand, you only need to call this function in the child	688	On the other hand, you only need to call this function in the child
536	process if and only if you want to use the event library in the child. If	689	process if and only if you want to use the event loop in the child. If
537	you just fork+exec, you don't have to call it at all.	690	you just fork+exec or create a new loop in the child, you don't have to
		691	call it at all (in fact, C<epoll> is so badly broken that it makes a
		692	difference, but libev will usually detect this case on its own and do a
		693	costly reset of the backend).
538		694
539	The function itself is quite fast and it's usually not a problem to call	695	The function itself is quite fast and it's usually not a problem to call
540	it just in case after a fork. To make this easy, the function will fit in	696	it just in case after a fork.
541	quite nicely into a call to C<pthread_atfork>:
542		697
		698	Example: Automate calling C<ev_loop_fork> on the default loop when
		699	using pthreads.
		700
		701	static void
		702	post_fork_child (void)
		703	{
		704	ev_loop_fork (EV_DEFAULT);
		705	}
		706
		707	...
543	pthread_atfork (0, 0, ev_default_fork);	708	pthread_atfork (0, 0, post_fork_child);
544
545	=item ev_loop_fork (loop)
546
547	Like C<ev_default_fork>, but acts on an event loop created by
548	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
549	after fork, and how you do this is entirely your own problem.
550		709
551	=item int ev_is_default_loop (loop)	710	=item int ev_is_default_loop (loop)
552		711
553	Returns true when the given loop actually is the default loop, false otherwise.	712	Returns true when the given loop is, in fact, the default loop, and false
		713	otherwise.
554		714
555	=item unsigned int ev_loop_count (loop)	715	=item unsigned int ev_iteration (loop)
556		716
557	Returns the count of loop iterations for the loop, which is identical to	717	Returns the current iteration count for the event loop, which is identical
558	the number of times libev did poll for new events. It starts at C<0> and	718	to the number of times libev did poll for new events. It starts at C<0>
559	happily wraps around with enough iterations.	719	and happily wraps around with enough iterations.
560		720
561	This value can sometimes be useful as a generation counter of sorts (it	721	This value can sometimes be useful as a generation counter of sorts (it
562	"ticks" the number of loop iterations), as it roughly corresponds with	722	"ticks" the number of loop iterations), as it roughly corresponds with
563	C<ev_prepare> and C<ev_check> calls.	723	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		724	prepare and check phases.
		725
		726	=item unsigned int ev_depth (loop)
		727
		728	Returns the number of times C<ev_run> was entered minus the number of
		729	times C<ev_run> was exited normally, in other words, the recursion depth.
		730
		731	Outside C<ev_run>, this number is zero. In a callback, this number is
		732	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		733	in which case it is higher.
		734
		735	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		736	throwing an exception etc.), doesn't count as "exit" - consider this
		737	as a hint to avoid such ungentleman-like behaviour unless it's really
		738	convenient, in which case it is fully supported.
564		739
565	=item unsigned int ev_backend (loop)	740	=item unsigned int ev_backend (loop)
566		741
567	Returns one of the C<EVBACKEND_*> flags indicating the event backend in	742	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
568	use.	743	use.
…		…
577		752
578	=item ev_now_update (loop)	753	=item ev_now_update (loop)
579		754
580	Establishes the current time by querying the kernel, updating the time	755	Establishes the current time by querying the kernel, updating the time
581	returned by C<ev_now ()> in the progress. This is a costly operation and	756	returned by C<ev_now ()> in the progress. This is a costly operation and
582	is usually done automatically within C<ev_loop ()>.	757	is usually done automatically within C<ev_run ()>.
583		758
584	This function is rarely useful, but when some event callback runs for a	759	This function is rarely useful, but when some event callback runs for a
585	very long time without entering the event loop, updating libev's idea of	760	very long time without entering the event loop, updating libev's idea of
586	the current time is a good idea.	761	the current time is a good idea.
587		762
588	See also "The special problem of time updates" in the C<ev_timer> section.	763	See also L<The special problem of time updates> in the C<ev_timer> section.
589		764
		765	=item ev_suspend (loop)
		766
		767	=item ev_resume (loop)
		768
		769	These two functions suspend and resume an event loop, for use when the
		770	loop is not used for a while and timeouts should not be processed.
		771
		772	A typical use case would be an interactive program such as a game: When
		773	the user presses C<^Z> to suspend the game and resumes it an hour later it
		774	would be best to handle timeouts as if no time had actually passed while
		775	the program was suspended. This can be achieved by calling C<ev_suspend>
		776	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		777	C<ev_resume> directly afterwards to resume timer processing.
		778
		779	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		780	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		781	will be rescheduled (that is, they will lose any events that would have
		782	occurred while suspended).
		783
		784	After calling C<ev_suspend> you B<must not> call I<any> function on the
		785	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		786	without a previous call to C<ev_suspend>.
		787
		788	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		789	event loop time (see C<ev_now_update>).
		790
590	=item ev_loop (loop, int flags)	791	=item ev_run (loop, int flags)
591		792
592	Finally, this is it, the event handler. This function usually is called	793	Finally, this is it, the event handler. This function usually is called
593	after you initialised all your watchers and you want to start handling	794	after you have initialised all your watchers and you want to start
594	events.	795	handling events. It will ask the operating system for any new events, call
		796	the watcher callbacks, an then repeat the whole process indefinitely: This
		797	is why event loops are called I<loops>.
595		798
596	If the flags argument is specified as C<0>, it will not return until	799	If the flags argument is specified as C<0>, it will keep handling events
597	either no event watchers are active anymore or C<ev_unloop> was called.	800	until either no event watchers are active anymore or C<ev_break> was
		801	called.
598		802
599	Please note that an explicit C<ev_unloop> is usually better than	803	Please note that an explicit C<ev_break> is usually better than
600	relying on all watchers to be stopped when deciding when a program has	804	relying on all watchers to be stopped when deciding when a program has
601	finished (especially in interactive programs), but having a program that	805	finished (especially in interactive programs), but having a program
602	automatically loops as long as it has to and no longer by virtue of	806	that automatically loops as long as it has to and no longer by virtue
603	relying on its watchers stopping correctly is a thing of beauty.	807	of relying on its watchers stopping correctly, that is truly a thing of
		808	beauty.
604		809
		810	This function is also I<mostly> exception-safe - you can break out of
		811	a C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		812	exception and so on. This does not decrement the C<ev_depth> value, nor
		813	will it clear any outstanding C<EVBREAK_ONE> breaks.
		814
605	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	815	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
606	those events and any outstanding ones, but will not block your process in	816	those events and any already outstanding ones, but will not wait and
607	case there are no events and will return after one iteration of the loop.	817	block your process in case there are no events and will return after one
		818	iteration of the loop. This is sometimes useful to poll and handle new
		819	events while doing lengthy calculations, to keep the program responsive.
608		820
609	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	821	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
610	necessary) and will handle those and any outstanding ones. It will block	822	necessary) and will handle those and any already outstanding ones. It
611	your process until at least one new event arrives, and will return after	823	will block your process until at least one new event arrives (which could
612	one iteration of the loop. This is useful if you are waiting for some	824	be an event internal to libev itself, so there is no guarantee that a
613	external event in conjunction with something not expressible using other	825	user-registered callback will be called), and will return after one
		826	iteration of the loop.
		827
		828	This is useful if you are waiting for some external event in conjunction
		829	with something not expressible using other libev watchers (i.e. "roll your
614	libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is	830	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
615	usually a better approach for this kind of thing.	831	usually a better approach for this kind of thing.
616		832
617	Here are the gory details of what C<ev_loop> does:	833	Here are the gory details of what C<ev_run> does (this is for your
		834	understanding, not a guarantee that things will work exactly like this in
		835	future versions):
618		836
		837	- Increment loop depth.
		838	- Reset the ev_break status.
619	- Before the first iteration, call any pending watchers.	839	- Before the first iteration, call any pending watchers.
		840	LOOP:
620	* If EVFLAG_FORKCHECK was used, check for a fork.	841	- If EVFLAG_FORKCHECK was used, check for a fork.
621	- If a fork was detected (by any means), queue and call all fork watchers.	842	- If a fork was detected (by any means), queue and call all fork watchers.
622	- Queue and call all prepare watchers.	843	- Queue and call all prepare watchers.
		844	- If ev_break was called, goto FINISH.
623	- If we have been forked, detach and recreate the kernel state	845	- If we have been forked, detach and recreate the kernel state
624	as to not disturb the other process.	846	as to not disturb the other process.
625	- Update the kernel state with all outstanding changes.	847	- Update the kernel state with all outstanding changes.
626	- Update the "event loop time" (ev_now ()).	848	- Update the "event loop time" (ev_now ()).
627	- Calculate for how long to sleep or block, if at all	849	- Calculate for how long to sleep or block, if at all
628	(active idle watchers, EVLOOP_NONBLOCK or not having	850	(active idle watchers, EVRUN_NOWAIT or not having
629	any active watchers at all will result in not sleeping).	851	any active watchers at all will result in not sleeping).
630	- Sleep if the I/O and timer collect interval say so.	852	- Sleep if the I/O and timer collect interval say so.
		853	- Increment loop iteration counter.
631	- Block the process, waiting for any events.	854	- Block the process, waiting for any events.
632	- Queue all outstanding I/O (fd) events.	855	- Queue all outstanding I/O (fd) events.
633	- Update the "event loop time" (ev_now ()), and do time jump adjustments.	856	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
634	- Queue all outstanding timers.	857	- Queue all expired timers.
635	- Queue all outstanding periodics.	858	- Queue all expired periodics.
636	- Unless any events are pending now, queue all idle watchers.	859	- Queue all idle watchers with priority higher than that of pending events.
637	- Queue all check watchers.	860	- Queue all check watchers.
638	- Call all queued watchers in reverse order (i.e. check watchers first).	861	- Call all queued watchers in reverse order (i.e. check watchers first).
639	Signals and child watchers are implemented as I/O watchers, and will	862	Signals and child watchers are implemented as I/O watchers, and will
640	be handled here by queueing them when their watcher gets executed.	863	be handled here by queueing them when their watcher gets executed.
641	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	864	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
642	were used, or there are no active watchers, return, otherwise	865	were used, or there are no active watchers, goto FINISH, otherwise
643	continue with step *.	866	continue with step LOOP.
		867	FINISH:
		868	- Reset the ev_break status iff it was EVBREAK_ONE.
		869	- Decrement the loop depth.
		870	- Return.
644		871
645	Example: Queue some jobs and then loop until no events are outstanding	872	Example: Queue some jobs and then loop until no events are outstanding
646	anymore.	873	anymore.
647		874
648	... queue jobs here, make sure they register event watchers as long	875	... queue jobs here, make sure they register event watchers as long
649	... as they still have work to do (even an idle watcher will do..)	876	... as they still have work to do (even an idle watcher will do..)
650	ev_loop (my_loop, 0);	877	ev_run (my_loop, 0);
651	... jobs done or somebody called unloop. yeah!	878	... jobs done or somebody called break. yeah!
652		879
653	=item ev_unloop (loop, how)	880	=item ev_break (loop, how)
654		881
655	Can be used to make a call to C<ev_loop> return early (but only after it	882	Can be used to make a call to C<ev_run> return early (but only after it
656	has processed all outstanding events). The C<how> argument must be either	883	has processed all outstanding events). The C<how> argument must be either
657	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	884	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
658	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	885	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
659		886
660	This "unloop state" will be cleared when entering C<ev_loop> again.	887	This "break state" will be cleared on the next call to C<ev_run>.
		888
		889	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		890	which case it will have no effect.
661		891
662	=item ev_ref (loop)	892	=item ev_ref (loop)
663		893
664	=item ev_unref (loop)	894	=item ev_unref (loop)
665		895
666	Ref/unref can be used to add or remove a reference count on the event	896	Ref/unref can be used to add or remove a reference count on the event
667	loop: Every watcher keeps one reference, and as long as the reference	897	loop: Every watcher keeps one reference, and as long as the reference
668	count is nonzero, C<ev_loop> will not return on its own. If you have	898	count is nonzero, C<ev_run> will not return on its own.
669	a watcher you never unregister that should not keep C<ev_loop> from	899
670	returning, ev_unref() after starting, and ev_ref() before stopping it. For	900	This is useful when you have a watcher that you never intend to
		901	unregister, but that nevertheless should not keep C<ev_run> from
		902	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		903	before stopping it.
		904
671	example, libev itself uses this for its internal signal pipe: It is not	905	As an example, libev itself uses this for its internal signal pipe: It
672	visible to the libev user and should not keep C<ev_loop> from exiting if	906	is not visible to the libev user and should not keep C<ev_run> from
673	no event watchers registered by it are active. It is also an excellent	907	exiting if no event watchers registered by it are active. It is also an
674	way to do this for generic recurring timers or from within third-party	908	excellent way to do this for generic recurring timers or from within
675	libraries. Just remember to I<unref after start> and I<ref before stop>	909	third-party libraries. Just remember to I<unref after start> and I<ref
676	(but only if the watcher wasn't active before, or was active before,	910	before stop> (but only if the watcher wasn't active before, or was active
677	respectively).	911	before, respectively. Note also that libev might stop watchers itself
		912	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		913	in the callback).
678		914
679	Example: Create a signal watcher, but keep it from keeping C<ev_loop>	915	Example: Create a signal watcher, but keep it from keeping C<ev_run>
680	running when nothing else is active.	916	running when nothing else is active.
681		917
682	struct ev_signal exitsig;	918	ev_signal exitsig;
683	ev_signal_init (&exitsig, sig_cb, SIGINT);	919	ev_signal_init (&exitsig, sig_cb, SIGINT);
684	ev_signal_start (loop, &exitsig);	920	ev_signal_start (loop, &exitsig);
685	evf_unref (loop);	921	ev_unref (loop);
686		922
687	Example: For some weird reason, unregister the above signal handler again.	923	Example: For some weird reason, unregister the above signal handler again.
688		924
689	ev_ref (loop);	925	ev_ref (loop);
690	ev_signal_stop (loop, &exitsig);	926	ev_signal_stop (loop, &exitsig);
…		…
701	Setting these to a higher value (the C<interval> I<must> be >= C<0>)	937	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
702	allows libev to delay invocation of I/O and timer/periodic callbacks	938	allows libev to delay invocation of I/O and timer/periodic callbacks
703	to increase efficiency of loop iterations (or to increase power-saving	939	to increase efficiency of loop iterations (or to increase power-saving
704	opportunities).	940	opportunities).
705		941
706	The background is that sometimes your program runs just fast enough to	942	The idea is that sometimes your program runs just fast enough to handle
707	handle one (or very few) event(s) per loop iteration. While this makes	943	one (or very few) event(s) per loop iteration. While this makes the
708	the program responsive, it also wastes a lot of CPU time to poll for new	944	program responsive, it also wastes a lot of CPU time to poll for new
709	events, especially with backends like C<select ()> which have a high	945	events, especially with backends like C<select ()> which have a high
710	overhead for the actual polling but can deliver many events at once.	946	overhead for the actual polling but can deliver many events at once.
711		947
712	By setting a higher I<io collect interval> you allow libev to spend more	948	By setting a higher I<io collect interval> you allow libev to spend more
713	time collecting I/O events, so you can handle more events per iteration,	949	time collecting I/O events, so you can handle more events per iteration,
714	at the cost of increasing latency. Timeouts (both C<ev_periodic> and	950	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
715	C<ev_timer>) will be not affected. Setting this to a non-null value will	951	C<ev_timer>) will be not affected. Setting this to a non-null value will
716	introduce an additional C<ev_sleep ()> call into most loop iterations.	952	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		953	sleep time ensures that libev will not poll for I/O events more often then
		954	once per this interval, on average.
717		955
718	Likewise, by setting a higher I<timeout collect interval> you allow libev	956	Likewise, by setting a higher I<timeout collect interval> you allow libev
719	to spend more time collecting timeouts, at the expense of increased	957	to spend more time collecting timeouts, at the expense of increased
720	latency (the watcher callback will be called later). C<ev_io> watchers	958	latency/jitter/inexactness (the watcher callback will be called
721	will not be affected. Setting this to a non-null value will not introduce	959	later). C<ev_io> watchers will not be affected. Setting this to a non-null
722	any overhead in libev.	960	value will not introduce any overhead in libev.
723		961
724	Many (busy) programs can usually benefit by setting the I/O collect	962	Many (busy) programs can usually benefit by setting the I/O collect
725	interval to a value near C<0.1> or so, which is often enough for	963	interval to a value near C<0.1> or so, which is often enough for
726	interactive servers (of course not for games), likewise for timeouts. It	964	interactive servers (of course not for games), likewise for timeouts. It
727	usually doesn't make much sense to set it to a lower value than C<0.01>,	965	usually doesn't make much sense to set it to a lower value than C<0.01>,
728	as this approaches the timing granularity of most systems.	966	as this approaches the timing granularity of most systems. Note that if
		967	you do transactions with the outside world and you can't increase the
		968	parallelity, then this setting will limit your transaction rate (if you
		969	need to poll once per transaction and the I/O collect interval is 0.01,
		970	then you can't do more than 100 transactions per second).
729		971
730	Setting the I<timeout collect interval> can improve the opportunity for	972	Setting the I<timeout collect interval> can improve the opportunity for
731	saving power, as the program will "bundle" timer callback invocations that	973	saving power, as the program will "bundle" timer callback invocations that
732	are "near" in time together, by delaying some, thus reducing the number of	974	are "near" in time together, by delaying some, thus reducing the number of
733	times the process sleeps and wakes up again. Another useful technique to	975	times the process sleeps and wakes up again. Another useful technique to
734	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure	976	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
735	they fire on, say, one-second boundaries only.	977	they fire on, say, one-second boundaries only.
736		978
		979	Example: we only need 0.1s timeout granularity, and we wish not to poll
		980	more often than 100 times per second:
		981
		982	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		983	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		984
		985	=item ev_invoke_pending (loop)
		986
		987	This call will simply invoke all pending watchers while resetting their
		988	pending state. Normally, C<ev_run> does this automatically when required,
		989	but when overriding the invoke callback this call comes handy. This
		990	function can be invoked from a watcher - this can be useful for example
		991	when you want to do some lengthy calculation and want to pass further
		992	event handling to another thread (you still have to make sure only one
		993	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		994
		995	=item int ev_pending_count (loop)
		996
		997	Returns the number of pending watchers - zero indicates that no watchers
		998	are pending.
		999
		1000	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1001
		1002	This overrides the invoke pending functionality of the loop: Instead of
		1003	invoking all pending watchers when there are any, C<ev_run> will call
		1004	this callback instead. This is useful, for example, when you want to
		1005	invoke the actual watchers inside another context (another thread etc.).
		1006
		1007	If you want to reset the callback, use C<ev_invoke_pending> as new
		1008	callback.
		1009
		1010	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		1011
		1012	Sometimes you want to share the same loop between multiple threads. This
		1013	can be done relatively simply by putting mutex_lock/unlock calls around
		1014	each call to a libev function.
		1015
		1016	However, C<ev_run> can run an indefinite time, so it is not feasible
		1017	to wait for it to return. One way around this is to wake up the event
		1018	loop via C<ev_break> and C<av_async_send>, another way is to set these
		1019	I<release> and I<acquire> callbacks on the loop.
		1020
		1021	When set, then C<release> will be called just before the thread is
		1022	suspended waiting for new events, and C<acquire> is called just
		1023	afterwards.
		1024
		1025	Ideally, C<release> will just call your mutex_unlock function, and
		1026	C<acquire> will just call the mutex_lock function again.
		1027
		1028	While event loop modifications are allowed between invocations of
		1029	C<release> and C<acquire> (that's their only purpose after all), no
		1030	modifications done will affect the event loop, i.e. adding watchers will
		1031	have no effect on the set of file descriptors being watched, or the time
		1032	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1033	to take note of any changes you made.
		1034
		1035	In theory, threads executing C<ev_run> will be async-cancel safe between
		1036	invocations of C<release> and C<acquire>.
		1037
		1038	See also the locking example in the C<THREADS> section later in this
		1039	document.
		1040
		1041	=item ev_set_userdata (loop, void *data)
		1042
		1043	=item void *ev_userdata (loop)
		1044
		1045	Set and retrieve a single C<void *> associated with a loop. When
		1046	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1047	C<0>.
		1048
		1049	These two functions can be used to associate arbitrary data with a loop,
		1050	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1051	C<acquire> callbacks described above, but of course can be (ab-)used for
		1052	any other purpose as well.
		1053
737	=item ev_loop_verify (loop)	1054	=item ev_verify (loop)
738		1055
739	This function only does something when C<EV_VERIFY> support has been	1056	This function only does something when C<EV_VERIFY> support has been
740	compiled in. It tries to go through all internal structures and checks	1057	compiled in, which is the default for non-minimal builds. It tries to go
741	them for validity. If anything is found to be inconsistent, it will print	1058	through all internal structures and checks them for validity. If anything
742	an error message to standard error and call C<abort ()>.	1059	is found to be inconsistent, it will print an error message to standard
		1060	error and call C<abort ()>.
743		1061
744	This can be used to catch bugs inside libev itself: under normal	1062	This can be used to catch bugs inside libev itself: under normal
745	circumstances, this function will never abort as of course libev keeps its	1063	circumstances, this function will never abort as of course libev keeps its
746	data structures consistent.	1064	data structures consistent.
747		1065
748	=back	1066	=back
749		1067
750		1068
751	=head1 ANATOMY OF A WATCHER	1069	=head1 ANATOMY OF A WATCHER
752		1070
		1071	In the following description, uppercase C<TYPE> in names stands for the
		1072	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1073	watchers and C<ev_io_start> for I/O watchers.
		1074
753	A watcher is a structure that you create and register to record your	1075	A watcher is an opaque structure that you allocate and register to record
754	interest in some event. For instance, if you want to wait for STDIN to	1076	your interest in some event. To make a concrete example, imagine you want
755	become readable, you would create an C<ev_io> watcher for that:	1077	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1078	for that:
756		1079
757	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1080	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
758	{	1081	{
759	ev_io_stop (w);	1082	ev_io_stop (w);
760	ev_unloop (loop, EVUNLOOP_ALL);	1083	ev_break (loop, EVBREAK_ALL);
761	}	1084	}
762		1085
763	struct ev_loop *loop = ev_default_loop (0);	1086	struct ev_loop *loop = ev_default_loop (0);
		1087
764	struct ev_io stdin_watcher;	1088	ev_io stdin_watcher;
		1089
765	ev_init (&stdin_watcher, my_cb);	1090	ev_init (&stdin_watcher, my_cb);
766	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1091	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
767	ev_io_start (loop, &stdin_watcher);	1092	ev_io_start (loop, &stdin_watcher);
		1093
768	ev_loop (loop, 0);	1094	ev_run (loop, 0);
769		1095
770	As you can see, you are responsible for allocating the memory for your	1096	As you can see, you are responsible for allocating the memory for your
771	watcher structures (and it is usually a bad idea to do this on the stack,	1097	watcher structures (and it is I<usually> a bad idea to do this on the
772	although this can sometimes be quite valid).	1098	stack).
773		1099
		1100	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1101	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1102
774	Each watcher structure must be initialised by a call to C<ev_init	1103	Each watcher structure must be initialised by a call to C<ev_init (watcher
775	(watcher *, callback)>, which expects a callback to be provided. This	1104	*, callback)>, which expects a callback to be provided. This callback is
776	callback gets invoked each time the event occurs (or, in the case of I/O	1105	invoked each time the event occurs (or, in the case of I/O watchers, each
777	watchers, each time the event loop detects that the file descriptor given	1106	time the event loop detects that the file descriptor given is readable
778	is readable and/or writable).	1107	and/or writable).
779		1108
780	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1109	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
781	with arguments specific to this watcher type. There is also a macro	1110	macro to configure it, with arguments specific to the watcher type. There
782	to combine initialisation and setting in one call: C<< ev_<type>_init	1111	is also a macro to combine initialisation and setting in one call: C<<
783	(watcher *, callback, ...) >>.	1112	ev_TYPE_init (watcher *, callback, ...) >>.
784		1113
785	To make the watcher actually watch out for events, you have to start it	1114	To make the watcher actually watch out for events, you have to start it
786	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1115	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
787	*) >>), and you can stop watching for events at any time by calling the	1116	*) >>), and you can stop watching for events at any time by calling the
788	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1117	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
789		1118
790	As long as your watcher is active (has been started but not stopped) you	1119	As long as your watcher is active (has been started but not stopped) you
791	must not touch the values stored in it. Most specifically you must never	1120	must not touch the values stored in it. Most specifically you must never
792	reinitialise it or call its C<set> macro.	1121	reinitialise it or call its C<ev_TYPE_set> macro.
793		1122
794	Each and every callback receives the event loop pointer as first, the	1123	Each and every callback receives the event loop pointer as first, the
795	registered watcher structure as second, and a bitset of received events as	1124	registered watcher structure as second, and a bitset of received events as
796	third argument.	1125	third argument.
797		1126
…		…
806	=item C<EV_WRITE>	1135	=item C<EV_WRITE>
807		1136
808	The file descriptor in the C<ev_io> watcher has become readable and/or	1137	The file descriptor in the C<ev_io> watcher has become readable and/or
809	writable.	1138	writable.
810		1139
811	=item C<EV_TIMEOUT>	1140	=item C<EV_TIMER>
812		1141
813	The C<ev_timer> watcher has timed out.	1142	The C<ev_timer> watcher has timed out.
814		1143
815	=item C<EV_PERIODIC>	1144	=item C<EV_PERIODIC>
816		1145
…		…
834		1163
835	=item C<EV_PREPARE>	1164	=item C<EV_PREPARE>
836		1165
837	=item C<EV_CHECK>	1166	=item C<EV_CHECK>
838		1167
839	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1168	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
840	to gather new events, and all C<ev_check> watchers are invoked just after	1169	to gather new events, and all C<ev_check> watchers are invoked just after
841	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1170	C<ev_run> has gathered them, but before it invokes any callbacks for any
842	received events. Callbacks of both watcher types can start and stop as	1171	received events. Callbacks of both watcher types can start and stop as
843	many watchers as they want, and all of them will be taken into account	1172	many watchers as they want, and all of them will be taken into account
844	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1173	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
845	C<ev_loop> from blocking).	1174	C<ev_run> from blocking).
846		1175
847	=item C<EV_EMBED>	1176	=item C<EV_EMBED>
848		1177
849	The embedded event loop specified in the C<ev_embed> watcher needs attention.	1178	The embedded event loop specified in the C<ev_embed> watcher needs attention.
850		1179
851	=item C<EV_FORK>	1180	=item C<EV_FORK>
852		1181
853	The event loop has been resumed in the child process after fork (see	1182	The event loop has been resumed in the child process after fork (see
854	C<ev_fork>).	1183	C<ev_fork>).
855		1184
		1185	=item C<EV_CLEANUP>
		1186
		1187	The event loop is about to be destroyed (see C<ev_cleanup>).
		1188
856	=item C<EV_ASYNC>	1189	=item C<EV_ASYNC>
857		1190
858	The given async watcher has been asynchronously notified (see C<ev_async>).	1191	The given async watcher has been asynchronously notified (see C<ev_async>).
		1192
		1193	=item C<EV_CUSTOM>
		1194
		1195	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1196	by libev users to signal watchers (e.g. via C<ev_feed_event>).
859		1197
860	=item C<EV_ERROR>	1198	=item C<EV_ERROR>
861		1199
862	An unspecified error has occurred, the watcher has been stopped. This might	1200	An unspecified error has occurred, the watcher has been stopped. This might
863	happen because the watcher could not be properly started because libev	1201	happen because the watcher could not be properly started because libev
864	ran out of memory, a file descriptor was found to be closed or any other	1202	ran out of memory, a file descriptor was found to be closed or any other
		1203	problem. Libev considers these application bugs.
		1204
865	problem. You best act on it by reporting the problem and somehow coping	1205	You best act on it by reporting the problem and somehow coping with the
866	with the watcher being stopped.	1206	watcher being stopped. Note that well-written programs should not receive
		1207	an error ever, so when your watcher receives it, this usually indicates a
		1208	bug in your program.
867		1209
868	Libev will usually signal a few "dummy" events together with an error,	1210	Libev will usually signal a few "dummy" events together with an error, for
869	for example it might indicate that a fd is readable or writable, and if	1211	example it might indicate that a fd is readable or writable, and if your
870	your callbacks is well-written it can just attempt the operation and cope	1212	callbacks is well-written it can just attempt the operation and cope with
871	with the error from read() or write(). This will not work in multi-threaded	1213	the error from read() or write(). This will not work in multi-threaded
872	programs, though, so beware.	1214	programs, though, as the fd could already be closed and reused for another
		1215	thing, so beware.
873		1216
874	=back	1217	=back
875		1218
876	=head2 GENERIC WATCHER FUNCTIONS	1219	=head2 GENERIC WATCHER FUNCTIONS
877
878	In the following description, C<TYPE> stands for the watcher type,
879	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
880		1220
881	=over 4	1221	=over 4
882		1222
883	=item C<ev_init> (ev_TYPE *watcher, callback)	1223	=item C<ev_init> (ev_TYPE *watcher, callback)
884		1224
…		…
890	which rolls both calls into one.	1230	which rolls both calls into one.
891		1231
892	You can reinitialise a watcher at any time as long as it has been stopped	1232	You can reinitialise a watcher at any time as long as it has been stopped
893	(or never started) and there are no pending events outstanding.	1233	(or never started) and there are no pending events outstanding.
894		1234
895	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,	1235	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
896	int revents)>.	1236	int revents)>.
897		1237
		1238	Example: Initialise an C<ev_io> watcher in two steps.
		1239
		1240	ev_io w;
		1241	ev_init (&w, my_cb);
		1242	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1243
898	=item C<ev_TYPE_set> (ev_TYPE *, [args])	1244	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
899		1245
900	This macro initialises the type-specific parts of a watcher. You need to	1246	This macro initialises the type-specific parts of a watcher. You need to
901	call C<ev_init> at least once before you call this macro, but you can	1247	call C<ev_init> at least once before you call this macro, but you can
902	call C<ev_TYPE_set> any number of times. You must not, however, call this	1248	call C<ev_TYPE_set> any number of times. You must not, however, call this
903	macro on a watcher that is active (it can be pending, however, which is a	1249	macro on a watcher that is active (it can be pending, however, which is a
904	difference to the C<ev_init> macro).	1250	difference to the C<ev_init> macro).
905		1251
906	Although some watcher types do not have type-specific arguments	1252	Although some watcher types do not have type-specific arguments
907	(e.g. C<ev_prepare>) you still need to call its C<set> macro.	1253	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
908		1254
		1255	See C<ev_init>, above, for an example.
		1256
909	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])	1257	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
910		1258
911	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro	1259	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
912	calls into a single call. This is the most convenient method to initialise	1260	calls into a single call. This is the most convenient method to initialise
913	a watcher. The same limitations apply, of course.	1261	a watcher. The same limitations apply, of course.
914		1262
		1263	Example: Initialise and set an C<ev_io> watcher in one step.
		1264
		1265	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1266
915	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)	1267	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
916		1268
917	Starts (activates) the given watcher. Only active watchers will receive	1269	Starts (activates) the given watcher. Only active watchers will receive
918	events. If the watcher is already active nothing will happen.	1270	events. If the watcher is already active nothing will happen.
919		1271
		1272	Example: Start the C<ev_io> watcher that is being abused as example in this
		1273	whole section.
		1274
		1275	ev_io_start (EV_DEFAULT_UC, &w);
		1276
920	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)	1277	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
921		1278
922	Stops the given watcher again (if active) and clears the pending	1279	Stops the given watcher if active, and clears the pending status (whether
		1280	the watcher was active or not).
		1281
923	status. It is possible that stopped watchers are pending (for example,	1282	It is possible that stopped watchers are pending - for example,
924	non-repeating timers are being stopped when they become pending), but	1283	non-repeating timers are being stopped when they become pending - but
925	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If	1284	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
926	you want to free or reuse the memory used by the watcher it is therefore a	1285	pending. If you want to free or reuse the memory used by the watcher it is
927	good idea to always call its C<ev_TYPE_stop> function.	1286	therefore a good idea to always call its C<ev_TYPE_stop> function.
928		1287
929	=item bool ev_is_active (ev_TYPE *watcher)	1288	=item bool ev_is_active (ev_TYPE *watcher)
930		1289
931	Returns a true value iff the watcher is active (i.e. it has been started	1290	Returns a true value iff the watcher is active (i.e. it has been started
932	and not yet been stopped). As long as a watcher is active you must not modify	1291	and not yet been stopped). As long as a watcher is active you must not modify
…		…
948	=item ev_cb_set (ev_TYPE *watcher, callback)	1307	=item ev_cb_set (ev_TYPE *watcher, callback)
949		1308
950	Change the callback. You can change the callback at virtually any time	1309	Change the callback. You can change the callback at virtually any time
951	(modulo threads).	1310	(modulo threads).
952		1311
953	=item ev_set_priority (ev_TYPE *watcher, priority)	1312	=item ev_set_priority (ev_TYPE *watcher, int priority)
954		1313
955	=item int ev_priority (ev_TYPE *watcher)	1314	=item int ev_priority (ev_TYPE *watcher)
956		1315
957	Set and query the priority of the watcher. The priority is a small	1316	Set and query the priority of the watcher. The priority is a small
958	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>	1317	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
959	(default: C<-2>). Pending watchers with higher priority will be invoked	1318	(default: C<-2>). Pending watchers with higher priority will be invoked
960	before watchers with lower priority, but priority will not keep watchers	1319	before watchers with lower priority, but priority will not keep watchers
961	from being executed (except for C<ev_idle> watchers).	1320	from being executed (except for C<ev_idle> watchers).
962		1321
963	This means that priorities are I<only> used for ordering callback
964	invocation after new events have been received. This is useful, for
965	example, to reduce latency after idling, or more often, to bind two
966	watchers on the same event and make sure one is called first.
967
968	If you need to suppress invocation when higher priority events are pending	1322	If you need to suppress invocation when higher priority events are pending
969	you need to look at C<ev_idle> watchers, which provide this functionality.	1323	you need to look at C<ev_idle> watchers, which provide this functionality.
970		1324
971	You I<must not> change the priority of a watcher as long as it is active or	1325	You I<must not> change the priority of a watcher as long as it is active or
972	pending.	1326	pending.
973		1327
		1328	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1329	fine, as long as you do not mind that the priority value you query might
		1330	or might not have been clamped to the valid range.
		1331
974	The default priority used by watchers when no priority has been set is	1332	The default priority used by watchers when no priority has been set is
975	always C<0>, which is supposed to not be too high and not be too low :).	1333	always C<0>, which is supposed to not be too high and not be too low :).
976		1334
977	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is	1335	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
978	fine, as long as you do not mind that the priority value you query might	1336	priorities.
979	or might not have been adjusted to be within valid range.
980		1337
981	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1338	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
982		1339
983	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1340	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
984	C<loop> nor C<revents> need to be valid as long as the watcher callback	1341	C<loop> nor C<revents> need to be valid as long as the watcher callback
985	can deal with that fact.	1342	can deal with that fact, as both are simply passed through to the
		1343	callback.
986		1344
987	=item int ev_clear_pending (loop, ev_TYPE *watcher)	1345	=item int ev_clear_pending (loop, ev_TYPE *watcher)
988		1346
989	If the watcher is pending, this function returns clears its pending status	1347	If the watcher is pending, this function clears its pending status and
990	and returns its C<revents> bitset (as if its callback was invoked). If the	1348	returns its C<revents> bitset (as if its callback was invoked). If the
991	watcher isn't pending it does nothing and returns C<0>.	1349	watcher isn't pending it does nothing and returns C<0>.
992		1350
		1351	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1352	callback to be invoked, which can be accomplished with this function.
		1353
		1354	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1355
		1356	Feeds the given event set into the event loop, as if the specified event
		1357	had happened for the specified watcher (which must be a pointer to an
		1358	initialised but not necessarily started event watcher). Obviously you must
		1359	not free the watcher as long as it has pending events.
		1360
		1361	Stopping the watcher, letting libev invoke it, or calling
		1362	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1363	not started in the first place.
		1364
		1365	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1366	functions that do not need a watcher.
		1367
993	=back	1368	=back
994		1369
		1370	See also the L<ASSOCIATING CUSTOM DATA WITH A WATCHER> and L<BUILDING YOUR
		1371	OWN COMPOSITE WATCHERS> idioms.
995		1372
996	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1373	=head2 WATCHER STATES
997		1374
998	Each watcher has, by default, a member C<void *data> that you can change	1375	There are various watcher states mentioned throughout this manual -
999	and read at any time, libev will completely ignore it. This can be used	1376	active, pending and so on. In this section these states and the rules to
1000	to associate arbitrary data with your watcher. If you need more data and	1377	transition between them will be described in more detail - and while these
1001	don't want to allocate memory and store a pointer to it in that data	1378	rules might look complicated, they usually do "the right thing".
1002	member, you can also "subclass" the watcher type and provide your own
1003	data:
1004		1379
1005	struct my_io	1380	=over 4
		1381
		1382	=item initialiased
		1383
		1384	Before a watcher can be registered with the event looop it has to be
		1385	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1386	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1387
		1388	In this state it is simply some block of memory that is suitable for
		1389	use in an event loop. It can be moved around, freed, reused etc. at
		1390	will - as long as you either keep the memory contents intact, or call
		1391	C<ev_TYPE_init> again.
		1392
		1393	=item started/running/active
		1394
		1395	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1396	property of the event loop, and is actively waiting for events. While in
		1397	this state it cannot be accessed (except in a few documented ways), moved,
		1398	freed or anything else - the only legal thing is to keep a pointer to it,
		1399	and call libev functions on it that are documented to work on active watchers.
		1400
		1401	=item pending
		1402
		1403	If a watcher is active and libev determines that an event it is interested
		1404	in has occurred (such as a timer expiring), it will become pending. It will
		1405	stay in this pending state until either it is stopped or its callback is
		1406	about to be invoked, so it is not normally pending inside the watcher
		1407	callback.
		1408
		1409	The watcher might or might not be active while it is pending (for example,
		1410	an expired non-repeating timer can be pending but no longer active). If it
		1411	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1412	but it is still property of the event loop at this time, so cannot be
		1413	moved, freed or reused. And if it is active the rules described in the
		1414	previous item still apply.
		1415
		1416	It is also possible to feed an event on a watcher that is not active (e.g.
		1417	via C<ev_feed_event>), in which case it becomes pending without being
		1418	active.
		1419
		1420	=item stopped
		1421
		1422	A watcher can be stopped implicitly by libev (in which case it might still
		1423	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1424	latter will clear any pending state the watcher might be in, regardless
		1425	of whether it was active or not, so stopping a watcher explicitly before
		1426	freeing it is often a good idea.
		1427
		1428	While stopped (and not pending) the watcher is essentially in the
		1429	initialised state, that is, it can be reused, moved, modified in any way
		1430	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1431	it again).
		1432
		1433	=back
		1434
		1435	=head2 WATCHER PRIORITY MODELS
		1436
		1437	Many event loops support I<watcher priorities>, which are usually small
		1438	integers that influence the ordering of event callback invocation
		1439	between watchers in some way, all else being equal.
		1440
		1441	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1442	description for the more technical details such as the actual priority
		1443	range.
		1444
		1445	There are two common ways how these these priorities are being interpreted
		1446	by event loops:
		1447
		1448	In the more common lock-out model, higher priorities "lock out" invocation
		1449	of lower priority watchers, which means as long as higher priority
		1450	watchers receive events, lower priority watchers are not being invoked.
		1451
		1452	The less common only-for-ordering model uses priorities solely to order
		1453	callback invocation within a single event loop iteration: Higher priority
		1454	watchers are invoked before lower priority ones, but they all get invoked
		1455	before polling for new events.
		1456
		1457	Libev uses the second (only-for-ordering) model for all its watchers
		1458	except for idle watchers (which use the lock-out model).
		1459
		1460	The rationale behind this is that implementing the lock-out model for
		1461	watchers is not well supported by most kernel interfaces, and most event
		1462	libraries will just poll for the same events again and again as long as
		1463	their callbacks have not been executed, which is very inefficient in the
		1464	common case of one high-priority watcher locking out a mass of lower
		1465	priority ones.
		1466
		1467	Static (ordering) priorities are most useful when you have two or more
		1468	watchers handling the same resource: a typical usage example is having an
		1469	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1470	timeouts. Under load, data might be received while the program handles
		1471	other jobs, but since timers normally get invoked first, the timeout
		1472	handler will be executed before checking for data. In that case, giving
		1473	the timer a lower priority than the I/O watcher ensures that I/O will be
		1474	handled first even under adverse conditions (which is usually, but not
		1475	always, what you want).
		1476
		1477	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1478	will only be executed when no same or higher priority watchers have
		1479	received events, they can be used to implement the "lock-out" model when
		1480	required.
		1481
		1482	For example, to emulate how many other event libraries handle priorities,
		1483	you can associate an C<ev_idle> watcher to each such watcher, and in
		1484	the normal watcher callback, you just start the idle watcher. The real
		1485	processing is done in the idle watcher callback. This causes libev to
		1486	continuously poll and process kernel event data for the watcher, but when
		1487	the lock-out case is known to be rare (which in turn is rare :), this is
		1488	workable.
		1489
		1490	Usually, however, the lock-out model implemented that way will perform
		1491	miserably under the type of load it was designed to handle. In that case,
		1492	it might be preferable to stop the real watcher before starting the
		1493	idle watcher, so the kernel will not have to process the event in case
		1494	the actual processing will be delayed for considerable time.
		1495
		1496	Here is an example of an I/O watcher that should run at a strictly lower
		1497	priority than the default, and which should only process data when no
		1498	other events are pending:
		1499
		1500	ev_idle idle; // actual processing watcher
		1501	ev_io io; // actual event watcher
		1502
		1503	static void
		1504	io_cb (EV_P_ ev_io *w, int revents)
1006	{	1505	{
1007	struct ev_io io;	1506	// stop the I/O watcher, we received the event, but
1008	int otherfd;	1507	// are not yet ready to handle it.
1009	void *somedata;	1508	ev_io_stop (EV_A_ w);
1010	struct whatever *mostinteresting;	1509
		1510	// start the idle watcher to handle the actual event.
		1511	// it will not be executed as long as other watchers
		1512	// with the default priority are receiving events.
		1513	ev_idle_start (EV_A_ &idle);
1011	}	1514	}
1012		1515
1013	And since your callback will be called with a pointer to the watcher, you	1516	static void
1014	can cast it back to your own type:	1517	idle_cb (EV_P_ ev_idle *w, int revents)
1015
1016	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
1017	{	1518	{
1018	struct my_io *w = (struct my_io *)w_;	1519	// actual processing
1019	...	1520	read (STDIN_FILENO, ...);
		1521
		1522	// have to start the I/O watcher again, as
		1523	// we have handled the event
		1524	ev_io_start (EV_P_ &io);
1020	}	1525	}
1021		1526
1022	More interesting and less C-conformant ways of casting your callback type	1527	// initialisation
1023	instead have been omitted.	1528	ev_idle_init (&idle, idle_cb);
		1529	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1530	ev_io_start (EV_DEFAULT_ &io);
1024		1531
1025	Another common scenario is having some data structure with multiple	1532	In the "real" world, it might also be beneficial to start a timer, so that
1026	watchers:	1533	low-priority connections can not be locked out forever under load. This
1027		1534	enables your program to keep a lower latency for important connections
1028	struct my_biggy	1535	during short periods of high load, while not completely locking out less
1029	{	1536	important ones.
1030	int some_data;
1031	ev_timer t1;
1032	ev_timer t2;
1033	}
1034
1035	In this case getting the pointer to C<my_biggy> is a bit more complicated,
1036	you need to use C<offsetof>:
1037
1038	#include <stddef.h>
1039
1040	static void
1041	t1_cb (EV_P_ struct ev_timer *w, int revents)
1042	{
1043	struct my_biggy big = (struct my_biggy *
1044	(((char *)w) - offsetof (struct my_biggy, t1));
1045	}
1046
1047	static void
1048	t2_cb (EV_P_ struct ev_timer *w, int revents)
1049	{
1050	struct my_biggy big = (struct my_biggy *
1051	(((char *)w) - offsetof (struct my_biggy, t2));
1052	}
1053		1537
1054		1538
1055	=head1 WATCHER TYPES	1539	=head1 WATCHER TYPES
1056		1540
1057	This section describes each watcher in detail, but will not repeat	1541	This section describes each watcher in detail, but will not repeat
…		…
1081	In general you can register as many read and/or write event watchers per	1565	In general you can register as many read and/or write event watchers per
1082	fd as you want (as long as you don't confuse yourself). Setting all file	1566	fd as you want (as long as you don't confuse yourself). Setting all file
1083	descriptors to non-blocking mode is also usually a good idea (but not	1567	descriptors to non-blocking mode is also usually a good idea (but not
1084	required if you know what you are doing).	1568	required if you know what you are doing).
1085		1569
1086	If you must do this, then force the use of a known-to-be-good backend
1087	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
1088	C<EVBACKEND_POLL>).
1089
1090	Another thing you have to watch out for is that it is quite easy to	1570	Another thing you have to watch out for is that it is quite easy to
1091	receive "spurious" readiness notifications, that is your callback might	1571	receive "spurious" readiness notifications, that is, your callback might
1092	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1572	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1093	because there is no data. Not only are some backends known to create a	1573	because there is no data. It is very easy to get into this situation even
1094	lot of those (for example Solaris ports), it is very easy to get into	1574	with a relatively standard program structure. Thus it is best to always
1095	this situation even with a relatively standard program structure. Thus	1575	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
1096	it is best to always use non-blocking I/O: An extra C<read>(2) returning
1097	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1576	preferable to a program hanging until some data arrives.
1098		1577
1099	If you cannot run the fd in non-blocking mode (for example you should not	1578	If you cannot run the fd in non-blocking mode (for example you should
1100	play around with an Xlib connection), then you have to separately re-test	1579	not play around with an Xlib connection), then you have to separately
1101	whether a file descriptor is really ready with a known-to-be good interface	1580	re-test whether a file descriptor is really ready with a known-to-be good
1102	such as poll (fortunately in our Xlib example, Xlib already does this on	1581	interface such as poll (fortunately in the case of Xlib, it already does
1103	its own, so its quite safe to use).	1582	this on its own, so its quite safe to use). Some people additionally
		1583	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1584	indefinitely.
		1585
		1586	But really, best use non-blocking mode.
1104		1587
1105	=head3 The special problem of disappearing file descriptors	1588	=head3 The special problem of disappearing file descriptors
1106		1589
1107	Some backends (e.g. kqueue, epoll) need to be told about closing a file	1590	Some backends (e.g. kqueue, epoll) need to be told about closing a file
1108	descriptor (either by calling C<close> explicitly or by any other means,	1591	descriptor (either due to calling C<close> explicitly or any other means,
1109	such as C<dup>). The reason is that you register interest in some file	1592	such as C<dup2>). The reason is that you register interest in some file
1110	descriptor, but when it goes away, the operating system will silently drop	1593	descriptor, but when it goes away, the operating system will silently drop
1111	this interest. If another file descriptor with the same number then is	1594	this interest. If another file descriptor with the same number then is
1112	registered with libev, there is no efficient way to see that this is, in	1595	registered with libev, there is no efficient way to see that this is, in
1113	fact, a different file descriptor.	1596	fact, a different file descriptor.
1114		1597
…		…
1132		1615
1133	There is no workaround possible except not registering events	1616	There is no workaround possible except not registering events
1134	for potentially C<dup ()>'ed file descriptors, or to resort to	1617	for potentially C<dup ()>'ed file descriptors, or to resort to
1135	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.	1618	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1136		1619
		1620	=head3 The special problem of files
		1621
		1622	Many people try to use C<select> (or libev) on file descriptors
		1623	representing files, and expect it to become ready when their program
		1624	doesn't block on disk accesses (which can take a long time on their own).
		1625
		1626	However, this cannot ever work in the "expected" way - you get a readiness
		1627	notification as soon as the kernel knows whether and how much data is
		1628	there, and in the case of open files, that's always the case, so you
		1629	always get a readiness notification instantly, and your read (or possibly
		1630	write) will still block on the disk I/O.
		1631
		1632	Another way to view it is that in the case of sockets, pipes, character
		1633	devices and so on, there is another party (the sender) that delivers data
		1634	on its own, but in the case of files, there is no such thing: the disk
		1635	will not send data on its own, simply because it doesn't know what you
		1636	wish to read - you would first have to request some data.
		1637
		1638	Since files are typically not-so-well supported by advanced notification
		1639	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1640	to files, even though you should not use it. The reason for this is
		1641	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1642	usually a tty, often a pipe, but also sometimes files or special devices
		1643	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1644	F</dev/urandom>), and even though the file might better be served with
		1645	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1646	it "just works" instead of freezing.
		1647
		1648	So avoid file descriptors pointing to files when you know it (e.g. use
		1649	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1650	when you rarely read from a file instead of from a socket, and want to
		1651	reuse the same code path.
		1652
1137	=head3 The special problem of fork	1653	=head3 The special problem of fork
1138		1654
1139	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit	1655	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
1140	useless behaviour. Libev fully supports fork, but needs to be told about	1656	useless behaviour. Libev fully supports fork, but needs to be told about
1141	it in the child.	1657	it in the child if you want to continue to use it in the child.
1142		1658
1143	To support fork in your programs, you either have to call	1659	To support fork in your child processes, you have to call C<ev_loop_fork
1144	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,	1660	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
1145	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or	1661	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
1146	C<EVBACKEND_POLL>.
1147		1662
1148	=head3 The special problem of SIGPIPE	1663	=head3 The special problem of SIGPIPE
1149		1664
1150	While not really specific to libev, it is easy to forget about SIGPIPE:	1665	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
1151	when writing to a pipe whose other end has been closed, your program gets	1666	when writing to a pipe whose other end has been closed, your program gets
1152	send a SIGPIPE, which, by default, aborts your program. For most programs	1667	sent a SIGPIPE, which, by default, aborts your program. For most programs
1153	this is sensible behaviour, for daemons, this is usually undesirable.	1668	this is sensible behaviour, for daemons, this is usually undesirable.
1154		1669
1155	So when you encounter spurious, unexplained daemon exits, make sure you	1670	So when you encounter spurious, unexplained daemon exits, make sure you
1156	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon	1671	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
1157	somewhere, as that would have given you a big clue).	1672	somewhere, as that would have given you a big clue).
1158		1673
		1674	=head3 The special problem of accept()ing when you can't
		1675
		1676	Many implementations of the POSIX C<accept> function (for example,
		1677	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1678	connection from the pending queue in all error cases.
		1679
		1680	For example, larger servers often run out of file descriptors (because
		1681	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1682	rejecting the connection, leading to libev signalling readiness on
		1683	the next iteration again (the connection still exists after all), and
		1684	typically causing the program to loop at 100% CPU usage.
		1685
		1686	Unfortunately, the set of errors that cause this issue differs between
		1687	operating systems, there is usually little the app can do to remedy the
		1688	situation, and no known thread-safe method of removing the connection to
		1689	cope with overload is known (to me).
		1690
		1691	One of the easiest ways to handle this situation is to just ignore it
		1692	- when the program encounters an overload, it will just loop until the
		1693	situation is over. While this is a form of busy waiting, no OS offers an
		1694	event-based way to handle this situation, so it's the best one can do.
		1695
		1696	A better way to handle the situation is to log any errors other than
		1697	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1698	messages, and continue as usual, which at least gives the user an idea of
		1699	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1700	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1701	usage.
		1702
		1703	If your program is single-threaded, then you could also keep a dummy file
		1704	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1705	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1706	close that fd, and create a new dummy fd. This will gracefully refuse
		1707	clients under typical overload conditions.
		1708
		1709	The last way to handle it is to simply log the error and C<exit>, as
		1710	is often done with C<malloc> failures, but this results in an easy
		1711	opportunity for a DoS attack.
1159		1712
1160	=head3 Watcher-Specific Functions	1713	=head3 Watcher-Specific Functions
1161		1714
1162	=over 4	1715	=over 4
1163		1716
1164	=item ev_io_init (ev_io *, callback, int fd, int events)	1717	=item ev_io_init (ev_io *, callback, int fd, int events)
1165		1718
1166	=item ev_io_set (ev_io *, int fd, int events)	1719	=item ev_io_set (ev_io *, int fd, int events)
1167		1720
1168	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to	1721	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
1169	receive events for and events is either C<EV_READ>, C<EV_WRITE> or	1722	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
1170	C<EV_READ | EV_WRITE> to receive the given events.	1723	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
1171		1724
1172	=item int fd [read-only]	1725	=item int fd [read-only]
1173		1726
1174	The file descriptor being watched.	1727	The file descriptor being watched.
1175		1728
…		…
1184	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well	1737	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
1185	readable, but only once. Since it is likely line-buffered, you could	1738	readable, but only once. Since it is likely line-buffered, you could
1186	attempt to read a whole line in the callback.	1739	attempt to read a whole line in the callback.
1187		1740
1188	static void	1741	static void
1189	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1742	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
1190	{	1743	{
1191	ev_io_stop (loop, w);	1744	ev_io_stop (loop, w);
1192	.. read from stdin here (or from w->fd) and haqndle any I/O errors	1745	.. read from stdin here (or from w->fd) and handle any I/O errors
1193	}	1746	}
1194		1747
1195	...	1748	...
1196	struct ev_loop *loop = ev_default_init (0);	1749	struct ev_loop *loop = ev_default_init (0);
1197	struct ev_io stdin_readable;	1750	ev_io stdin_readable;
1198	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);	1751	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
1199	ev_io_start (loop, &stdin_readable);	1752	ev_io_start (loop, &stdin_readable);
1200	ev_loop (loop, 0);	1753	ev_run (loop, 0);
1201		1754
1202		1755
1203	=head2 C<ev_timer> - relative and optionally repeating timeouts	1756	=head2 C<ev_timer> - relative and optionally repeating timeouts
1204		1757
1205	Timer watchers are simple relative timers that generate an event after a	1758	Timer watchers are simple relative timers that generate an event after a
1206	given time, and optionally repeating in regular intervals after that.	1759	given time, and optionally repeating in regular intervals after that.
1207		1760
1208	The timers are based on real time, that is, if you register an event that	1761	The timers are based on real time, that is, if you register an event that
1209	times out after an hour and you reset your system clock to January last	1762	times out after an hour and you reset your system clock to January last
1210	year, it will still time out after (roughly) and hour. "Roughly" because	1763	year, it will still time out after (roughly) one hour. "Roughly" because
1211	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1764	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1212	monotonic clock option helps a lot here).	1765	monotonic clock option helps a lot here).
1213		1766
1214	The callback is guaranteed to be invoked only after its timeout has passed,	1767	The callback is guaranteed to be invoked only I<after> its timeout has
1215	but if multiple timers become ready during the same loop iteration then	1768	passed (not I<at>, so on systems with very low-resolution clocks this
1216	order of execution is undefined.	1769	might introduce a small delay). If multiple timers become ready during the
		1770	same loop iteration then the ones with earlier time-out values are invoked
		1771	before ones of the same priority with later time-out values (but this is
		1772	no longer true when a callback calls C<ev_run> recursively).
		1773
		1774	=head3 Be smart about timeouts
		1775
		1776	Many real-world problems involve some kind of timeout, usually for error
		1777	recovery. A typical example is an HTTP request - if the other side hangs,
		1778	you want to raise some error after a while.
		1779
		1780	What follows are some ways to handle this problem, from obvious and
		1781	inefficient to smart and efficient.
		1782
		1783	In the following, a 60 second activity timeout is assumed - a timeout that
		1784	gets reset to 60 seconds each time there is activity (e.g. each time some
		1785	data or other life sign was received).
		1786
		1787	=over 4
		1788
		1789	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1790
		1791	This is the most obvious, but not the most simple way: In the beginning,
		1792	start the watcher:
		1793
		1794	ev_timer_init (timer, callback, 60., 0.);
		1795	ev_timer_start (loop, timer);
		1796
		1797	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1798	and start it again:
		1799
		1800	ev_timer_stop (loop, timer);
		1801	ev_timer_set (timer, 60., 0.);
		1802	ev_timer_start (loop, timer);
		1803
		1804	This is relatively simple to implement, but means that each time there is
		1805	some activity, libev will first have to remove the timer from its internal
		1806	data structure and then add it again. Libev tries to be fast, but it's
		1807	still not a constant-time operation.
		1808
		1809	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1810
		1811	This is the easiest way, and involves using C<ev_timer_again> instead of
		1812	C<ev_timer_start>.
		1813
		1814	To implement this, configure an C<ev_timer> with a C<repeat> value
		1815	of C<60> and then call C<ev_timer_again> at start and each time you
		1816	successfully read or write some data. If you go into an idle state where
		1817	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1818	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1819
		1820	That means you can ignore both the C<ev_timer_start> function and the
		1821	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1822	member and C<ev_timer_again>.
		1823
		1824	At start:
		1825
		1826	ev_init (timer, callback);
		1827	timer->repeat = 60.;
		1828	ev_timer_again (loop, timer);
		1829
		1830	Each time there is some activity:
		1831
		1832	ev_timer_again (loop, timer);
		1833
		1834	It is even possible to change the time-out on the fly, regardless of
		1835	whether the watcher is active or not:
		1836
		1837	timer->repeat = 30.;
		1838	ev_timer_again (loop, timer);
		1839
		1840	This is slightly more efficient then stopping/starting the timer each time
		1841	you want to modify its timeout value, as libev does not have to completely
		1842	remove and re-insert the timer from/into its internal data structure.
		1843
		1844	It is, however, even simpler than the "obvious" way to do it.
		1845
		1846	=item 3. Let the timer time out, but then re-arm it as required.
		1847
		1848	This method is more tricky, but usually most efficient: Most timeouts are
		1849	relatively long compared to the intervals between other activity - in
		1850	our example, within 60 seconds, there are usually many I/O events with
		1851	associated activity resets.
		1852
		1853	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1854	but remember the time of last activity, and check for a real timeout only
		1855	within the callback:
		1856
		1857	ev_tstamp last_activity; // time of last activity
		1858
		1859	static void
		1860	callback (EV_P_ ev_timer *w, int revents)
		1861	{
		1862	ev_tstamp now = ev_now (EV_A);
		1863	ev_tstamp timeout = last_activity + 60.;
		1864
		1865	// if last_activity + 60. is older than now, we did time out
		1866	if (timeout < now)
		1867	{
		1868	// timeout occurred, take action
		1869	}
		1870	else
		1871	{
		1872	// callback was invoked, but there was some activity, re-arm
		1873	// the watcher to fire in last_activity + 60, which is
		1874	// guaranteed to be in the future, so "again" is positive:
		1875	w->repeat = timeout - now;
		1876	ev_timer_again (EV_A_ w);
		1877	}
		1878	}
		1879
		1880	To summarise the callback: first calculate the real timeout (defined
		1881	as "60 seconds after the last activity"), then check if that time has
		1882	been reached, which means something I<did>, in fact, time out. Otherwise
		1883	the callback was invoked too early (C<timeout> is in the future), so
		1884	re-schedule the timer to fire at that future time, to see if maybe we have
		1885	a timeout then.
		1886
		1887	Note how C<ev_timer_again> is used, taking advantage of the
		1888	C<ev_timer_again> optimisation when the timer is already running.
		1889
		1890	This scheme causes more callback invocations (about one every 60 seconds
		1891	minus half the average time between activity), but virtually no calls to
		1892	libev to change the timeout.
		1893
		1894	To start the timer, simply initialise the watcher and set C<last_activity>
		1895	to the current time (meaning we just have some activity :), then call the
		1896	callback, which will "do the right thing" and start the timer:
		1897
		1898	ev_init (timer, callback);
		1899	last_activity = ev_now (loop);
		1900	callback (loop, timer, EV_TIMER);
		1901
		1902	And when there is some activity, simply store the current time in
		1903	C<last_activity>, no libev calls at all:
		1904
		1905	last_activity = ev_now (loop);
		1906
		1907	This technique is slightly more complex, but in most cases where the
		1908	time-out is unlikely to be triggered, much more efficient.
		1909
		1910	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1911	callback :) - just change the timeout and invoke the callback, which will
		1912	fix things for you.
		1913
		1914	=item 4. Wee, just use a double-linked list for your timeouts.
		1915
		1916	If there is not one request, but many thousands (millions...), all
		1917	employing some kind of timeout with the same timeout value, then one can
		1918	do even better:
		1919
		1920	When starting the timeout, calculate the timeout value and put the timeout
		1921	at the I<end> of the list.
		1922
		1923	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1924	the list is expected to fire (for example, using the technique #3).
		1925
		1926	When there is some activity, remove the timer from the list, recalculate
		1927	the timeout, append it to the end of the list again, and make sure to
		1928	update the C<ev_timer> if it was taken from the beginning of the list.
		1929
		1930	This way, one can manage an unlimited number of timeouts in O(1) time for
		1931	starting, stopping and updating the timers, at the expense of a major
		1932	complication, and having to use a constant timeout. The constant timeout
		1933	ensures that the list stays sorted.
		1934
		1935	=back
		1936
		1937	So which method the best?
		1938
		1939	Method #2 is a simple no-brain-required solution that is adequate in most
		1940	situations. Method #3 requires a bit more thinking, but handles many cases
		1941	better, and isn't very complicated either. In most case, choosing either
		1942	one is fine, with #3 being better in typical situations.
		1943
		1944	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1945	rather complicated, but extremely efficient, something that really pays
		1946	off after the first million or so of active timers, i.e. it's usually
		1947	overkill :)
1217		1948
1218	=head3 The special problem of time updates	1949	=head3 The special problem of time updates
1219		1950
1220	Establishing the current time is a costly operation (it usually takes at	1951	Establishing the current time is a costly operation (it usually takes at
1221	least two system calls): EV therefore updates its idea of the current	1952	least two system calls): EV therefore updates its idea of the current
1222	time only before and after C<ev_loop> polls for new events, which causes	1953	time only before and after C<ev_run> collects new events, which causes a
1223	a growing difference between C<ev_now ()> and C<ev_time ()> when handling	1954	growing difference between C<ev_now ()> and C<ev_time ()> when handling
1224	lots of events.	1955	lots of events in one iteration.
1225		1956
1226	The relative timeouts are calculated relative to the C<ev_now ()>	1957	The relative timeouts are calculated relative to the C<ev_now ()>
1227	time. This is usually the right thing as this timestamp refers to the time	1958	time. This is usually the right thing as this timestamp refers to the time
1228	of the event triggering whatever timeout you are modifying/starting. If	1959	of the event triggering whatever timeout you are modifying/starting. If
1229	you suspect event processing to be delayed and you I<need> to base the	1960	you suspect event processing to be delayed and you I<need> to base the
1230	timeout on the current time, use something like this to adjust for this:	1961	timeout on the current time, use something like this to adjust for this:
1231		1962
1232	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1963	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1233		1964
1234	If the event loop is suspended for a long time, one can also force an	1965	If the event loop is suspended for a long time, you can also force an
1235	update of the time returned by C<ev_now ()> by calling C<ev_now_update	1966	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1236	()>.	1967	()>.
		1968
		1969	=head3 The special problems of suspended animation
		1970
		1971	When you leave the server world it is quite customary to hit machines that
		1972	can suspend/hibernate - what happens to the clocks during such a suspend?
		1973
		1974	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1975	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1976	to run until the system is suspended, but they will not advance while the
		1977	system is suspended. That means, on resume, it will be as if the program
		1978	was frozen for a few seconds, but the suspend time will not be counted
		1979	towards C<ev_timer> when a monotonic clock source is used. The real time
		1980	clock advanced as expected, but if it is used as sole clocksource, then a
		1981	long suspend would be detected as a time jump by libev, and timers would
		1982	be adjusted accordingly.
		1983
		1984	I would not be surprised to see different behaviour in different between
		1985	operating systems, OS versions or even different hardware.
		1986
		1987	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1988	time jump in the monotonic clocks and the realtime clock. If the program
		1989	is suspended for a very long time, and monotonic clock sources are in use,
		1990	then you can expect C<ev_timer>s to expire as the full suspension time
		1991	will be counted towards the timers. When no monotonic clock source is in
		1992	use, then libev will again assume a timejump and adjust accordingly.
		1993
		1994	It might be beneficial for this latter case to call C<ev_suspend>
		1995	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1996	deterministic behaviour in this case (you can do nothing against
		1997	C<SIGSTOP>).
1237		1998
1238	=head3 Watcher-Specific Functions and Data Members	1999	=head3 Watcher-Specific Functions and Data Members
1239		2000
1240	=over 4	2001	=over 4
1241		2002
…		…
1265	If the timer is started but non-repeating, stop it (as if it timed out).	2026	If the timer is started but non-repeating, stop it (as if it timed out).
1266		2027
1267	If the timer is repeating, either start it if necessary (with the	2028	If the timer is repeating, either start it if necessary (with the
1268	C<repeat> value), or reset the running timer to the C<repeat> value.	2029	C<repeat> value), or reset the running timer to the C<repeat> value.
1269		2030
1270	This sounds a bit complicated, but here is a useful and typical	2031	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
1271	example: Imagine you have a TCP connection and you want a so-called idle	2032	usage example.
1272	timeout, that is, you want to be called when there have been, say, 60
1273	seconds of inactivity on the socket. The easiest way to do this is to
1274	configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1275	C<ev_timer_again> each time you successfully read or write some data. If
1276	you go into an idle state where you do not expect data to travel on the
1277	socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1278	automatically restart it if need be.
1279		2033
1280	That means you can ignore the C<after> value and C<ev_timer_start>	2034	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
1281	altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1282		2035
1283	ev_timer_init (timer, callback, 0., 5.);	2036	Returns the remaining time until a timer fires. If the timer is active,
1284	ev_timer_again (loop, timer);	2037	then this time is relative to the current event loop time, otherwise it's
1285	...	2038	the timeout value currently configured.
1286	timer->again = 17.;
1287	ev_timer_again (loop, timer);
1288	...
1289	timer->again = 10.;
1290	ev_timer_again (loop, timer);
1291		2039
1292	This is more slightly efficient then stopping/starting the timer each time	2040	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
1293	you want to modify its timeout value.	2041	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2042	will return C<4>. When the timer expires and is restarted, it will return
		2043	roughly C<7> (likely slightly less as callback invocation takes some time,
		2044	too), and so on.
1294		2045
1295	=item ev_tstamp repeat [read-write]	2046	=item ev_tstamp repeat [read-write]
1296		2047
1297	The current C<repeat> value. Will be used each time the watcher times out	2048	The current C<repeat> value. Will be used each time the watcher times out
1298	or C<ev_timer_again> is called and determines the next timeout (if any),	2049	or C<ev_timer_again> is called, and determines the next timeout (if any),
1299	which is also when any modifications are taken into account.	2050	which is also when any modifications are taken into account.
1300		2051
1301	=back	2052	=back
1302		2053
1303	=head3 Examples	2054	=head3 Examples
1304		2055
1305	Example: Create a timer that fires after 60 seconds.	2056	Example: Create a timer that fires after 60 seconds.
1306		2057
1307	static void	2058	static void
1308	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2059	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
1309	{	2060	{
1310	.. one minute over, w is actually stopped right here	2061	.. one minute over, w is actually stopped right here
1311	}	2062	}
1312		2063
1313	struct ev_timer mytimer;	2064	ev_timer mytimer;
1314	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);	2065	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1315	ev_timer_start (loop, &mytimer);	2066	ev_timer_start (loop, &mytimer);
1316		2067
1317	Example: Create a timeout timer that times out after 10 seconds of	2068	Example: Create a timeout timer that times out after 10 seconds of
1318	inactivity.	2069	inactivity.
1319		2070
1320	static void	2071	static void
1321	timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2072	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
1322	{	2073	{
1323	.. ten seconds without any activity	2074	.. ten seconds without any activity
1324	}	2075	}
1325		2076
1326	struct ev_timer mytimer;	2077	ev_timer mytimer;
1327	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */	2078	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1328	ev_timer_again (&mytimer); /* start timer */	2079	ev_timer_again (&mytimer); /* start timer */
1329	ev_loop (loop, 0);	2080	ev_run (loop, 0);
1330		2081
1331	// and in some piece of code that gets executed on any "activity":	2082	// and in some piece of code that gets executed on any "activity":
1332	// reset the timeout to start ticking again at 10 seconds	2083	// reset the timeout to start ticking again at 10 seconds
1333	ev_timer_again (&mytimer);	2084	ev_timer_again (&mytimer);
1334		2085
…		…
1336	=head2 C<ev_periodic> - to cron or not to cron?	2087	=head2 C<ev_periodic> - to cron or not to cron?
1337		2088
1338	Periodic watchers are also timers of a kind, but they are very versatile	2089	Periodic watchers are also timers of a kind, but they are very versatile
1339	(and unfortunately a bit complex).	2090	(and unfortunately a bit complex).
1340		2091
1341	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2092	Unlike C<ev_timer>, periodic watchers are not based on real time (or
1342	but on wall clock time (absolute time). You can tell a periodic watcher	2093	relative time, the physical time that passes) but on wall clock time
1343	to trigger after some specific point in time. For example, if you tell a	2094	(absolute time, the thing you can read on your calender or clock). The
1344	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()	2095	difference is that wall clock time can run faster or slower than real
1345	+ 10.>, that is, an absolute time not a delay) and then reset your system	2096	time, and time jumps are not uncommon (e.g. when you adjust your
1346	clock to January of the previous year, then it will take more than year	2097	wrist-watch).
1347	to trigger the event (unlike an C<ev_timer>, which would still trigger
1348	roughly 10 seconds later as it uses a relative timeout).
1349		2098
		2099	You can tell a periodic watcher to trigger after some specific point
		2100	in time: for example, if you tell a periodic watcher to trigger "in 10
		2101	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2102	not a delay) and then reset your system clock to January of the previous
		2103	year, then it will take a year or more to trigger the event (unlike an
		2104	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2105	it, as it uses a relative timeout).
		2106
1350	C<ev_periodic>s can also be used to implement vastly more complex timers,	2107	C<ev_periodic> watchers can also be used to implement vastly more complex
1351	such as triggering an event on each "midnight, local time", or other	2108	timers, such as triggering an event on each "midnight, local time", or
1352	complicated, rules.	2109	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		2110	those cannot react to time jumps.
1353		2111
1354	As with timers, the callback is guaranteed to be invoked only when the	2112	As with timers, the callback is guaranteed to be invoked only when the
1355	time (C<at>) has passed, but if multiple periodic timers become ready	2113	point in time where it is supposed to trigger has passed. If multiple
1356	during the same loop iteration then order of execution is undefined.	2114	timers become ready during the same loop iteration then the ones with
		2115	earlier time-out values are invoked before ones with later time-out values
		2116	(but this is no longer true when a callback calls C<ev_run> recursively).
1357		2117
1358	=head3 Watcher-Specific Functions and Data Members	2118	=head3 Watcher-Specific Functions and Data Members
1359		2119
1360	=over 4	2120	=over 4
1361		2121
1362	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2122	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1363		2123
1364	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2124	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1365		2125
1366	Lots of arguments, lets sort it out... There are basically three modes of	2126	Lots of arguments, let's sort it out... There are basically three modes of
1367	operation, and we will explain them from simplest to complex:	2127	operation, and we will explain them from simplest to most complex:
1368		2128
1369	=over 4	2129	=over 4
1370		2130
1371	=item * absolute timer (at = time, interval = reschedule_cb = 0)	2131	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
1372		2132
1373	In this configuration the watcher triggers an event after the wall clock	2133	In this configuration the watcher triggers an event after the wall clock
1374	time C<at> has passed and doesn't repeat. It will not adjust when a time	2134	time C<offset> has passed. It will not repeat and will not adjust when a
1375	jump occurs, that is, if it is to be run at January 1st 2011 then it will	2135	time jump occurs, that is, if it is to be run at January 1st 2011 then it
1376	run when the system time reaches or surpasses this time.	2136	will be stopped and invoked when the system clock reaches or surpasses
		2137	this point in time.
1377		2138
1378	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	2139	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
1379		2140
1380	In this mode the watcher will always be scheduled to time out at the next	2141	In this mode the watcher will always be scheduled to time out at the next
1381	C<at + N * interval> time (for some integer N, which can also be negative)	2142	C<offset + N * interval> time (for some integer N, which can also be
1382	and then repeat, regardless of any time jumps.	2143	negative) and then repeat, regardless of any time jumps. The C<offset>
		2144	argument is merely an offset into the C<interval> periods.
1383		2145
1384	This can be used to create timers that do not drift with respect to system	2146	This can be used to create timers that do not drift with respect to the
1385	time, for example, here is a C<ev_periodic> that triggers each hour, on	2147	system clock, for example, here is an C<ev_periodic> that triggers each
1386	the hour:	2148	hour, on the hour (with respect to UTC):
1387		2149
1388	ev_periodic_set (&periodic, 0., 3600., 0);	2150	ev_periodic_set (&periodic, 0., 3600., 0);
1389		2151
1390	This doesn't mean there will always be 3600 seconds in between triggers,	2152	This doesn't mean there will always be 3600 seconds in between triggers,
1391	but only that the callback will be called when the system time shows a	2153	but only that the callback will be called when the system time shows a
1392	full hour (UTC), or more correctly, when the system time is evenly divisible	2154	full hour (UTC), or more correctly, when the system time is evenly divisible
1393	by 3600.	2155	by 3600.
1394		2156
1395	Another way to think about it (for the mathematically inclined) is that	2157	Another way to think about it (for the mathematically inclined) is that
1396	C<ev_periodic> will try to run the callback in this mode at the next possible	2158	C<ev_periodic> will try to run the callback in this mode at the next possible
1397	time where C<time = at (mod interval)>, regardless of any time jumps.	2159	time where C<time = offset (mod interval)>, regardless of any time jumps.
1398		2160
1399	For numerical stability it is preferable that the C<at> value is near	2161	The C<interval> I<MUST> be positive, and for numerical stability, the
1400	C<ev_now ()> (the current time), but there is no range requirement for	2162	interval value should be higher than C<1/8192> (which is around 100
1401	this value, and in fact is often specified as zero.	2163	microseconds) and C<offset> should be higher than C<0> and should have
		2164	at most a similar magnitude as the current time (say, within a factor of
		2165	ten). Typical values for offset are, in fact, C<0> or something between
		2166	C<0> and C<interval>, which is also the recommended range.
1402		2167
1403	Note also that there is an upper limit to how often a timer can fire (CPU	2168	Note also that there is an upper limit to how often a timer can fire (CPU
1404	speed for example), so if C<interval> is very small then timing stability	2169	speed for example), so if C<interval> is very small then timing stability
1405	will of course deteriorate. Libev itself tries to be exact to be about one	2170	will of course deteriorate. Libev itself tries to be exact to be about one
1406	millisecond (if the OS supports it and the machine is fast enough).	2171	millisecond (if the OS supports it and the machine is fast enough).
1407		2172
1408	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)	2173	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
1409		2174
1410	In this mode the values for C<interval> and C<at> are both being	2175	In this mode the values for C<interval> and C<offset> are both being
1411	ignored. Instead, each time the periodic watcher gets scheduled, the	2176	ignored. Instead, each time the periodic watcher gets scheduled, the
1412	reschedule callback will be called with the watcher as first, and the	2177	reschedule callback will be called with the watcher as first, and the
1413	current time as second argument.	2178	current time as second argument.
1414		2179
1415	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2180	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
1416	ever, or make ANY event loop modifications whatsoever>.	2181	or make ANY other event loop modifications whatsoever, unless explicitly
		2182	allowed by documentation here>.
1417		2183
1418	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop	2184	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
1419	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the	2185	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
1420	only event loop modification you are allowed to do).	2186	only event loop modification you are allowed to do).
1421		2187
1422	The callback prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic	2188	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
1423	*w, ev_tstamp now)>, e.g.:	2189	*w, ev_tstamp now)>, e.g.:
1424		2190
		2191	static ev_tstamp
1425	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2192	my_rescheduler (ev_periodic *w, ev_tstamp now)
1426	{	2193	{
1427	return now + 60.;	2194	return now + 60.;
1428	}	2195	}
1429		2196
1430	It must return the next time to trigger, based on the passed time value	2197	It must return the next time to trigger, based on the passed time value
…		…
1450	a different time than the last time it was called (e.g. in a crond like	2217	a different time than the last time it was called (e.g. in a crond like
1451	program when the crontabs have changed).	2218	program when the crontabs have changed).
1452		2219
1453	=item ev_tstamp ev_periodic_at (ev_periodic *)	2220	=item ev_tstamp ev_periodic_at (ev_periodic *)
1454		2221
1455	When active, returns the absolute time that the watcher is supposed to	2222	When active, returns the absolute time that the watcher is supposed
1456	trigger next.	2223	to trigger next. This is not the same as the C<offset> argument to
		2224	C<ev_periodic_set>, but indeed works even in interval and manual
		2225	rescheduling modes.
1457		2226
1458	=item ev_tstamp offset [read-write]	2227	=item ev_tstamp offset [read-write]
1459		2228
1460	When repeating, this contains the offset value, otherwise this is the	2229	When repeating, this contains the offset value, otherwise this is the
1461	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	2230	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2231	although libev might modify this value for better numerical stability).
1462		2232
1463	Can be modified any time, but changes only take effect when the periodic	2233	Can be modified any time, but changes only take effect when the periodic
1464	timer fires or C<ev_periodic_again> is being called.	2234	timer fires or C<ev_periodic_again> is being called.
1465		2235
1466	=item ev_tstamp interval [read-write]	2236	=item ev_tstamp interval [read-write]
1467		2237
1468	The current interval value. Can be modified any time, but changes only	2238	The current interval value. Can be modified any time, but changes only
1469	take effect when the periodic timer fires or C<ev_periodic_again> is being	2239	take effect when the periodic timer fires or C<ev_periodic_again> is being
1470	called.	2240	called.
1471		2241
1472	=item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]	2242	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
1473		2243
1474	The current reschedule callback, or C<0>, if this functionality is	2244	The current reschedule callback, or C<0>, if this functionality is
1475	switched off. Can be changed any time, but changes only take effect when	2245	switched off. Can be changed any time, but changes only take effect when
1476	the periodic timer fires or C<ev_periodic_again> is being called.	2246	the periodic timer fires or C<ev_periodic_again> is being called.
1477		2247
1478	=back	2248	=back
1479		2249
1480	=head3 Examples	2250	=head3 Examples
1481		2251
1482	Example: Call a callback every hour, or, more precisely, whenever the	2252	Example: Call a callback every hour, or, more precisely, whenever the
1483	system clock is divisible by 3600. The callback invocation times have	2253	system time is divisible by 3600. The callback invocation times have
1484	potentially a lot of jitter, but good long-term stability.	2254	potentially a lot of jitter, but good long-term stability.
1485		2255
1486	static void	2256	static void
1487	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)	2257	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
1488	{	2258	{
1489	... its now a full hour (UTC, or TAI or whatever your clock follows)	2259	... its now a full hour (UTC, or TAI or whatever your clock follows)
1490	}	2260	}
1491		2261
1492	struct ev_periodic hourly_tick;	2262	ev_periodic hourly_tick;
1493	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);	2263	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1494	ev_periodic_start (loop, &hourly_tick);	2264	ev_periodic_start (loop, &hourly_tick);
1495		2265
1496	Example: The same as above, but use a reschedule callback to do it:	2266	Example: The same as above, but use a reschedule callback to do it:
1497		2267
1498	#include <math.h>	2268	#include <math.h>
1499		2269
1500	static ev_tstamp	2270	static ev_tstamp
1501	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)	2271	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
1502	{	2272	{
1503	return fmod (now, 3600.) + 3600.;	2273	return now + (3600. - fmod (now, 3600.));
1504	}	2274	}
1505		2275
1506	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);	2276	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1507		2277
1508	Example: Call a callback every hour, starting now:	2278	Example: Call a callback every hour, starting now:
1509		2279
1510	struct ev_periodic hourly_tick;	2280	ev_periodic hourly_tick;
1511	ev_periodic_init (&hourly_tick, clock_cb,	2281	ev_periodic_init (&hourly_tick, clock_cb,
1512	fmod (ev_now (loop), 3600.), 3600., 0);	2282	fmod (ev_now (loop), 3600.), 3600., 0);
1513	ev_periodic_start (loop, &hourly_tick);	2283	ev_periodic_start (loop, &hourly_tick);
1514		2284
1515		2285
1516	=head2 C<ev_signal> - signal me when a signal gets signalled!	2286	=head2 C<ev_signal> - signal me when a signal gets signalled!
1517		2287
1518	Signal watchers will trigger an event when the process receives a specific	2288	Signal watchers will trigger an event when the process receives a specific
1519	signal one or more times. Even though signals are very asynchronous, libev	2289	signal one or more times. Even though signals are very asynchronous, libev
1520	will try it's best to deliver signals synchronously, i.e. as part of the	2290	will try its best to deliver signals synchronously, i.e. as part of the
1521	normal event processing, like any other event.	2291	normal event processing, like any other event.
1522		2292
		2293	If you want signals to be delivered truly asynchronously, just use
		2294	C<sigaction> as you would do without libev and forget about sharing
		2295	the signal. You can even use C<ev_async> from a signal handler to
		2296	synchronously wake up an event loop.
		2297
1523	You can configure as many watchers as you like per signal. Only when the	2298	You can configure as many watchers as you like for the same signal, but
		2299	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2300	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2301	C<SIGINT> in both the default loop and another loop at the same time. At
		2302	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2303
1524	first watcher gets started will libev actually register a signal watcher	2304	When the first watcher gets started will libev actually register something
1525	with the kernel (thus it coexists with your own signal handlers as long	2305	with the kernel (thus it coexists with your own signal handlers as long as
1526	as you don't register any with libev). Similarly, when the last signal	2306	you don't register any with libev for the same signal).
1527	watcher for a signal is stopped libev will reset the signal handler to
1528	SIG_DFL (regardless of what it was set to before).
1529		2307
1530	If possible and supported, libev will install its handlers with	2308	If possible and supported, libev will install its handlers with
1531	C<SA_RESTART> behaviour enabled, so system calls should not be unduly	2309	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
1532	interrupted. If you have a problem with system calls getting interrupted by	2310	not be unduly interrupted. If you have a problem with system calls getting
1533	signals you can block all signals in an C<ev_check> watcher and unblock	2311	interrupted by signals you can block all signals in an C<ev_check> watcher
1534	them in an C<ev_prepare> watcher.	2312	and unblock them in an C<ev_prepare> watcher.
		2313
		2314	=head3 The special problem of inheritance over fork/execve/pthread_create
		2315
		2316	Both the signal mask (C<sigprocmask>) and the signal disposition
		2317	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2318	stopping it again), that is, libev might or might not block the signal,
		2319	and might or might not set or restore the installed signal handler (but
		2320	see C<EVFLAG_NOSIGMASK>).
		2321
		2322	While this does not matter for the signal disposition (libev never
		2323	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2324	C<execve>), this matters for the signal mask: many programs do not expect
		2325	certain signals to be blocked.
		2326
		2327	This means that before calling C<exec> (from the child) you should reset
		2328	the signal mask to whatever "default" you expect (all clear is a good
		2329	choice usually).
		2330
		2331	The simplest way to ensure that the signal mask is reset in the child is
		2332	to install a fork handler with C<pthread_atfork> that resets it. That will
		2333	catch fork calls done by libraries (such as the libc) as well.
		2334
		2335	In current versions of libev, the signal will not be blocked indefinitely
		2336	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2337	the window of opportunity for problems, it will not go away, as libev
		2338	I<has> to modify the signal mask, at least temporarily.
		2339
		2340	So I can't stress this enough: I<If you do not reset your signal mask when
		2341	you expect it to be empty, you have a race condition in your code>. This
		2342	is not a libev-specific thing, this is true for most event libraries.
		2343
		2344	=head3 The special problem of threads signal handling
		2345
		2346	POSIX threads has problematic signal handling semantics, specifically,
		2347	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2348	threads in a process block signals, which is hard to achieve.
		2349
		2350	When you want to use sigwait (or mix libev signal handling with your own
		2351	for the same signals), you can tackle this problem by globally blocking
		2352	all signals before creating any threads (or creating them with a fully set
		2353	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2354	loops. Then designate one thread as "signal receiver thread" which handles
		2355	these signals. You can pass on any signals that libev might be interested
		2356	in by calling C<ev_feed_signal>.
1535		2357
1536	=head3 Watcher-Specific Functions and Data Members	2358	=head3 Watcher-Specific Functions and Data Members
1537		2359
1538	=over 4	2360	=over 4
1539		2361
…		…
1550		2372
1551	=back	2373	=back
1552		2374
1553	=head3 Examples	2375	=head3 Examples
1554		2376
1555	Example: Try to exit cleanly on SIGINT and SIGTERM.	2377	Example: Try to exit cleanly on SIGINT.
1556		2378
1557	static void	2379	static void
1558	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)	2380	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
1559	{	2381	{
1560	ev_unloop (loop, EVUNLOOP_ALL);	2382	ev_break (loop, EVBREAK_ALL);
1561	}	2383	}
1562		2384
1563	struct ev_signal signal_watcher;	2385	ev_signal signal_watcher;
1564	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	2386	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1565	ev_signal_start (loop, &sigint_cb);	2387	ev_signal_start (loop, &signal_watcher);
1566		2388
1567		2389
1568	=head2 C<ev_child> - watch out for process status changes	2390	=head2 C<ev_child> - watch out for process status changes
1569		2391
1570	Child watchers trigger when your process receives a SIGCHLD in response to	2392	Child watchers trigger when your process receives a SIGCHLD in response to
1571	some child status changes (most typically when a child of yours dies). It	2393	some child status changes (most typically when a child of yours dies or
1572	is permissible to install a child watcher I<after> the child has been	2394	exits). It is permissible to install a child watcher I<after> the child
1573	forked (which implies it might have already exited), as long as the event	2395	has been forked (which implies it might have already exited), as long
1574	loop isn't entered (or is continued from a watcher).	2396	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2397	forking and then immediately registering a watcher for the child is fine,
		2398	but forking and registering a watcher a few event loop iterations later or
		2399	in the next callback invocation is not.
1575		2400
1576	Only the default event loop is capable of handling signals, and therefore	2401	Only the default event loop is capable of handling signals, and therefore
1577	you can only register child watchers in the default event loop.	2402	you can only register child watchers in the default event loop.
1578		2403
		2404	Due to some design glitches inside libev, child watchers will always be
		2405	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2406	libev)
		2407
1579	=head3 Process Interaction	2408	=head3 Process Interaction
1580		2409
1581	Libev grabs C<SIGCHLD> as soon as the default event loop is	2410	Libev grabs C<SIGCHLD> as soon as the default event loop is
1582	initialised. This is necessary to guarantee proper behaviour even if	2411	initialised. This is necessary to guarantee proper behaviour even if the
1583	the first child watcher is started after the child exits. The occurrence	2412	first child watcher is started after the child exits. The occurrence
1584	of C<SIGCHLD> is recorded asynchronously, but child reaping is done	2413	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
1585	synchronously as part of the event loop processing. Libev always reaps all	2414	synchronously as part of the event loop processing. Libev always reaps all
1586	children, even ones not watched.	2415	children, even ones not watched.
1587		2416
1588	=head3 Overriding the Built-In Processing	2417	=head3 Overriding the Built-In Processing
…		…
1598	=head3 Stopping the Child Watcher	2427	=head3 Stopping the Child Watcher
1599		2428
1600	Currently, the child watcher never gets stopped, even when the	2429	Currently, the child watcher never gets stopped, even when the
1601	child terminates, so normally one needs to stop the watcher in the	2430	child terminates, so normally one needs to stop the watcher in the
1602	callback. Future versions of libev might stop the watcher automatically	2431	callback. Future versions of libev might stop the watcher automatically
1603	when a child exit is detected.	2432	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2433	problem).
1604		2434
1605	=head3 Watcher-Specific Functions and Data Members	2435	=head3 Watcher-Specific Functions and Data Members
1606		2436
1607	=over 4	2437	=over 4
1608		2438
…		…
1640	its completion.	2470	its completion.
1641		2471
1642	ev_child cw;	2472	ev_child cw;
1643		2473
1644	static void	2474	static void
1645	child_cb (EV_P_ struct ev_child *w, int revents)	2475	child_cb (EV_P_ ev_child *w, int revents)
1646	{	2476	{
1647	ev_child_stop (EV_A_ w);	2477	ev_child_stop (EV_A_ w);
1648	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);	2478	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1649	}	2479	}
1650		2480
…		…
1665		2495
1666		2496
1667	=head2 C<ev_stat> - did the file attributes just change?	2497	=head2 C<ev_stat> - did the file attributes just change?
1668		2498
1669	This watches a file system path for attribute changes. That is, it calls	2499	This watches a file system path for attribute changes. That is, it calls
1670	C<stat> regularly (or when the OS says it changed) and sees if it changed	2500	C<stat> on that path in regular intervals (or when the OS says it changed)
1671	compared to the last time, invoking the callback if it did.	2501	and sees if it changed compared to the last time, invoking the callback if
		2502	it did.
1672		2503
1673	The path does not need to exist: changing from "path exists" to "path does	2504	The path does not need to exist: changing from "path exists" to "path does
1674	not exist" is a status change like any other. The condition "path does	2505	not exist" is a status change like any other. The condition "path does not
1675	not exist" is signified by the C<st_nlink> field being zero (which is	2506	exist" (or more correctly "path cannot be stat'ed") is signified by the
1676	otherwise always forced to be at least one) and all the other fields of	2507	C<st_nlink> field being zero (which is otherwise always forced to be at
1677	the stat buffer having unspecified contents.	2508	least one) and all the other fields of the stat buffer having unspecified
		2509	contents.
1678		2510
1679	The path I<should> be absolute and I<must not> end in a slash. If it is	2511	The path I<must not> end in a slash or contain special components such as
		2512	C<.> or C<..>. The path I<should> be absolute: If it is relative and
1680	relative and your working directory changes, the behaviour is undefined.	2513	your working directory changes, then the behaviour is undefined.
1681		2514
1682	Since there is no standard to do this, the portable implementation simply	2515	Since there is no portable change notification interface available, the
1683	calls C<stat (2)> regularly on the path to see if it changed somehow. You	2516	portable implementation simply calls C<stat(2)> regularly on the path
1684	can specify a recommended polling interval for this case. If you specify	2517	to see if it changed somehow. You can specify a recommended polling
1685	a polling interval of C<0> (highly recommended!) then a I<suitable,	2518	interval for this case. If you specify a polling interval of C<0> (highly
1686	unspecified default> value will be used (which you can expect to be around	2519	recommended!) then a I<suitable, unspecified default> value will be used
1687	five seconds, although this might change dynamically). Libev will also	2520	(which you can expect to be around five seconds, although this might
1688	impose a minimum interval which is currently around C<0.1>, but thats	2521	change dynamically). Libev will also impose a minimum interval which is
1689	usually overkill.	2522	currently around C<0.1>, but that's usually overkill.
1690		2523
1691	This watcher type is not meant for massive numbers of stat watchers,	2524	This watcher type is not meant for massive numbers of stat watchers,
1692	as even with OS-supported change notifications, this can be	2525	as even with OS-supported change notifications, this can be
1693	resource-intensive.	2526	resource-intensive.
1694		2527
1695	At the time of this writing, only the Linux inotify interface is	2528	At the time of this writing, the only OS-specific interface implemented
1696	implemented (implementing kqueue support is left as an exercise for the	2529	is the Linux inotify interface (implementing kqueue support is left as an
1697	reader, note, however, that the author sees no way of implementing ev_stat	2530	exercise for the reader. Note, however, that the author sees no way of
1698	semantics with kqueue). Inotify will be used to give hints only and should	2531	implementing C<ev_stat> semantics with kqueue, except as a hint).
1699	not change the semantics of C<ev_stat> watchers, which means that libev
1700	sometimes needs to fall back to regular polling again even with inotify,
1701	but changes are usually detected immediately, and if the file exists there
1702	will be no polling.
1703		2532
1704	=head3 ABI Issues (Largefile Support)	2533	=head3 ABI Issues (Largefile Support)
1705		2534
1706	Libev by default (unless the user overrides this) uses the default	2535	Libev by default (unless the user overrides this) uses the default
1707	compilation environment, which means that on systems with large file	2536	compilation environment, which means that on systems with large file
1708	support disabled by default, you get the 32 bit version of the stat	2537	support disabled by default, you get the 32 bit version of the stat
1709	structure. When using the library from programs that change the ABI to	2538	structure. When using the library from programs that change the ABI to
1710	use 64 bit file offsets the programs will fail. In that case you have to	2539	use 64 bit file offsets the programs will fail. In that case you have to
1711	compile libev with the same flags to get binary compatibility. This is	2540	compile libev with the same flags to get binary compatibility. This is
1712	obviously the case with any flags that change the ABI, but the problem is	2541	obviously the case with any flags that change the ABI, but the problem is
1713	most noticeably disabled with ev_stat and large file support.	2542	most noticeably displayed with ev_stat and large file support.
1714		2543
1715	The solution for this is to lobby your distribution maker to make large	2544	The solution for this is to lobby your distribution maker to make large
1716	file interfaces available by default (as e.g. FreeBSD does) and not	2545	file interfaces available by default (as e.g. FreeBSD does) and not
1717	optional. Libev cannot simply switch on large file support because it has	2546	optional. Libev cannot simply switch on large file support because it has
1718	to exchange stat structures with application programs compiled using the	2547	to exchange stat structures with application programs compiled using the
1719	default compilation environment.	2548	default compilation environment.
1720		2549
1721	=head3 Inotify	2550	=head3 Inotify and Kqueue
1722		2551
1723	When C<inotify (7)> support has been compiled into libev (generally only	2552	When C<inotify (7)> support has been compiled into libev and present at
1724	available on Linux) and present at runtime, it will be used to speed up	2553	runtime, it will be used to speed up change detection where possible. The
1725	change detection where possible. The inotify descriptor will be created lazily	2554	inotify descriptor will be created lazily when the first C<ev_stat>
1726	when the first C<ev_stat> watcher is being started.	2555	watcher is being started.
1727		2556
1728	Inotify presence does not change the semantics of C<ev_stat> watchers	2557	Inotify presence does not change the semantics of C<ev_stat> watchers
1729	except that changes might be detected earlier, and in some cases, to avoid	2558	except that changes might be detected earlier, and in some cases, to avoid
1730	making regular C<stat> calls. Even in the presence of inotify support	2559	making regular C<stat> calls. Even in the presence of inotify support
1731	there are many cases where libev has to resort to regular C<stat> polling.	2560	there are many cases where libev has to resort to regular C<stat> polling,
		2561	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2562	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2563	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2564	xfs are fully working) libev usually gets away without polling.
1732		2565
1733	(There is no support for kqueue, as apparently it cannot be used to	2566	There is no support for kqueue, as apparently it cannot be used to
1734	implement this functionality, due to the requirement of having a file	2567	implement this functionality, due to the requirement of having a file
1735	descriptor open on the object at all times).	2568	descriptor open on the object at all times, and detecting renames, unlinks
		2569	etc. is difficult.
		2570
		2571	=head3 C<stat ()> is a synchronous operation
		2572
		2573	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2574	the process. The exception are C<ev_stat> watchers - those call C<stat
		2575	()>, which is a synchronous operation.
		2576
		2577	For local paths, this usually doesn't matter: unless the system is very
		2578	busy or the intervals between stat's are large, a stat call will be fast,
		2579	as the path data is usually in memory already (except when starting the
		2580	watcher).
		2581
		2582	For networked file systems, calling C<stat ()> can block an indefinite
		2583	time due to network issues, and even under good conditions, a stat call
		2584	often takes multiple milliseconds.
		2585
		2586	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2587	paths, although this is fully supported by libev.
1736		2588
1737	=head3 The special problem of stat time resolution	2589	=head3 The special problem of stat time resolution
1738		2590
1739	The C<stat ()> system call only supports full-second resolution portably, and	2591	The C<stat ()> system call only supports full-second resolution portably,
1740	even on systems where the resolution is higher, many file systems still	2592	and even on systems where the resolution is higher, most file systems
1741	only support whole seconds.	2593	still only support whole seconds.
1742		2594
1743	That means that, if the time is the only thing that changes, you can	2595	That means that, if the time is the only thing that changes, you can
1744	easily miss updates: on the first update, C<ev_stat> detects a change and	2596	easily miss updates: on the first update, C<ev_stat> detects a change and
1745	calls your callback, which does something. When there is another update	2597	calls your callback, which does something. When there is another update
1746	within the same second, C<ev_stat> will be unable to detect it as the stat	2598	within the same second, C<ev_stat> will be unable to detect unless the
1747	data does not change.	2599	stat data does change in other ways (e.g. file size).
1748		2600
1749	The solution to this is to delay acting on a change for slightly more	2601	The solution to this is to delay acting on a change for slightly more
1750	than a second (or till slightly after the next full second boundary), using	2602	than a second (or till slightly after the next full second boundary), using
1751	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);	2603	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
1752	ev_timer_again (loop, w)>).	2604	ev_timer_again (loop, w)>).
…		…
1772	C<path>. The C<interval> is a hint on how quickly a change is expected to	2624	C<path>. The C<interval> is a hint on how quickly a change is expected to
1773	be detected and should normally be specified as C<0> to let libev choose	2625	be detected and should normally be specified as C<0> to let libev choose
1774	a suitable value. The memory pointed to by C<path> must point to the same	2626	a suitable value. The memory pointed to by C<path> must point to the same
1775	path for as long as the watcher is active.	2627	path for as long as the watcher is active.
1776		2628
1777	The callback will receive C<EV_STAT> when a change was detected, relative	2629	The callback will receive an C<EV_STAT> event when a change was detected,
1778	to the attributes at the time the watcher was started (or the last change	2630	relative to the attributes at the time the watcher was started (or the
1779	was detected).	2631	last change was detected).
1780		2632
1781	=item ev_stat_stat (loop, ev_stat *)	2633	=item ev_stat_stat (loop, ev_stat *)
1782		2634
1783	Updates the stat buffer immediately with new values. If you change the	2635	Updates the stat buffer immediately with new values. If you change the
1784	watched path in your callback, you could call this function to avoid	2636	watched path in your callback, you could call this function to avoid
…		…
1867		2719
1868		2720
1869	=head2 C<ev_idle> - when you've got nothing better to do...	2721	=head2 C<ev_idle> - when you've got nothing better to do...
1870		2722
1871	Idle watchers trigger events when no other events of the same or higher	2723	Idle watchers trigger events when no other events of the same or higher
1872	priority are pending (prepare, check and other idle watchers do not	2724	priority are pending (prepare, check and other idle watchers do not count
1873	count).	2725	as receiving "events").
1874		2726
1875	That is, as long as your process is busy handling sockets or timeouts	2727	That is, as long as your process is busy handling sockets or timeouts
1876	(or even signals, imagine) of the same or higher priority it will not be	2728	(or even signals, imagine) of the same or higher priority it will not be
1877	triggered. But when your process is idle (or only lower-priority watchers	2729	triggered. But when your process is idle (or only lower-priority watchers
1878	are pending), the idle watchers are being called once per event loop	2730	are pending), the idle watchers are being called once per event loop
…		…
1889		2741
1890	=head3 Watcher-Specific Functions and Data Members	2742	=head3 Watcher-Specific Functions and Data Members
1891		2743
1892	=over 4	2744	=over 4
1893		2745
1894	=item ev_idle_init (ev_signal *, callback)	2746	=item ev_idle_init (ev_idle *, callback)
1895		2747
1896	Initialises and configures the idle watcher - it has no parameters of any	2748	Initialises and configures the idle watcher - it has no parameters of any
1897	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2749	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1898	believe me.	2750	believe me.
1899		2751
…		…
1903		2755
1904	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the	2756	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1905	callback, free it. Also, use no error checking, as usual.	2757	callback, free it. Also, use no error checking, as usual.
1906		2758
1907	static void	2759	static void
1908	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)	2760	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
1909	{	2761	{
1910	free (w);	2762	free (w);
1911	// now do something you wanted to do when the program has	2763	// now do something you wanted to do when the program has
1912	// no longer anything immediate to do.	2764	// no longer anything immediate to do.
1913	}	2765	}
1914		2766
1915	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));	2767	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
1916	ev_idle_init (idle_watcher, idle_cb);	2768	ev_idle_init (idle_watcher, idle_cb);
1917	ev_idle_start (loop, idle_cb);	2769	ev_idle_start (loop, idle_watcher);
1918		2770
1919		2771
1920	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	2772	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1921		2773
1922	Prepare and check watchers are usually (but not always) used in tandem:	2774	Prepare and check watchers are usually (but not always) used in pairs:
1923	prepare watchers get invoked before the process blocks and check watchers	2775	prepare watchers get invoked before the process blocks and check watchers
1924	afterwards.	2776	afterwards.
1925		2777
1926	You I<must not> call C<ev_loop> or similar functions that enter	2778	You I<must not> call C<ev_run> or similar functions that enter
1927	the current event loop from either C<ev_prepare> or C<ev_check>	2779	the current event loop from either C<ev_prepare> or C<ev_check>
1928	watchers. Other loops than the current one are fine, however. The	2780	watchers. Other loops than the current one are fine, however. The
1929	rationale behind this is that you do not need to check for recursion in	2781	rationale behind this is that you do not need to check for recursion in
1930	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,	2782	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1931	C<ev_check> so if you have one watcher of each kind they will always be	2783	C<ev_check> so if you have one watcher of each kind they will always be
1932	called in pairs bracketing the blocking call.	2784	called in pairs bracketing the blocking call.
1933		2785
1934	Their main purpose is to integrate other event mechanisms into libev and	2786	Their main purpose is to integrate other event mechanisms into libev and
1935	their use is somewhat advanced. This could be used, for example, to track	2787	their use is somewhat advanced. They could be used, for example, to track
1936	variable changes, implement your own watchers, integrate net-snmp or a	2788	variable changes, implement your own watchers, integrate net-snmp or a
1937	coroutine library and lots more. They are also occasionally useful if	2789	coroutine library and lots more. They are also occasionally useful if
1938	you cache some data and want to flush it before blocking (for example,	2790	you cache some data and want to flush it before blocking (for example,
1939	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>	2791	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1940	watcher).	2792	watcher).
1941		2793
1942	This is done by examining in each prepare call which file descriptors need	2794	This is done by examining in each prepare call which file descriptors
1943	to be watched by the other library, registering C<ev_io> watchers for	2795	need to be watched by the other library, registering C<ev_io> watchers
1944	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2796	for them and starting an C<ev_timer> watcher for any timeouts (many
1945	provide just this functionality). Then, in the check watcher you check for	2797	libraries provide exactly this functionality). Then, in the check watcher,
1946	any events that occurred (by checking the pending status of all watchers	2798	you check for any events that occurred (by checking the pending status
1947	and stopping them) and call back into the library. The I/O and timer	2799	of all watchers and stopping them) and call back into the library. The
1948	callbacks will never actually be called (but must be valid nevertheless,	2800	I/O and timer callbacks will never actually be called (but must be valid
1949	because you never know, you know?).	2801	nevertheless, because you never know, you know?).
1950		2802
1951	As another example, the Perl Coro module uses these hooks to integrate	2803	As another example, the Perl Coro module uses these hooks to integrate
1952	coroutines into libev programs, by yielding to other active coroutines	2804	coroutines into libev programs, by yielding to other active coroutines
1953	during each prepare and only letting the process block if no coroutines	2805	during each prepare and only letting the process block if no coroutines
1954	are ready to run (it's actually more complicated: it only runs coroutines	2806	are ready to run (it's actually more complicated: it only runs coroutines
…		…
1957	loop from blocking if lower-priority coroutines are active, thus mapping	2809	loop from blocking if lower-priority coroutines are active, thus mapping
1958	low-priority coroutines to idle/background tasks).	2810	low-priority coroutines to idle/background tasks).
1959		2811
1960	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	2812	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1961	priority, to ensure that they are being run before any other watchers	2813	priority, to ensure that they are being run before any other watchers
		2814	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2815
1962	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,	2816	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
1963	too) should not activate ("feed") events into libev. While libev fully	2817	activate ("feed") events into libev. While libev fully supports this, they
1964	supports this, they might get executed before other C<ev_check> watchers	2818	might get executed before other C<ev_check> watchers did their job. As
1965	did their job. As C<ev_check> watchers are often used to embed other	2819	C<ev_check> watchers are often used to embed other (non-libev) event
1966	(non-libev) event loops those other event loops might be in an unusable	2820	loops those other event loops might be in an unusable state until their
1967	state until their C<ev_check> watcher ran (always remind yourself to	2821	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
1968	coexist peacefully with others).	2822	others).
1969		2823
1970	=head3 Watcher-Specific Functions and Data Members	2824	=head3 Watcher-Specific Functions and Data Members
1971		2825
1972	=over 4	2826	=over 4
1973		2827
…		…
1975		2829
1976	=item ev_check_init (ev_check *, callback)	2830	=item ev_check_init (ev_check *, callback)
1977		2831
1978	Initialises and configures the prepare or check watcher - they have no	2832	Initialises and configures the prepare or check watcher - they have no
1979	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2833	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1980	macros, but using them is utterly, utterly and completely pointless.	2834	macros, but using them is utterly, utterly, utterly and completely
		2835	pointless.
1981		2836
1982	=back	2837	=back
1983		2838
1984	=head3 Examples	2839	=head3 Examples
1985		2840
…		…
1998		2853
1999	static ev_io iow [nfd];	2854	static ev_io iow [nfd];
2000	static ev_timer tw;	2855	static ev_timer tw;
2001		2856
2002	static void	2857	static void
2003	io_cb (ev_loop *loop, ev_io *w, int revents)	2858	io_cb (struct ev_loop *loop, ev_io *w, int revents)
2004	{	2859	{
2005	}	2860	}
2006		2861
2007	// create io watchers for each fd and a timer before blocking	2862	// create io watchers for each fd and a timer before blocking
2008	static void	2863	static void
2009	adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)	2864	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
2010	{	2865	{
2011	int timeout = 3600000;	2866	int timeout = 3600000;
2012	struct pollfd fds [nfd];	2867	struct pollfd fds [nfd];
2013	// actual code will need to loop here and realloc etc.	2868	// actual code will need to loop here and realloc etc.
2014	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));	2869	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2015		2870
2016	/* the callback is illegal, but won't be called as we stop during check */	2871	/* the callback is illegal, but won't be called as we stop during check */
2017	ev_timer_init (&tw, 0, timeout * 1e-3);	2872	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
2018	ev_timer_start (loop, &tw);	2873	ev_timer_start (loop, &tw);
2019		2874
2020	// create one ev_io per pollfd	2875	// create one ev_io per pollfd
2021	for (int i = 0; i < nfd; ++i)	2876	for (int i = 0; i < nfd; ++i)
2022	{	2877	{
…		…
2029	}	2884	}
2030	}	2885	}
2031		2886
2032	// stop all watchers after blocking	2887	// stop all watchers after blocking
2033	static void	2888	static void
2034	adns_check_cb (ev_loop *loop, ev_check *w, int revents)	2889	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
2035	{	2890	{
2036	ev_timer_stop (loop, &tw);	2891	ev_timer_stop (loop, &tw);
2037		2892
2038	for (int i = 0; i < nfd; ++i)	2893	for (int i = 0; i < nfd; ++i)
2039	{	2894	{
…		…
2078	}	2933	}
2079		2934
2080	// do not ever call adns_afterpoll	2935	// do not ever call adns_afterpoll
2081		2936
2082	Method 4: Do not use a prepare or check watcher because the module you	2937	Method 4: Do not use a prepare or check watcher because the module you
2083	want to embed is too inflexible to support it. Instead, you can override	2938	want to embed is not flexible enough to support it. Instead, you can
2084	their poll function. The drawback with this solution is that the main	2939	override their poll function. The drawback with this solution is that the
2085	loop is now no longer controllable by EV. The C<Glib::EV> module does	2940	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
2086	this.	2941	this approach, effectively embedding EV as a client into the horrible
		2942	libglib event loop.
2087		2943
2088	static gint	2944	static gint
2089	event_poll_func (GPollFD *fds, guint nfds, gint timeout)	2945	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
2090	{	2946	{
2091	int got_events = 0;	2947	int got_events = 0;
…		…
2095		2951
2096	if (timeout >= 0)	2952	if (timeout >= 0)
2097	// create/start timer	2953	// create/start timer
2098		2954
2099	// poll	2955	// poll
2100	ev_loop (EV_A_ 0);	2956	ev_run (EV_A_ 0);
2101		2957
2102	// stop timer again	2958	// stop timer again
2103	if (timeout >= 0)	2959	if (timeout >= 0)
2104	ev_timer_stop (EV_A_ &to);	2960	ev_timer_stop (EV_A_ &to);
2105		2961
…		…
2122	prioritise I/O.	2978	prioritise I/O.
2123		2979
2124	As an example for a bug workaround, the kqueue backend might only support	2980	As an example for a bug workaround, the kqueue backend might only support
2125	sockets on some platform, so it is unusable as generic backend, but you	2981	sockets on some platform, so it is unusable as generic backend, but you
2126	still want to make use of it because you have many sockets and it scales	2982	still want to make use of it because you have many sockets and it scales
2127	so nicely. In this case, you would create a kqueue-based loop and embed it	2983	so nicely. In this case, you would create a kqueue-based loop and embed
2128	into your default loop (which might use e.g. poll). Overall operation will	2984	it into your default loop (which might use e.g. poll). Overall operation
2129	be a bit slower because first libev has to poll and then call kevent, but	2985	will be a bit slower because first libev has to call C<poll> and then
2130	at least you can use both at what they are best.	2986	C<kevent>, but at least you can use both mechanisms for what they are
		2987	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
2131		2988
2132	As for prioritising I/O: rarely you have the case where some fds have	2989	As for prioritising I/O: under rare circumstances you have the case where
2133	to be watched and handled very quickly (with low latency), and even	2990	some fds have to be watched and handled very quickly (with low latency),
2134	priorities and idle watchers might have too much overhead. In this case	2991	and even priorities and idle watchers might have too much overhead. In
2135	you would put all the high priority stuff in one loop and all the rest in	2992	this case you would put all the high priority stuff in one loop and all
2136	a second one, and embed the second one in the first.	2993	the rest in a second one, and embed the second one in the first.
2137		2994
2138	As long as the watcher is active, the callback will be invoked every time	2995	As long as the watcher is active, the callback will be invoked every
2139	there might be events pending in the embedded loop. The callback must then	2996	time there might be events pending in the embedded loop. The callback
2140	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	2997	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
2141	their callbacks (you could also start an idle watcher to give the embedded	2998	sweep and invoke their callbacks (the callback doesn't need to invoke the
2142	loop strictly lower priority for example). You can also set the callback	2999	C<ev_embed_sweep> function directly, it could also start an idle watcher
2143	to C<0>, in which case the embed watcher will automatically execute the	3000	to give the embedded loop strictly lower priority for example).
2144	embedded loop sweep.
2145		3001
2146	As long as the watcher is started it will automatically handle events. The	3002	You can also set the callback to C<0>, in which case the embed watcher
2147	callback will be invoked whenever some events have been handled. You can	3003	will automatically execute the embedded loop sweep whenever necessary.
2148	set the callback to C<0> to avoid having to specify one if you are not
2149	interested in that.
2150		3004
2151	Also, there have not currently been made special provisions for forking:	3005	Fork detection will be handled transparently while the C<ev_embed> watcher
2152	when you fork, you not only have to call C<ev_loop_fork> on both loops,	3006	is active, i.e., the embedded loop will automatically be forked when the
2153	but you will also have to stop and restart any C<ev_embed> watchers	3007	embedding loop forks. In other cases, the user is responsible for calling
2154	yourself.	3008	C<ev_loop_fork> on the embedded loop.
2155		3009
2156	Unfortunately, not all backends are embeddable, only the ones returned by	3010	Unfortunately, not all backends are embeddable: only the ones returned by
2157	C<ev_embeddable_backends> are, which, unfortunately, does not include any	3011	C<ev_embeddable_backends> are, which, unfortunately, does not include any
2158	portable one.	3012	portable one.
2159		3013
2160	So when you want to use this feature you will always have to be prepared	3014	So when you want to use this feature you will always have to be prepared
2161	that you cannot get an embeddable loop. The recommended way to get around	3015	that you cannot get an embeddable loop. The recommended way to get around
2162	this is to have a separate variables for your embeddable loop, try to	3016	this is to have a separate variables for your embeddable loop, try to
2163	create it, and if that fails, use the normal loop for everything.	3017	create it, and if that fails, use the normal loop for everything.
		3018
		3019	=head3 C<ev_embed> and fork
		3020
		3021	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3022	automatically be applied to the embedded loop as well, so no special
		3023	fork handling is required in that case. When the watcher is not running,
		3024	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3025	as applicable.
2164		3026
2165	=head3 Watcher-Specific Functions and Data Members	3027	=head3 Watcher-Specific Functions and Data Members
2166		3028
2167	=over 4	3029	=over 4
2168		3030
…		…
2177	if you do not want that, you need to temporarily stop the embed watcher).	3039	if you do not want that, you need to temporarily stop the embed watcher).
2178		3040
2179	=item ev_embed_sweep (loop, ev_embed *)	3041	=item ev_embed_sweep (loop, ev_embed *)
2180		3042
2181	Make a single, non-blocking sweep over the embedded loop. This works	3043	Make a single, non-blocking sweep over the embedded loop. This works
2182	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most	3044	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
2183	appropriate way for embedded loops.	3045	appropriate way for embedded loops.
2184		3046
2185	=item struct ev_loop *other [read-only]	3047	=item struct ev_loop *other [read-only]
2186		3048
2187	The embedded event loop.	3049	The embedded event loop.
…		…
2196	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be	3058	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
2197	used).	3059	used).
2198		3060
2199	struct ev_loop *loop_hi = ev_default_init (0);	3061	struct ev_loop *loop_hi = ev_default_init (0);
2200	struct ev_loop *loop_lo = 0;	3062	struct ev_loop *loop_lo = 0;
2201	struct ev_embed embed;	3063	ev_embed embed;
2202		3064
2203	// see if there is a chance of getting one that works	3065	// see if there is a chance of getting one that works
2204	// (remember that a flags value of 0 means autodetection)	3066	// (remember that a flags value of 0 means autodetection)
2205	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()	3067	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
2206	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())	3068	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
…		…
2220	kqueue implementation). Store the kqueue/socket-only event loop in	3082	kqueue implementation). Store the kqueue/socket-only event loop in
2221	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).	3083	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
2222		3084
2223	struct ev_loop *loop = ev_default_init (0);	3085	struct ev_loop *loop = ev_default_init (0);
2224	struct ev_loop *loop_socket = 0;	3086	struct ev_loop *loop_socket = 0;
2225	struct ev_embed embed;	3087	ev_embed embed;
2226		3088
2227	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)	3089	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
2228	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))	3090	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
2229	{	3091	{
2230	ev_embed_init (&embed, 0, loop_socket);	3092	ev_embed_init (&embed, 0, loop_socket);
…		…
2245	event loop blocks next and before C<ev_check> watchers are being called,	3107	event loop blocks next and before C<ev_check> watchers are being called,
2246	and only in the child after the fork. If whoever good citizen calling	3108	and only in the child after the fork. If whoever good citizen calling
2247	C<ev_default_fork> cheats and calls it in the wrong process, the fork	3109	C<ev_default_fork> cheats and calls it in the wrong process, the fork
2248	handlers will be invoked, too, of course.	3110	handlers will be invoked, too, of course.
2249		3111
		3112	=head3 The special problem of life after fork - how is it possible?
		3113
		3114	Most uses of C<fork()> consist of forking, then some simple calls to set
		3115	up/change the process environment, followed by a call to C<exec()>. This
		3116	sequence should be handled by libev without any problems.
		3117
		3118	This changes when the application actually wants to do event handling
		3119	in the child, or both parent in child, in effect "continuing" after the
		3120	fork.
		3121
		3122	The default mode of operation (for libev, with application help to detect
		3123	forks) is to duplicate all the state in the child, as would be expected
		3124	when I<either> the parent I<or> the child process continues.
		3125
		3126	When both processes want to continue using libev, then this is usually the
		3127	wrong result. In that case, usually one process (typically the parent) is
		3128	supposed to continue with all watchers in place as before, while the other
		3129	process typically wants to start fresh, i.e. without any active watchers.
		3130
		3131	The cleanest and most efficient way to achieve that with libev is to
		3132	simply create a new event loop, which of course will be "empty", and
		3133	use that for new watchers. This has the advantage of not touching more
		3134	memory than necessary, and thus avoiding the copy-on-write, and the
		3135	disadvantage of having to use multiple event loops (which do not support
		3136	signal watchers).
		3137
		3138	When this is not possible, or you want to use the default loop for
		3139	other reasons, then in the process that wants to start "fresh", call
		3140	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3141	Destroying the default loop will "orphan" (not stop) all registered
		3142	watchers, so you have to be careful not to execute code that modifies
		3143	those watchers. Note also that in that case, you have to re-register any
		3144	signal watchers.
		3145
2250	=head3 Watcher-Specific Functions and Data Members	3146	=head3 Watcher-Specific Functions and Data Members
2251		3147
2252	=over 4	3148	=over 4
2253		3149
2254	=item ev_fork_init (ev_signal *, callback)	3150	=item ev_fork_init (ev_fork *, callback)
2255		3151
2256	Initialises and configures the fork watcher - it has no parameters of any	3152	Initialises and configures the fork watcher - it has no parameters of any
2257	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,	3153	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
2258	believe me.	3154	really.
2259		3155
2260	=back	3156	=back
2261		3157
2262		3158
		3159	=head2 C<ev_cleanup> - even the best things end
		3160
		3161	Cleanup watchers are called just before the event loop is being destroyed
		3162	by a call to C<ev_loop_destroy>.
		3163
		3164	While there is no guarantee that the event loop gets destroyed, cleanup
		3165	watchers provide a convenient method to install cleanup hooks for your
		3166	program, worker threads and so on - you just to make sure to destroy the
		3167	loop when you want them to be invoked.
		3168
		3169	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3170	all other watchers, they do not keep a reference to the event loop (which
		3171	makes a lot of sense if you think about it). Like all other watchers, you
		3172	can call libev functions in the callback, except C<ev_cleanup_start>.
		3173
		3174	=head3 Watcher-Specific Functions and Data Members
		3175
		3176	=over 4
		3177
		3178	=item ev_cleanup_init (ev_cleanup *, callback)
		3179
		3180	Initialises and configures the cleanup watcher - it has no parameters of
		3181	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3182	pointless, I assure you.
		3183
		3184	=back
		3185
		3186	Example: Register an atexit handler to destroy the default loop, so any
		3187	cleanup functions are called.
		3188
		3189	static void
		3190	program_exits (void)
		3191	{
		3192	ev_loop_destroy (EV_DEFAULT_UC);
		3193	}
		3194
		3195	...
		3196	atexit (program_exits);
		3197
		3198
2263	=head2 C<ev_async> - how to wake up another event loop	3199	=head2 C<ev_async> - how to wake up an event loop
2264		3200
2265	In general, you cannot use an C<ev_loop> from multiple threads or other	3201	In general, you cannot use an C<ev_loop> from multiple threads or other
2266	asynchronous sources such as signal handlers (as opposed to multiple event	3202	asynchronous sources such as signal handlers (as opposed to multiple event
2267	loops - those are of course safe to use in different threads).	3203	loops - those are of course safe to use in different threads).
2268		3204
2269	Sometimes, however, you need to wake up another event loop you do not	3205	Sometimes, however, you need to wake up an event loop you do not control,
2270	control, for example because it belongs to another thread. This is what	3206	for example because it belongs to another thread. This is what C<ev_async>
2271	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you	3207	watchers do: as long as the C<ev_async> watcher is active, you can signal
2272	can signal it by calling C<ev_async_send>, which is thread- and signal	3208	it by calling C<ev_async_send>, which is thread- and signal safe.
2273	safe.
2274		3209
2275	This functionality is very similar to C<ev_signal> watchers, as signals,	3210	This functionality is very similar to C<ev_signal> watchers, as signals,
2276	too, are asynchronous in nature, and signals, too, will be compressed	3211	too, are asynchronous in nature, and signals, too, will be compressed
2277	(i.e. the number of callback invocations may be less than the number of	3212	(i.e. the number of callback invocations may be less than the number of
2278	C<ev_async_sent> calls).	3213	C<ev_async_sent> calls). In fact, you could use signal watchers as a kind
		3214	of "global async watchers" by using a watcher on an otherwise unused
		3215	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3216	even without knowing which loop owns the signal.
2279		3217
2280	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not	3218	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2281	just the default loop.	3219	just the default loop.
2282		3220
2283	=head3 Queueing	3221	=head3 Queueing
2284		3222
2285	C<ev_async> does not support queueing of data in any way. The reason	3223	C<ev_async> does not support queueing of data in any way. The reason
2286	is that the author does not know of a simple (or any) algorithm for a	3224	is that the author does not know of a simple (or any) algorithm for a
2287	multiple-writer-single-reader queue that works in all cases and doesn't	3225	multiple-writer-single-reader queue that works in all cases and doesn't
2288	need elaborate support such as pthreads.	3226	need elaborate support such as pthreads or unportable memory access
		3227	semantics.
2289		3228
2290	That means that if you want to queue data, you have to provide your own	3229	That means that if you want to queue data, you have to provide your own
2291	queue. But at least I can tell you would implement locking around your	3230	queue. But at least I can tell you how to implement locking around your
2292	queue:	3231	queue:
2293		3232
2294	=over 4	3233	=over 4
2295		3234
2296	=item queueing from a signal handler context	3235	=item queueing from a signal handler context
2297		3236
2298	To implement race-free queueing, you simply add to the queue in the signal	3237	To implement race-free queueing, you simply add to the queue in the signal
2299	handler but you block the signal handler in the watcher callback. Here is an example that does that for	3238	handler but you block the signal handler in the watcher callback. Here is
2300	some fictitious SIGUSR1 handler:	3239	an example that does that for some fictitious SIGUSR1 handler:
2301		3240
2302	static ev_async mysig;	3241	static ev_async mysig;
2303		3242
2304	static void	3243	static void
2305	sigusr1_handler (void)	3244	sigusr1_handler (void)
…		…
2371	=over 4	3310	=over 4
2372		3311
2373	=item ev_async_init (ev_async *, callback)	3312	=item ev_async_init (ev_async *, callback)
2374		3313
2375	Initialises and configures the async watcher - it has no parameters of any	3314	Initialises and configures the async watcher - it has no parameters of any
2376	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,	3315	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
2377	believe me.	3316	trust me.
2378		3317
2379	=item ev_async_send (loop, ev_async *)	3318	=item ev_async_send (loop, ev_async *)
2380		3319
2381	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	3320	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2382	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	3321	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3322	returns.
		3323
2383	C<ev_feed_event>, this call is safe to do in other threads, signal or	3324	Unlike C<ev_feed_event>, this call is safe to do from other threads,
2384	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	3325	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
2385	section below on what exactly this means).	3326	embedding section below on what exactly this means).
2386		3327
		3328	Note that, as with other watchers in libev, multiple events might get
		3329	compressed into a single callback invocation (another way to look at this
		3330	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3331	reset when the event loop detects that).
		3332
2387	This call incurs the overhead of a system call only once per loop iteration,	3333	This call incurs the overhead of a system call only once per event loop
2388	so while the overhead might be noticeable, it doesn't apply to repeated	3334	iteration, so while the overhead might be noticeable, it doesn't apply to
2389	calls to C<ev_async_send>.	3335	repeated calls to C<ev_async_send> for the same event loop.
2390		3336
2391	=item bool = ev_async_pending (ev_async *)	3337	=item bool = ev_async_pending (ev_async *)
2392		3338
2393	Returns a non-zero value when C<ev_async_send> has been called on the	3339	Returns a non-zero value when C<ev_async_send> has been called on the
2394	watcher but the event has not yet been processed (or even noted) by the	3340	watcher but the event has not yet been processed (or even noted) by the
…		…
2397	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When	3343	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2398	the loop iterates next and checks for the watcher to have become active,	3344	the loop iterates next and checks for the watcher to have become active,
2399	it will reset the flag again. C<ev_async_pending> can be used to very	3345	it will reset the flag again. C<ev_async_pending> can be used to very
2400	quickly check whether invoking the loop might be a good idea.	3346	quickly check whether invoking the loop might be a good idea.
2401		3347
2402	Not that this does I<not> check whether the watcher itself is pending, only	3348	Not that this does I<not> check whether the watcher itself is pending,
2403	whether it has been requested to make this watcher pending.	3349	only whether it has been requested to make this watcher pending: there
		3350	is a time window between the event loop checking and resetting the async
		3351	notification, and the callback being invoked.
2404		3352
2405	=back	3353	=back
2406		3354
2407		3355
2408	=head1 OTHER FUNCTIONS	3356	=head1 OTHER FUNCTIONS
…		…
2412	=over 4	3360	=over 4
2413		3361
2414	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3362	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2415		3363
2416	This function combines a simple timer and an I/O watcher, calls your	3364	This function combines a simple timer and an I/O watcher, calls your
2417	callback on whichever event happens first and automatically stop both	3365	callback on whichever event happens first and automatically stops both
2418	watchers. This is useful if you want to wait for a single event on an fd	3366	watchers. This is useful if you want to wait for a single event on an fd
2419	or timeout without having to allocate/configure/start/stop/free one or	3367	or timeout without having to allocate/configure/start/stop/free one or
2420	more watchers yourself.	3368	more watchers yourself.
2421		3369
2422	If C<fd> is less than 0, then no I/O watcher will be started and events	3370	If C<fd> is less than 0, then no I/O watcher will be started and the
2423	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3371	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
2424	C<events> set will be created and started.	3372	the given C<fd> and C<events> set will be created and started.
2425		3373
2426	If C<timeout> is less than 0, then no timeout watcher will be	3374	If C<timeout> is less than 0, then no timeout watcher will be
2427	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3375	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2428	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3376	repeat = 0) will be started. C<0> is a valid timeout.
2429	dubious value.
2430		3377
2431	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3378	The callback has the type C<void (*cb)(int revents, void *arg)> and is
2432	passed an C<revents> set like normal event callbacks (a combination of	3379	passed an C<revents> set like normal event callbacks (a combination of
2433	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3380	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
2434	value passed to C<ev_once>:	3381	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3382	a timeout and an io event at the same time - you probably should give io
		3383	events precedence.
		3384
		3385	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
2435		3386
2436	static void stdin_ready (int revents, void *arg)	3387	static void stdin_ready (int revents, void *arg)
2437	{	3388	{
		3389	if (revents & EV_READ)
		3390	/* stdin might have data for us, joy! */;
2438	if (revents & EV_TIMEOUT)	3391	else if (revents & EV_TIMER)
2439	/* doh, nothing entered */;	3392	/* doh, nothing entered */;
2440	else if (revents & EV_READ)
2441	/* stdin might have data for us, joy! */;
2442	}	3393	}
2443		3394
2444	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3395	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2445		3396
2446	=item ev_feed_event (ev_loop *, watcher *, int revents)
2447
2448	Feeds the given event set into the event loop, as if the specified event
2449	had happened for the specified watcher (which must be a pointer to an
2450	initialised but not necessarily started event watcher).
2451
2452	=item ev_feed_fd_event (ev_loop *, int fd, int revents)	3397	=item ev_feed_fd_event (loop, int fd, int revents)
2453		3398
2454	Feed an event on the given fd, as if a file descriptor backend detected	3399	Feed an event on the given fd, as if a file descriptor backend detected
2455	the given events it.	3400	the given events it.
2456		3401
2457	=item ev_feed_signal_event (ev_loop *loop, int signum)	3402	=item ev_feed_signal_event (loop, int signum)
2458		3403
2459	Feed an event as if the given signal occurred (C<loop> must be the default	3404	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
2460	loop!).	3405	which is async-safe.
2461		3406
2462	=back	3407	=back
		3408
		3409
		3410	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3411
		3412	This section explains some common idioms that are not immediately
		3413	obvious. Note that examples are sprinkled over the whole manual, and this
		3414	section only contains stuff that wouldn't fit anywhere else.
		3415
		3416	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3417
		3418	Each watcher has, by default, a C<void *data> member that you can read
		3419	or modify at any time: libev will completely ignore it. This can be used
		3420	to associate arbitrary data with your watcher. If you need more data and
		3421	don't want to allocate memory separately and store a pointer to it in that
		3422	data member, you can also "subclass" the watcher type and provide your own
		3423	data:
		3424
		3425	struct my_io
		3426	{
		3427	ev_io io;
		3428	int otherfd;
		3429	void *somedata;
		3430	struct whatever *mostinteresting;
		3431	};
		3432
		3433	...
		3434	struct my_io w;
		3435	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3436
		3437	And since your callback will be called with a pointer to the watcher, you
		3438	can cast it back to your own type:
		3439
		3440	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3441	{
		3442	struct my_io *w = (struct my_io *)w_;
		3443	...
		3444	}
		3445
		3446	More interesting and less C-conformant ways of casting your callback
		3447	function type instead have been omitted.
		3448
		3449	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3450
		3451	Another common scenario is to use some data structure with multiple
		3452	embedded watchers, in effect creating your own watcher that combines
		3453	multiple libev event sources into one "super-watcher":
		3454
		3455	struct my_biggy
		3456	{
		3457	int some_data;
		3458	ev_timer t1;
		3459	ev_timer t2;
		3460	}
		3461
		3462	In this case getting the pointer to C<my_biggy> is a bit more
		3463	complicated: Either you store the address of your C<my_biggy> struct in
		3464	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3465	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3466	real programmers):
		3467
		3468	#include <stddef.h>
		3469
		3470	static void
		3471	t1_cb (EV_P_ ev_timer *w, int revents)
		3472	{
		3473	struct my_biggy big = (struct my_biggy *)
		3474	(((char *)w) - offsetof (struct my_biggy, t1));
		3475	}
		3476
		3477	static void
		3478	t2_cb (EV_P_ ev_timer *w, int revents)
		3479	{
		3480	struct my_biggy big = (struct my_biggy *)
		3481	(((char *)w) - offsetof (struct my_biggy, t2));
		3482	}
		3483
		3484	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3485
		3486	Often (especially in GUI toolkits) there are places where you have
		3487	I<modal> interaction, which is most easily implemented by recursively
		3488	invoking C<ev_run>.
		3489
		3490	This brings the problem of exiting - a callback might want to finish the
		3491	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3492	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3493	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3494	other combination: In these cases, C<ev_break> will not work alone.
		3495
		3496	The solution is to maintain "break this loop" variable for each C<ev_run>
		3497	invocation, and use a loop around C<ev_run> until the condition is
		3498	triggered, using C<EVRUN_ONCE>:
		3499
		3500	// main loop
		3501	int exit_main_loop = 0;
		3502
		3503	while (!exit_main_loop)
		3504	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3505
		3506	// in a model watcher
		3507	int exit_nested_loop = 0;
		3508
		3509	while (!exit_nested_loop)
		3510	ev_run (EV_A_ EVRUN_ONCE);
		3511
		3512	To exit from any of these loops, just set the corresponding exit variable:
		3513
		3514	// exit modal loop
		3515	exit_nested_loop = 1;
		3516
		3517	// exit main program, after modal loop is finished
		3518	exit_main_loop = 1;
		3519
		3520	// exit both
		3521	exit_main_loop = exit_nested_loop = 1;
		3522
		3523	=head2 THREAD LOCKING EXAMPLE
		3524
		3525	Here is a fictitious example of how to run an event loop in a different
		3526	thread from where callbacks are being invoked and watchers are
		3527	created/added/removed.
		3528
		3529	For a real-world example, see the C<EV::Loop::Async> perl module,
		3530	which uses exactly this technique (which is suited for many high-level
		3531	languages).
		3532
		3533	The example uses a pthread mutex to protect the loop data, a condition
		3534	variable to wait for callback invocations, an async watcher to notify the
		3535	event loop thread and an unspecified mechanism to wake up the main thread.
		3536
		3537	First, you need to associate some data with the event loop:
		3538
		3539	typedef struct {
		3540	mutex_t lock; /* global loop lock */
		3541	ev_async async_w;
		3542	thread_t tid;
		3543	cond_t invoke_cv;
		3544	} userdata;
		3545
		3546	void prepare_loop (EV_P)
		3547	{
		3548	// for simplicity, we use a static userdata struct.
		3549	static userdata u;
		3550
		3551	ev_async_init (&u->async_w, async_cb);
		3552	ev_async_start (EV_A_ &u->async_w);
		3553
		3554	pthread_mutex_init (&u->lock, 0);
		3555	pthread_cond_init (&u->invoke_cv, 0);
		3556
		3557	// now associate this with the loop
		3558	ev_set_userdata (EV_A_ u);
		3559	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3560	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3561
		3562	// then create the thread running ev_run
		3563	pthread_create (&u->tid, 0, l_run, EV_A);
		3564	}
		3565
		3566	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3567	solely to wake up the event loop so it takes notice of any new watchers
		3568	that might have been added:
		3569
		3570	static void
		3571	async_cb (EV_P_ ev_async *w, int revents)
		3572	{
		3573	// just used for the side effects
		3574	}
		3575
		3576	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3577	protecting the loop data, respectively.
		3578
		3579	static void
		3580	l_release (EV_P)
		3581	{
		3582	userdata *u = ev_userdata (EV_A);
		3583	pthread_mutex_unlock (&u->lock);
		3584	}
		3585
		3586	static void
		3587	l_acquire (EV_P)
		3588	{
		3589	userdata *u = ev_userdata (EV_A);
		3590	pthread_mutex_lock (&u->lock);
		3591	}
		3592
		3593	The event loop thread first acquires the mutex, and then jumps straight
		3594	into C<ev_run>:
		3595
		3596	void *
		3597	l_run (void *thr_arg)
		3598	{
		3599	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3600
		3601	l_acquire (EV_A);
		3602	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3603	ev_run (EV_A_ 0);
		3604	l_release (EV_A);
		3605
		3606	return 0;
		3607	}
		3608
		3609	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3610	signal the main thread via some unspecified mechanism (signals? pipe
		3611	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3612	have been called (in a while loop because a) spurious wakeups are possible
		3613	and b) skipping inter-thread-communication when there are no pending
		3614	watchers is very beneficial):
		3615
		3616	static void
		3617	l_invoke (EV_P)
		3618	{
		3619	userdata *u = ev_userdata (EV_A);
		3620
		3621	while (ev_pending_count (EV_A))
		3622	{
		3623	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3624	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3625	}
		3626	}
		3627
		3628	Now, whenever the main thread gets told to invoke pending watchers, it
		3629	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3630	thread to continue:
		3631
		3632	static void
		3633	real_invoke_pending (EV_P)
		3634	{
		3635	userdata *u = ev_userdata (EV_A);
		3636
		3637	pthread_mutex_lock (&u->lock);
		3638	ev_invoke_pending (EV_A);
		3639	pthread_cond_signal (&u->invoke_cv);
		3640	pthread_mutex_unlock (&u->lock);
		3641	}
		3642
		3643	Whenever you want to start/stop a watcher or do other modifications to an
		3644	event loop, you will now have to lock:
		3645
		3646	ev_timer timeout_watcher;
		3647	userdata *u = ev_userdata (EV_A);
		3648
		3649	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3650
		3651	pthread_mutex_lock (&u->lock);
		3652	ev_timer_start (EV_A_ &timeout_watcher);
		3653	ev_async_send (EV_A_ &u->async_w);
		3654	pthread_mutex_unlock (&u->lock);
		3655
		3656	Note that sending the C<ev_async> watcher is required because otherwise
		3657	an event loop currently blocking in the kernel will have no knowledge
		3658	about the newly added timer. By waking up the loop it will pick up any new
		3659	watchers in the next event loop iteration.
		3660
		3661	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3662
		3663	While the overhead of a callback that e.g. schedules a thread is small, it
		3664	is still an overhead. If you embed libev, and your main usage is with some
		3665	kind of threads or coroutines, you might want to customise libev so that
		3666	doesn't need callbacks anymore.
		3667
		3668	Imagine you have coroutines that you can switch to using a function
		3669	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3670	and that due to some magic, the currently active coroutine is stored in a
		3671	global called C<current_coro>. Then you can build your own "wait for libev
		3672	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3673	the differing C<;> conventions):
		3674
		3675	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3676	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3677
		3678	That means instead of having a C callback function, you store the
		3679	coroutine to switch to in each watcher, and instead of having libev call
		3680	your callback, you instead have it switch to that coroutine.
		3681
		3682	A coroutine might now wait for an event with a function called
		3683	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3684	matter when, or whether the watcher is active or not when this function is
		3685	called):
		3686
		3687	void
		3688	wait_for_event (ev_watcher *w)
		3689	{
		3690	ev_cb_set (w) = current_coro;
		3691	switch_to (libev_coro);
		3692	}
		3693
		3694	That basically suspends the coroutine inside C<wait_for_event> and
		3695	continues the libev coroutine, which, when appropriate, switches back to
		3696	this or any other coroutine. I am sure if you sue this your own :)
		3697
		3698	You can do similar tricks if you have, say, threads with an event queue -
		3699	instead of storing a coroutine, you store the queue object and instead of
		3700	switching to a coroutine, you push the watcher onto the queue and notify
		3701	any waiters.
		3702
		3703	To embed libev, see L<EMBEDDING>, but in short, it's easiest to create two
		3704	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3705
		3706	// my_ev.h
		3707	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3708	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb);
		3709	#include "../libev/ev.h"
		3710
		3711	// my_ev.c
		3712	#define EV_H "my_ev.h"
		3713	#include "../libev/ev.c"
		3714
		3715	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3716	F<my_ev.c> into your project. When properly specifying include paths, you
		3717	can even use F<ev.h> as header file name directly.
2463		3718
2464		3719
2465	=head1 LIBEVENT EMULATION	3720	=head1 LIBEVENT EMULATION
2466		3721
2467	Libev offers a compatibility emulation layer for libevent. It cannot	3722	Libev offers a compatibility emulation layer for libevent. It cannot
2468	emulate the internals of libevent, so here are some usage hints:	3723	emulate the internals of libevent, so here are some usage hints:
2469		3724
2470	=over 4	3725	=over 4
		3726
		3727	=item * Only the libevent-1.4.1-beta API is being emulated.
		3728
		3729	This was the newest libevent version available when libev was implemented,
		3730	and is still mostly unchanged in 2010.
2471		3731
2472	=item * Use it by including <event.h>, as usual.	3732	=item * Use it by including <event.h>, as usual.
2473		3733
2474	=item * The following members are fully supported: ev_base, ev_callback,	3734	=item * The following members are fully supported: ev_base, ev_callback,
2475	ev_arg, ev_fd, ev_res, ev_events.	3735	ev_arg, ev_fd, ev_res, ev_events.
…		…
2481	=item * Priorities are not currently supported. Initialising priorities	3741	=item * Priorities are not currently supported. Initialising priorities
2482	will fail and all watchers will have the same priority, even though there	3742	will fail and all watchers will have the same priority, even though there
2483	is an ev_pri field.	3743	is an ev_pri field.
2484		3744
2485	=item * In libevent, the last base created gets the signals, in libev, the	3745	=item * In libevent, the last base created gets the signals, in libev, the
2486	first base created (== the default loop) gets the signals.	3746	base that registered the signal gets the signals.
2487		3747
2488	=item * Other members are not supported.	3748	=item * Other members are not supported.
2489		3749
2490	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3750	=item * The libev emulation is I<not> ABI compatible to libevent, you need
2491	to use the libev header file and library.	3751	to use the libev header file and library.
…		…
2510	Care has been taken to keep the overhead low. The only data member the C++	3770	Care has been taken to keep the overhead low. The only data member the C++
2511	classes add (compared to plain C-style watchers) is the event loop pointer	3771	classes add (compared to plain C-style watchers) is the event loop pointer
2512	that the watcher is associated with (or no additional members at all if	3772	that the watcher is associated with (or no additional members at all if
2513	you disable C<EV_MULTIPLICITY> when embedding libev).	3773	you disable C<EV_MULTIPLICITY> when embedding libev).
2514		3774
2515	Currently, functions, and static and non-static member functions can be	3775	Currently, functions, static and non-static member functions and classes
2516	used as callbacks. Other types should be easy to add as long as they only	3776	with C<operator ()> can be used as callbacks. Other types should be easy
2517	need one additional pointer for context. If you need support for other	3777	to add as long as they only need one additional pointer for context. If
2518	types of functors please contact the author (preferably after implementing	3778	you need support for other types of functors please contact the author
2519	it).	3779	(preferably after implementing it).
2520		3780
2521	Here is a list of things available in the C<ev> namespace:	3781	Here is a list of things available in the C<ev> namespace:
2522		3782
2523	=over 4	3783	=over 4
2524		3784
…		…
2542		3802
2543	=over 4	3803	=over 4
2544		3804
2545	=item ev::TYPE::TYPE ()	3805	=item ev::TYPE::TYPE ()
2546		3806
2547	=item ev::TYPE::TYPE (struct ev_loop *)	3807	=item ev::TYPE::TYPE (loop)
2548		3808
2549	=item ev::TYPE::~TYPE	3809	=item ev::TYPE::~TYPE
2550		3810
2551	The constructor (optionally) takes an event loop to associate the watcher	3811	The constructor (optionally) takes an event loop to associate the watcher
2552	with. If it is omitted, it will use C<EV_DEFAULT>.	3812	with. If it is omitted, it will use C<EV_DEFAULT>.
…		…
2584		3844
2585	myclass obj;	3845	myclass obj;
2586	ev::io iow;	3846	ev::io iow;
2587	iow.set <myclass, &myclass::io_cb> (&obj);	3847	iow.set <myclass, &myclass::io_cb> (&obj);
2588		3848
		3849	=item w->set (object *)
		3850
		3851	This is a variation of a method callback - leaving out the method to call
		3852	will default the method to C<operator ()>, which makes it possible to use
		3853	functor objects without having to manually specify the C<operator ()> all
		3854	the time. Incidentally, you can then also leave out the template argument
		3855	list.
		3856
		3857	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3858	int revents)>.
		3859
		3860	See the method-C<set> above for more details.
		3861
		3862	Example: use a functor object as callback.
		3863
		3864	struct myfunctor
		3865	{
		3866	void operator() (ev::io &w, int revents)
		3867	{
		3868	...
		3869	}
		3870	}
		3871
		3872	myfunctor f;
		3873
		3874	ev::io w;
		3875	w.set (&f);
		3876
2589	=item w->set<function> (void *data = 0)	3877	=item w->set<function> (void *data = 0)
2590		3878
2591	Also sets a callback, but uses a static method or plain function as	3879	Also sets a callback, but uses a static method or plain function as
2592	callback. The optional C<data> argument will be stored in the watcher's	3880	callback. The optional C<data> argument will be stored in the watcher's
2593	C<data> member and is free for you to use.	3881	C<data> member and is free for you to use.
2594		3882
2595	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.	3883	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2596		3884
2597	See the method-C<set> above for more details.	3885	See the method-C<set> above for more details.
2598		3886
2599	Example:	3887	Example: Use a plain function as callback.
2600		3888
2601	static void io_cb (ev::io &w, int revents) { }	3889	static void io_cb (ev::io &w, int revents) { }
2602	iow.set <io_cb> ();	3890	iow.set <io_cb> ();
2603		3891
2604	=item w->set (struct ev_loop *)	3892	=item w->set (loop)
2605		3893
2606	Associates a different C<struct ev_loop> with this watcher. You can only	3894	Associates a different C<struct ev_loop> with this watcher. You can only
2607	do this when the watcher is inactive (and not pending either).	3895	do this when the watcher is inactive (and not pending either).
2608		3896
2609	=item w->set ([arguments])	3897	=item w->set ([arguments])
2610		3898
2611	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be	3899	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
2612	called at least once. Unlike the C counterpart, an active watcher gets	3900	method or a suitable start method must be called at least once. Unlike the
2613	automatically stopped and restarted when reconfiguring it with this	3901	C counterpart, an active watcher gets automatically stopped and restarted
2614	method.	3902	when reconfiguring it with this method.
2615		3903
2616	=item w->start ()	3904	=item w->start ()
2617		3905
2618	Starts the watcher. Note that there is no C<loop> argument, as the	3906	Starts the watcher. Note that there is no C<loop> argument, as the
2619	constructor already stores the event loop.	3907	constructor already stores the event loop.
2620		3908
		3909	=item w->start ([arguments])
		3910
		3911	Instead of calling C<set> and C<start> methods separately, it is often
		3912	convenient to wrap them in one call. Uses the same type of arguments as
		3913	the configure C<set> method of the watcher.
		3914
2621	=item w->stop ()	3915	=item w->stop ()
2622		3916
2623	Stops the watcher if it is active. Again, no C<loop> argument.	3917	Stops the watcher if it is active. Again, no C<loop> argument.
2624		3918
2625	=item w->again () (C<ev::timer>, C<ev::periodic> only)	3919	=item w->again () (C<ev::timer>, C<ev::periodic> only)
…		…
2637		3931
2638	=back	3932	=back
2639		3933
2640	=back	3934	=back
2641		3935
2642	Example: Define a class with an IO and idle watcher, start one of them in	3936	Example: Define a class with two I/O and idle watchers, start the I/O
2643	the constructor.	3937	watchers in the constructor.
2644		3938
2645	class myclass	3939	class myclass
2646	{	3940	{
2647	ev::io io; void io_cb (ev::io &w, int revents);	3941	ev::io io ; void io_cb (ev::io &w, int revents);
		3942	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
2648	ev:idle idle void idle_cb (ev::idle &w, int revents);	3943	ev::idle idle; void idle_cb (ev::idle &w, int revents);
2649		3944
2650	myclass (int fd)	3945	myclass (int fd)
2651	{	3946	{
2652	io .set <myclass, &myclass::io_cb > (this);	3947	io .set <myclass, &myclass::io_cb > (this);
		3948	io2 .set <myclass, &myclass::io2_cb > (this);
2653	idle.set <myclass, &myclass::idle_cb> (this);	3949	idle.set <myclass, &myclass::idle_cb> (this);
2654		3950
2655	io.start (fd, ev::READ);	3951	io.set (fd, ev::WRITE); // configure the watcher
		3952	io.start (); // start it whenever convenient
		3953
		3954	io2.start (fd, ev::READ); // set + start in one call
2656	}	3955	}
2657	};	3956	};
2658		3957
2659		3958
2660	=head1 OTHER LANGUAGE BINDINGS	3959	=head1 OTHER LANGUAGE BINDINGS
…		…
2669	=item Perl	3968	=item Perl
2670		3969
2671	The EV module implements the full libev API and is actually used to test	3970	The EV module implements the full libev API and is actually used to test
2672	libev. EV is developed together with libev. Apart from the EV core module,	3971	libev. EV is developed together with libev. Apart from the EV core module,
2673	there are additional modules that implement libev-compatible interfaces	3972	there are additional modules that implement libev-compatible interfaces
2674	to C<libadns> (C<EV::ADNS>), C<Net::SNMP> (C<Net::SNMP::EV>) and the	3973	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
2675	C<libglib> event core (C<Glib::EV> and C<EV::Glib>).	3974	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3975	and C<EV::Glib>).
2676		3976
2677	It can be found and installed via CPAN, its homepage is at	3977	It can be found and installed via CPAN, its homepage is at
2678	L<http://software.schmorp.de/pkg/EV>.	3978	L<http://software.schmorp.de/pkg/EV>.
2679		3979
2680	=item Python	3980	=item Python
2681		3981
2682	Python bindings can be found at L<http://code.google.com/p/pyev/>. It	3982	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
2683	seems to be quite complete and well-documented. Note, however, that the	3983	seems to be quite complete and well-documented.
2684	patch they require for libev is outright dangerous as it breaks the ABI
2685	for everybody else, and therefore, should never be applied in an installed
2686	libev (if python requires an incompatible ABI then it needs to embed
2687	libev).
2688		3984
2689	=item Ruby	3985	=item Ruby
2690		3986
2691	Tony Arcieri has written a ruby extension that offers access to a subset	3987	Tony Arcieri has written a ruby extension that offers access to a subset
2692	of the libev API and adds file handle abstractions, asynchronous DNS and	3988	of the libev API and adds file handle abstractions, asynchronous DNS and
2693	more on top of it. It can be found via gem servers. Its homepage is at	3989	more on top of it. It can be found via gem servers. Its homepage is at
2694	L<http://rev.rubyforge.org/>.	3990	L<http://rev.rubyforge.org/>.
2695		3991
		3992	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3993	makes rev work even on mingw.
		3994
		3995	=item Haskell
		3996
		3997	A haskell binding to libev is available at
		3998	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3999
2696	=item D	4000	=item D
2697		4001
2698	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	4002	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
2699	be found at L<http://proj.llucax.com.ar/wiki/evd>.	4003	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		4004
		4005	=item Ocaml
		4006
		4007	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4008	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4009
		4010	=item Lua
		4011
		4012	Brian Maher has written a partial interface to libev for lua (at the
		4013	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4014	L<http://github.com/brimworks/lua-ev>.
2700		4015
2701	=back	4016	=back
2702		4017
2703		4018
2704	=head1 MACRO MAGIC	4019	=head1 MACRO MAGIC
…		…
2718	loop argument"). The C<EV_A> form is used when this is the sole argument,	4033	loop argument"). The C<EV_A> form is used when this is the sole argument,
2719	C<EV_A_> is used when other arguments are following. Example:	4034	C<EV_A_> is used when other arguments are following. Example:
2720		4035
2721	ev_unref (EV_A);	4036	ev_unref (EV_A);
2722	ev_timer_add (EV_A_ watcher);	4037	ev_timer_add (EV_A_ watcher);
2723	ev_loop (EV_A_ 0);	4038	ev_run (EV_A_ 0);
2724		4039
2725	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,	4040	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2726	which is often provided by the following macro.	4041	which is often provided by the following macro.
2727		4042
2728	=item C<EV_P>, C<EV_P_>	4043	=item C<EV_P>, C<EV_P_>
…		…
2768	}	4083	}
2769		4084
2770	ev_check check;	4085	ev_check check;
2771	ev_check_init (&check, check_cb);	4086	ev_check_init (&check, check_cb);
2772	ev_check_start (EV_DEFAULT_ &check);	4087	ev_check_start (EV_DEFAULT_ &check);
2773	ev_loop (EV_DEFAULT_ 0);	4088	ev_run (EV_DEFAULT_ 0);
2774		4089
2775	=head1 EMBEDDING	4090	=head1 EMBEDDING
2776		4091
2777	Libev can (and often is) directly embedded into host	4092	Libev can (and often is) directly embedded into host
2778	applications. Examples of applications that embed it include the Deliantra	4093	applications. Examples of applications that embed it include the Deliantra
…		…
2805		4120
2806	#define EV_STANDALONE 1	4121	#define EV_STANDALONE 1
2807	#include "ev.h"	4122	#include "ev.h"
2808		4123
2809	Both header files and implementation files can be compiled with a C++	4124	Both header files and implementation files can be compiled with a C++
2810	compiler (at least, thats a stated goal, and breakage will be treated	4125	compiler (at least, that's a stated goal, and breakage will be treated
2811	as a bug).	4126	as a bug).
2812		4127
2813	You need the following files in your source tree, or in a directory	4128	You need the following files in your source tree, or in a directory
2814	in your include path (e.g. in libev/ when using -Ilibev):	4129	in your include path (e.g. in libev/ when using -Ilibev):
2815		4130
…		…
2858	libev.m4	4173	libev.m4
2859		4174
2860	=head2 PREPROCESSOR SYMBOLS/MACROS	4175	=head2 PREPROCESSOR SYMBOLS/MACROS
2861		4176
2862	Libev can be configured via a variety of preprocessor symbols you have to	4177	Libev can be configured via a variety of preprocessor symbols you have to
2863	define before including any of its files. The default in the absence of	4178	define before including (or compiling) any of its files. The default in
2864	autoconf is noted for every option.	4179	the absence of autoconf is documented for every option.
		4180
		4181	Symbols marked with "(h)" do not change the ABI, and can have different
		4182	values when compiling libev vs. including F<ev.h>, so it is permissible
		4183	to redefine them before including F<ev.h> without breaking compatibility
		4184	to a compiled library. All other symbols change the ABI, which means all
		4185	users of libev and the libev code itself must be compiled with compatible
		4186	settings.
2865		4187
2866	=over 4	4188	=over 4
2867		4189
		4190	=item EV_COMPAT3 (h)
		4191
		4192	Backwards compatibility is a major concern for libev. This is why this
		4193	release of libev comes with wrappers for the functions and symbols that
		4194	have been renamed between libev version 3 and 4.
		4195
		4196	You can disable these wrappers (to test compatibility with future
		4197	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4198	sources. This has the additional advantage that you can drop the C<struct>
		4199	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4200	typedef in that case.
		4201
		4202	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4203	and in some even more future version the compatibility code will be
		4204	removed completely.
		4205
2868	=item EV_STANDALONE	4206	=item EV_STANDALONE (h)
2869		4207
2870	Must always be C<1> if you do not use autoconf configuration, which	4208	Must always be C<1> if you do not use autoconf configuration, which
2871	keeps libev from including F<config.h>, and it also defines dummy	4209	keeps libev from including F<config.h>, and it also defines dummy
2872	implementations for some libevent functions (such as logging, which is not	4210	implementations for some libevent functions (such as logging, which is not
2873	supported). It will also not define any of the structs usually found in	4211	supported). It will also not define any of the structs usually found in
2874	F<event.h> that are not directly supported by the libev core alone.	4212	F<event.h> that are not directly supported by the libev core alone.
2875		4213
		4214	In standalone mode, libev will still try to automatically deduce the
		4215	configuration, but has to be more conservative.
		4216
		4217	=item EV_USE_FLOOR
		4218
		4219	If defined to be C<1>, libev will use the C<floor ()> function for its
		4220	periodic reschedule calculations, otherwise libev will fall back on a
		4221	portable (slower) implementation. If you enable this, you usually have to
		4222	link against libm or something equivalent. Enabling this when the C<floor>
		4223	function is not available will fail, so the safe default is to not enable
		4224	this.
		4225
2876	=item EV_USE_MONOTONIC	4226	=item EV_USE_MONOTONIC
2877		4227
2878	If defined to be C<1>, libev will try to detect the availability of the	4228	If defined to be C<1>, libev will try to detect the availability of the
2879	monotonic clock option at both compile time and runtime. Otherwise no use	4229	monotonic clock option at both compile time and runtime. Otherwise no
2880	of the monotonic clock option will be attempted. If you enable this, you	4230	use of the monotonic clock option will be attempted. If you enable this,
2881	usually have to link against librt or something similar. Enabling it when	4231	you usually have to link against librt or something similar. Enabling it
2882	the functionality isn't available is safe, though, although you have	4232	when the functionality isn't available is safe, though, although you have
2883	to make sure you link against any libraries where the C<clock_gettime>	4233	to make sure you link against any libraries where the C<clock_gettime>
2884	function is hiding in (often F<-lrt>).	4234	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
2885		4235
2886	=item EV_USE_REALTIME	4236	=item EV_USE_REALTIME
2887		4237
2888	If defined to be C<1>, libev will try to detect the availability of the	4238	If defined to be C<1>, libev will try to detect the availability of the
2889	real-time clock option at compile time (and assume its availability at	4239	real-time clock option at compile time (and assume its availability
2890	runtime if successful). Otherwise no use of the real-time clock option will	4240	at runtime if successful). Otherwise no use of the real-time clock
2891	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	4241	option will be attempted. This effectively replaces C<gettimeofday>
2892	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the	4242	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
2893	note about libraries in the description of C<EV_USE_MONOTONIC>, though.	4243	correctness. See the note about libraries in the description of
		4244	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4245	C<EV_USE_CLOCK_SYSCALL>.
		4246
		4247	=item EV_USE_CLOCK_SYSCALL
		4248
		4249	If defined to be C<1>, libev will try to use a direct syscall instead
		4250	of calling the system-provided C<clock_gettime> function. This option
		4251	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4252	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4253	programs needlessly. Using a direct syscall is slightly slower (in
		4254	theory), because no optimised vdso implementation can be used, but avoids
		4255	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4256	higher, as it simplifies linking (no need for C<-lrt>).
2894		4257
2895	=item EV_USE_NANOSLEEP	4258	=item EV_USE_NANOSLEEP
2896		4259
2897	If defined to be C<1>, libev will assume that C<nanosleep ()> is available	4260	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
2898	and will use it for delays. Otherwise it will use C<select ()>.	4261	and will use it for delays. Otherwise it will use C<select ()>.
…		…
2914		4277
2915	=item EV_SELECT_USE_FD_SET	4278	=item EV_SELECT_USE_FD_SET
2916		4279
2917	If defined to C<1>, then the select backend will use the system C<fd_set>	4280	If defined to C<1>, then the select backend will use the system C<fd_set>
2918	structure. This is useful if libev doesn't compile due to a missing	4281	structure. This is useful if libev doesn't compile due to a missing
2919	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on	4282	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
2920	exotic systems. This usually limits the range of file descriptors to some	4283	on exotic systems. This usually limits the range of file descriptors to
2921	low limit such as 1024 or might have other limitations (winsocket only	4284	some low limit such as 1024 or might have other limitations (winsocket
2922	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might	4285	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
2923	influence the size of the C<fd_set> used.	4286	configures the maximum size of the C<fd_set>.
2924		4287
2925	=item EV_SELECT_IS_WINSOCKET	4288	=item EV_SELECT_IS_WINSOCKET
2926		4289
2927	When defined to C<1>, the select backend will assume that	4290	When defined to C<1>, the select backend will assume that
2928	select/socket/connect etc. don't understand file descriptors but	4291	select/socket/connect etc. don't understand file descriptors but
…		…
2930	be used is the winsock select). This means that it will call	4293	be used is the winsock select). This means that it will call
2931	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,	4294	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2932	it is assumed that all these functions actually work on fds, even	4295	it is assumed that all these functions actually work on fds, even
2933	on win32. Should not be defined on non-win32 platforms.	4296	on win32. Should not be defined on non-win32 platforms.
2934		4297
2935	=item EV_FD_TO_WIN32_HANDLE	4298	=item EV_FD_TO_WIN32_HANDLE(fd)
2936		4299
2937	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map	4300	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
2938	file descriptors to socket handles. When not defining this symbol (the	4301	file descriptors to socket handles. When not defining this symbol (the
2939	default), then libev will call C<_get_osfhandle>, which is usually	4302	default), then libev will call C<_get_osfhandle>, which is usually
2940	correct. In some cases, programs use their own file descriptor management,	4303	correct. In some cases, programs use their own file descriptor management,
2941	in which case they can provide this function to map fds to socket handles.	4304	in which case they can provide this function to map fds to socket handles.
		4305
		4306	=item EV_WIN32_HANDLE_TO_FD(handle)
		4307
		4308	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4309	using the standard C<_open_osfhandle> function. For programs implementing
		4310	their own fd to handle mapping, overwriting this function makes it easier
		4311	to do so. This can be done by defining this macro to an appropriate value.
		4312
		4313	=item EV_WIN32_CLOSE_FD(fd)
		4314
		4315	If programs implement their own fd to handle mapping on win32, then this
		4316	macro can be used to override the C<close> function, useful to unregister
		4317	file descriptors again. Note that the replacement function has to close
		4318	the underlying OS handle.
2942		4319
2943	=item EV_USE_POLL	4320	=item EV_USE_POLL
2944		4321
2945	If defined to be C<1>, libev will compile in support for the C<poll>(2)	4322	If defined to be C<1>, libev will compile in support for the C<poll>(2)
2946	backend. Otherwise it will be enabled on non-win32 platforms. It	4323	backend. Otherwise it will be enabled on non-win32 platforms. It
…		…
2993	as well as for signal and thread safety in C<ev_async> watchers.	4370	as well as for signal and thread safety in C<ev_async> watchers.
2994		4371
2995	In the absence of this define, libev will use C<sig_atomic_t volatile>	4372	In the absence of this define, libev will use C<sig_atomic_t volatile>
2996	(from F<signal.h>), which is usually good enough on most platforms.	4373	(from F<signal.h>), which is usually good enough on most platforms.
2997		4374
2998	=item EV_H	4375	=item EV_H (h)
2999		4376
3000	The name of the F<ev.h> header file used to include it. The default if	4377	The name of the F<ev.h> header file used to include it. The default if
3001	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be	4378	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
3002	used to virtually rename the F<ev.h> header file in case of conflicts.	4379	used to virtually rename the F<ev.h> header file in case of conflicts.
3003		4380
3004	=item EV_CONFIG_H	4381	=item EV_CONFIG_H (h)
3005		4382
3006	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override	4383	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
3007	F<ev.c>'s idea of where to find the F<config.h> file, similarly to	4384	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
3008	C<EV_H>, above.	4385	C<EV_H>, above.
3009		4386
3010	=item EV_EVENT_H	4387	=item EV_EVENT_H (h)
3011		4388
3012	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea	4389	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
3013	of how the F<event.h> header can be found, the default is C<"event.h">.	4390	of how the F<event.h> header can be found, the default is C<"event.h">.
3014		4391
3015	=item EV_PROTOTYPES	4392	=item EV_PROTOTYPES (h)
3016		4393
3017	If defined to be C<0>, then F<ev.h> will not define any function	4394	If defined to be C<0>, then F<ev.h> will not define any function
3018	prototypes, but still define all the structs and other symbols. This is	4395	prototypes, but still define all the structs and other symbols. This is
3019	occasionally useful if you want to provide your own wrapper functions	4396	occasionally useful if you want to provide your own wrapper functions
3020	around libev functions.	4397	around libev functions.
…		…
3039	When doing priority-based operations, libev usually has to linearly search	4416	When doing priority-based operations, libev usually has to linearly search
3040	all the priorities, so having many of them (hundreds) uses a lot of space	4417	all the priorities, so having many of them (hundreds) uses a lot of space
3041	and time, so using the defaults of five priorities (-2 .. +2) is usually	4418	and time, so using the defaults of five priorities (-2 .. +2) is usually
3042	fine.	4419	fine.
3043		4420
3044	If your embedding application does not need any priorities, defining these both to	4421	If your embedding application does not need any priorities, defining these
3045	C<0> will save some memory and CPU.	4422	both to C<0> will save some memory and CPU.
3046		4423
3047	=item EV_PERIODIC_ENABLE	4424	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4425	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4426	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
3048		4427
3049	If undefined or defined to be C<1>, then periodic timers are supported. If	4428	If undefined or defined to be C<1> (and the platform supports it), then
3050	defined to be C<0>, then they are not. Disabling them saves a few kB of	4429	the respective watcher type is supported. If defined to be C<0>, then it
3051	code.	4430	is not. Disabling watcher types mainly saves code size.
3052		4431
3053	=item EV_IDLE_ENABLE	4432	=item EV_FEATURES
3054
3055	If undefined or defined to be C<1>, then idle watchers are supported. If
3056	defined to be C<0>, then they are not. Disabling them saves a few kB of
3057	code.
3058
3059	=item EV_EMBED_ENABLE
3060
3061	If undefined or defined to be C<1>, then embed watchers are supported. If
3062	defined to be C<0>, then they are not.
3063
3064	=item EV_STAT_ENABLE
3065
3066	If undefined or defined to be C<1>, then stat watchers are supported. If
3067	defined to be C<0>, then they are not.
3068
3069	=item EV_FORK_ENABLE
3070
3071	If undefined or defined to be C<1>, then fork watchers are supported. If
3072	defined to be C<0>, then they are not.
3073
3074	=item EV_ASYNC_ENABLE
3075
3076	If undefined or defined to be C<1>, then async watchers are supported. If
3077	defined to be C<0>, then they are not.
3078
3079	=item EV_MINIMAL
3080		4433
3081	If you need to shave off some kilobytes of code at the expense of some	4434	If you need to shave off some kilobytes of code at the expense of some
3082	speed, define this symbol to C<1>. Currently this is used to override some	4435	speed (but with the full API), you can define this symbol to request
3083	inlining decisions, saves roughly 30% code size on amd64. It also selects a	4436	certain subsets of functionality. The default is to enable all features
3084	much smaller 2-heap for timer management over the default 4-heap.	4437	that can be enabled on the platform.
		4438
		4439	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4440	with some broad features you want) and then selectively re-enable
		4441	additional parts you want, for example if you want everything minimal,
		4442	but multiple event loop support, async and child watchers and the poll
		4443	backend, use this:
		4444
		4445	#define EV_FEATURES 0
		4446	#define EV_MULTIPLICITY 1
		4447	#define EV_USE_POLL 1
		4448	#define EV_CHILD_ENABLE 1
		4449	#define EV_ASYNC_ENABLE 1
		4450
		4451	The actual value is a bitset, it can be a combination of the following
		4452	values:
		4453
		4454	=over 4
		4455
		4456	=item C<1> - faster/larger code
		4457
		4458	Use larger code to speed up some operations.
		4459
		4460	Currently this is used to override some inlining decisions (enlarging the
		4461	code size by roughly 30% on amd64).
		4462
		4463	When optimising for size, use of compiler flags such as C<-Os> with
		4464	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4465	assertions.
		4466
		4467	=item C<2> - faster/larger data structures
		4468
		4469	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4470	hash table sizes and so on. This will usually further increase code size
		4471	and can additionally have an effect on the size of data structures at
		4472	runtime.
		4473
		4474	=item C<4> - full API configuration
		4475
		4476	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4477	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4478
		4479	=item C<8> - full API
		4480
		4481	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4482	details on which parts of the API are still available without this
		4483	feature, and do not complain if this subset changes over time.
		4484
		4485	=item C<16> - enable all optional watcher types
		4486
		4487	Enables all optional watcher types. If you want to selectively enable
		4488	only some watcher types other than I/O and timers (e.g. prepare,
		4489	embed, async, child...) you can enable them manually by defining
		4490	C<EV_watchertype_ENABLE> to C<1> instead.
		4491
		4492	=item C<32> - enable all backends
		4493
		4494	This enables all backends - without this feature, you need to enable at
		4495	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4496
		4497	=item C<64> - enable OS-specific "helper" APIs
		4498
		4499	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4500	default.
		4501
		4502	=back
		4503
		4504	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4505	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4506	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4507	watchers, timers and monotonic clock support.
		4508
		4509	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4510	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4511	your program might be left out as well - a binary starting a timer and an
		4512	I/O watcher then might come out at only 5Kb.
		4513
		4514	=item EV_AVOID_STDIO
		4515
		4516	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4517	functions (printf, scanf, perror etc.). This will increase the code size
		4518	somewhat, but if your program doesn't otherwise depend on stdio and your
		4519	libc allows it, this avoids linking in the stdio library which is quite
		4520	big.
		4521
		4522	Note that error messages might become less precise when this option is
		4523	enabled.
		4524
		4525	=item EV_NSIG
		4526
		4527	The highest supported signal number, +1 (or, the number of
		4528	signals): Normally, libev tries to deduce the maximum number of signals
		4529	automatically, but sometimes this fails, in which case it can be
		4530	specified. Also, using a lower number than detected (C<32> should be
		4531	good for about any system in existence) can save some memory, as libev
		4532	statically allocates some 12-24 bytes per signal number.
3085		4533
3086	=item EV_PID_HASHSIZE	4534	=item EV_PID_HASHSIZE
3087		4535
3088	C<ev_child> watchers use a small hash table to distribute workload by	4536	C<ev_child> watchers use a small hash table to distribute workload by
3089	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more	4537	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
3090	than enough. If you need to manage thousands of children you might want to	4538	usually more than enough. If you need to manage thousands of children you
3091	increase this value (I<must> be a power of two).	4539	might want to increase this value (I<must> be a power of two).
3092		4540
3093	=item EV_INOTIFY_HASHSIZE	4541	=item EV_INOTIFY_HASHSIZE
3094		4542
3095	C<ev_stat> watchers use a small hash table to distribute workload by	4543	C<ev_stat> watchers use a small hash table to distribute workload by
3096	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),	4544	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
3097	usually more than enough. If you need to manage thousands of C<ev_stat>	4545	disabled), usually more than enough. If you need to manage thousands of
3098	watchers you might want to increase this value (I<must> be a power of	4546	C<ev_stat> watchers you might want to increase this value (I<must> be a
3099	two).	4547	power of two).
3100		4548
3101	=item EV_USE_4HEAP	4549	=item EV_USE_4HEAP
3102		4550
3103	Heaps are not very cache-efficient. To improve the cache-efficiency of the	4551	Heaps are not very cache-efficient. To improve the cache-efficiency of the
3104	timer and periodics heap, libev uses a 4-heap when this symbol is defined	4552	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
3105	to C<1>. The 4-heap uses more complicated (longer) code but has	4553	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
3106	noticeably faster performance with many (thousands) of watchers.	4554	faster performance with many (thousands) of watchers.
3107		4555
3108	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>	4556	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3109	(disabled).	4557	will be C<0>.
3110		4558
3111	=item EV_HEAP_CACHE_AT	4559	=item EV_HEAP_CACHE_AT
3112		4560
3113	Heaps are not very cache-efficient. To improve the cache-efficiency of the	4561	Heaps are not very cache-efficient. To improve the cache-efficiency of the
3114	timer and periodics heap, libev can cache the timestamp (I<at>) within	4562	timer and periodics heaps, libev can cache the timestamp (I<at>) within
3115	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),	4563	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
3116	which uses 8-12 bytes more per watcher and a few hundred bytes more code,	4564	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
3117	but avoids random read accesses on heap changes. This improves performance	4565	but avoids random read accesses on heap changes. This improves performance
3118	noticeably with with many (hundreds) of watchers.	4566	noticeably with many (hundreds) of watchers.
3119		4567
3120	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>	4568	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3121	(disabled).	4569	will be C<0>.
3122		4570
3123	=item EV_VERIFY	4571	=item EV_VERIFY
3124		4572
3125	Controls how much internal verification (see C<ev_loop_verify ()>) will	4573	Controls how much internal verification (see C<ev_verify ()>) will
3126	be done: If set to C<0>, no internal verification code will be compiled	4574	be done: If set to C<0>, no internal verification code will be compiled
3127	in. If set to C<1>, then verification code will be compiled in, but not	4575	in. If set to C<1>, then verification code will be compiled in, but not
3128	called. If set to C<2>, then the internal verification code will be	4576	called. If set to C<2>, then the internal verification code will be
3129	called once per loop, which can slow down libev. If set to C<3>, then the	4577	called once per loop, which can slow down libev. If set to C<3>, then the
3130	verification code will be called very frequently, which will slow down	4578	verification code will be called very frequently, which will slow down
3131	libev considerably.	4579	libev considerably.
3132		4580
3133	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be	4581	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
3134	C<0.>	4582	will be C<0>.
3135		4583
3136	=item EV_COMMON	4584	=item EV_COMMON
3137		4585
3138	By default, all watchers have a C<void *data> member. By redefining	4586	By default, all watchers have a C<void *data> member. By redefining
3139	this macro to a something else you can include more and other types of	4587	this macro to something else you can include more and other types of
3140	members. You have to define it each time you include one of the files,	4588	members. You have to define it each time you include one of the files,
3141	though, and it must be identical each time.	4589	though, and it must be identical each time.
3142		4590
3143	For example, the perl EV module uses something like this:	4591	For example, the perl EV module uses something like this:
3144		4592
…		…
3156	and the way callbacks are invoked and set. Must expand to a struct member	4604	and the way callbacks are invoked and set. Must expand to a struct member
3157	definition and a statement, respectively. See the F<ev.h> header file for	4605	definition and a statement, respectively. See the F<ev.h> header file for
3158	their default definitions. One possible use for overriding these is to	4606	their default definitions. One possible use for overriding these is to
3159	avoid the C<struct ev_loop *> as first argument in all cases, or to use	4607	avoid the C<struct ev_loop *> as first argument in all cases, or to use
3160	method calls instead of plain function calls in C++.	4608	method calls instead of plain function calls in C++.
		4609
		4610	=back
3161		4611
3162	=head2 EXPORTED API SYMBOLS	4612	=head2 EXPORTED API SYMBOLS
3163		4613
3164	If you need to re-export the API (e.g. via a DLL) and you need a list of	4614	If you need to re-export the API (e.g. via a DLL) and you need a list of
3165	exported symbols, you can use the provided F<Symbol.*> files which list	4615	exported symbols, you can use the provided F<Symbol.*> files which list
…		…
3195	file.	4645	file.
3196		4646
3197	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file	4647	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
3198	that everybody includes and which overrides some configure choices:	4648	that everybody includes and which overrides some configure choices:
3199		4649
3200	#define EV_MINIMAL 1	4650	#define EV_FEATURES 8
3201	#define EV_USE_POLL 0	4651	#define EV_USE_SELECT 1
3202	#define EV_MULTIPLICITY 0
3203	#define EV_PERIODIC_ENABLE 0	4652	#define EV_PREPARE_ENABLE 1
		4653	#define EV_IDLE_ENABLE 1
3204	#define EV_STAT_ENABLE 0	4654	#define EV_SIGNAL_ENABLE 1
3205	#define EV_FORK_ENABLE 0	4655	#define EV_CHILD_ENABLE 1
		4656	#define EV_USE_STDEXCEPT 0
3206	#define EV_CONFIG_H <config.h>	4657	#define EV_CONFIG_H <config.h>
3207	#define EV_MINPRI 0
3208	#define EV_MAXPRI 0
3209		4658
3210	#include "ev++.h"	4659	#include "ev++.h"
3211		4660
3212	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4661	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3213		4662
3214	#include "ev_cpp.h"	4663	#include "ev_cpp.h"
3215	#include "ev.c"	4664	#include "ev.c"
3216		4665
		4666	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
3217		4667
3218	=head1 THREADS AND COROUTINES	4668	=head2 THREADS AND COROUTINES
3219		4669
3220	=head2 THREADS	4670	=head3 THREADS
3221		4671
3222	Libev itself is completely thread-safe, but it uses no locking. This	4672	All libev functions are reentrant and thread-safe unless explicitly
		4673	documented otherwise, but libev implements no locking itself. This means
3223	means that you can use as many loops as you want in parallel, as long as	4674	that you can use as many loops as you want in parallel, as long as there
3224	only one thread ever calls into one libev function with the same loop	4675	are no concurrent calls into any libev function with the same loop
3225	parameter.	4676	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4677	of course): libev guarantees that different event loops share no data
		4678	structures that need any locking.
3226		4679
3227	Or put differently: calls with different loop parameters can be done in	4680	Or to put it differently: calls with different loop parameters can be done
3228	parallel from multiple threads, calls with the same loop parameter must be	4681	concurrently from multiple threads, calls with the same loop parameter
3229	done serially (but can be done from different threads, as long as only one	4682	must be done serially (but can be done from different threads, as long as
3230	thread ever is inside a call at any point in time, e.g. by using a mutex	4683	only one thread ever is inside a call at any point in time, e.g. by using
3231	per loop).	4684	a mutex per loop).
		4685
		4686	Specifically to support threads (and signal handlers), libev implements
		4687	so-called C<ev_async> watchers, which allow some limited form of
		4688	concurrency on the same event loop, namely waking it up "from the
		4689	outside".
3232		4690
3233	If you want to know which design (one loop, locking, or multiple loops	4691	If you want to know which design (one loop, locking, or multiple loops
3234	without or something else still) is best for your problem, then I cannot	4692	without or something else still) is best for your problem, then I cannot
3235	help you. I can give some generic advice however:	4693	help you, but here is some generic advice:
3236		4694
3237	=over 4	4695	=over 4
3238		4696
3239	=item * most applications have a main thread: use the default libev loop	4697	=item * most applications have a main thread: use the default libev loop
3240	in that thread, or create a separate thread running only the default loop.	4698	in that thread, or create a separate thread running only the default loop.
…		…
3252		4710
3253	Choosing a model is hard - look around, learn, know that usually you can do	4711	Choosing a model is hard - look around, learn, know that usually you can do
3254	better than you currently do :-)	4712	better than you currently do :-)
3255		4713
3256	=item * often you need to talk to some other thread which blocks in the	4714	=item * often you need to talk to some other thread which blocks in the
		4715	event loop.
		4716
3257	event loop - C<ev_async> watchers can be used to wake them up from other	4717	C<ev_async> watchers can be used to wake them up from other threads safely
3258	threads safely (or from signal contexts...).	4718	(or from signal contexts...).
		4719
		4720	An example use would be to communicate signals or other events that only
		4721	work in the default loop by registering the signal watcher with the
		4722	default loop and triggering an C<ev_async> watcher from the default loop
		4723	watcher callback into the event loop interested in the signal.
3259		4724
3260	=back	4725	=back
3261		4726
		4727	See also L<THREAD LOCKING EXAMPLE>.
		4728
3262	=head2 COROUTINES	4729	=head3 COROUTINES
3263		4730
3264	Libev is much more accommodating to coroutines ("cooperative threads"):	4731	Libev is very accommodating to coroutines ("cooperative threads"):
3265	libev fully supports nesting calls to it's functions from different	4732	libev fully supports nesting calls to its functions from different
3266	coroutines (e.g. you can call C<ev_loop> on the same loop from two	4733	coroutines (e.g. you can call C<ev_run> on the same loop from two
3267	different coroutines and switch freely between both coroutines running the	4734	different coroutines, and switch freely between both coroutines running
3268	loop, as long as you don't confuse yourself). The only exception is that	4735	the loop, as long as you don't confuse yourself). The only exception is
3269	you must not do this from C<ev_periodic> reschedule callbacks.	4736	that you must not do this from C<ev_periodic> reschedule callbacks.
3270		4737
3271	Care has been invested into making sure that libev does not keep local	4738	Care has been taken to ensure that libev does not keep local state inside
3272	state inside C<ev_loop>, and other calls do not usually allow coroutine	4739	C<ev_run>, and other calls do not usually allow for coroutine switches as
3273	switches.	4740	they do not call any callbacks.
3274		4741
		4742	=head2 COMPILER WARNINGS
3275		4743
3276	=head1 COMPLEXITIES	4744	Depending on your compiler and compiler settings, you might get no or a
		4745	lot of warnings when compiling libev code. Some people are apparently
		4746	scared by this.
3277		4747
3278	In this section the complexities of (many of) the algorithms used inside	4748	However, these are unavoidable for many reasons. For one, each compiler
3279	libev will be explained. For complexity discussions about backends see the	4749	has different warnings, and each user has different tastes regarding
3280	documentation for C<ev_default_init>.	4750	warning options. "Warn-free" code therefore cannot be a goal except when
		4751	targeting a specific compiler and compiler-version.
3281		4752
3282	All of the following are about amortised time: If an array needs to be	4753	Another reason is that some compiler warnings require elaborate
3283	extended, libev needs to realloc and move the whole array, but this	4754	workarounds, or other changes to the code that make it less clear and less
3284	happens asymptotically never with higher number of elements, so O(1) might	4755	maintainable.
3285	mean it might do a lengthy realloc operation in rare cases, but on average
3286	it is much faster and asymptotically approaches constant time.
3287		4756
3288	=over 4	4757	And of course, some compiler warnings are just plain stupid, or simply
		4758	wrong (because they don't actually warn about the condition their message
		4759	seems to warn about). For example, certain older gcc versions had some
		4760	warnings that resulted in an extreme number of false positives. These have
		4761	been fixed, but some people still insist on making code warn-free with
		4762	such buggy versions.
3289		4763
3290	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	4764	While libev is written to generate as few warnings as possible,
		4765	"warn-free" code is not a goal, and it is recommended not to build libev
		4766	with any compiler warnings enabled unless you are prepared to cope with
		4767	them (e.g. by ignoring them). Remember that warnings are just that:
		4768	warnings, not errors, or proof of bugs.
3291		4769
3292	This means that, when you have a watcher that triggers in one hour and
3293	there are 100 watchers that would trigger before that then inserting will
3294	have to skip roughly seven (C<ld 100>) of these watchers.
3295		4770
3296	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)	4771	=head2 VALGRIND
3297		4772
3298	That means that changing a timer costs less than removing/adding them	4773	Valgrind has a special section here because it is a popular tool that is
3299	as only the relative motion in the event queue has to be paid for.	4774	highly useful. Unfortunately, valgrind reports are very hard to interpret.
3300		4775
3301	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)	4776	If you think you found a bug (memory leak, uninitialised data access etc.)
		4777	in libev, then check twice: If valgrind reports something like:
3302		4778
3303	These just add the watcher into an array or at the head of a list.	4779	==2274== definitely lost: 0 bytes in 0 blocks.
		4780	==2274== possibly lost: 0 bytes in 0 blocks.
		4781	==2274== still reachable: 256 bytes in 1 blocks.
3304		4782
3305	=item Stopping check/prepare/idle/fork/async watchers: O(1)	4783	Then there is no memory leak, just as memory accounted to global variables
		4784	is not a memleak - the memory is still being referenced, and didn't leak.
3306		4785
3307	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	4786	Similarly, under some circumstances, valgrind might report kernel bugs
		4787	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4788	although an acceptable workaround has been found here), or it might be
		4789	confused.
3308		4790
3309	These watchers are stored in lists then need to be walked to find the	4791	Keep in mind that valgrind is a very good tool, but only a tool. Don't
3310	correct watcher to remove. The lists are usually short (you don't usually	4792	make it into some kind of religion.
3311	have many watchers waiting for the same fd or signal).
3312		4793
3313	=item Finding the next timer in each loop iteration: O(1)	4794	If you are unsure about something, feel free to contact the mailing list
		4795	with the full valgrind report and an explanation on why you think this
		4796	is a bug in libev (best check the archives, too :). However, don't be
		4797	annoyed when you get a brisk "this is no bug" answer and take the chance
		4798	of learning how to interpret valgrind properly.
3314		4799
3315	By virtue of using a binary or 4-heap, the next timer is always found at a	4800	If you need, for some reason, empty reports from valgrind for your project
3316	fixed position in the storage array.	4801	I suggest using suppression lists.
3317		4802
3318	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3319		4803
3320	A change means an I/O watcher gets started or stopped, which requires	4804	=head1 PORTABILITY NOTES
3321	libev to recalculate its status (and possibly tell the kernel, depending
3322	on backend and whether C<ev_io_set> was used).
3323		4805
3324	=item Activating one watcher (putting it into the pending state): O(1)	4806	=head2 GNU/LINUX 32 BIT LIMITATIONS
3325		4807
3326	=item Priority handling: O(number_of_priorities)	4808	GNU/Linux is the only common platform that supports 64 bit file/large file
		4809	interfaces but I<disables> them by default.
3327		4810
3328	Priorities are implemented by allocating some space for each	4811	That means that libev compiled in the default environment doesn't support
3329	priority. When doing priority-based operations, libev usually has to	4812	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
3330	linearly search all the priorities, but starting/stopping and activating
3331	watchers becomes O(1) w.r.t. priority handling.
3332		4813
3333	=item Sending an ev_async: O(1)	4814	Unfortunately, many programs try to work around this GNU/Linux issue
		4815	by enabling the large file API, which makes them incompatible with the
		4816	standard libev compiled for their system.
3334		4817
3335	=item Processing ev_async_send: O(number_of_async_watchers)	4818	Likewise, libev cannot enable the large file API itself as this would
		4819	suddenly make it incompatible to the default compile time environment,
		4820	i.e. all programs not using special compile switches.
3336		4821
3337	=item Processing signals: O(max_signal_number)	4822	=head2 OS/X AND DARWIN BUGS
3338		4823
3339	Sending involves a system call I<iff> there were no other C<ev_async_send>	4824	The whole thing is a bug if you ask me - basically any system interface
3340	calls in the current loop iteration. Checking for async and signal events	4825	you touch is broken, whether it is locales, poll, kqueue or even the
3341	involves iterating over all running async watchers or all signal numbers.	4826	OpenGL drivers.
3342		4827
3343	=back	4828	=head3 C<kqueue> is buggy
3344		4829
		4830	The kqueue syscall is broken in all known versions - most versions support
		4831	only sockets, many support pipes.
3345		4832
		4833	Libev tries to work around this by not using C<kqueue> by default on this
		4834	rotten platform, but of course you can still ask for it when creating a
		4835	loop - embedding a socket-only kqueue loop into a select-based one is
		4836	probably going to work well.
		4837
		4838	=head3 C<poll> is buggy
		4839
		4840	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4841	implementation by something calling C<kqueue> internally around the 10.5.6
		4842	release, so now C<kqueue> I<and> C<poll> are broken.
		4843
		4844	Libev tries to work around this by not using C<poll> by default on
		4845	this rotten platform, but of course you can still ask for it when creating
		4846	a loop.
		4847
		4848	=head3 C<select> is buggy
		4849
		4850	All that's left is C<select>, and of course Apple found a way to fuck this
		4851	one up as well: On OS/X, C<select> actively limits the number of file
		4852	descriptors you can pass in to 1024 - your program suddenly crashes when
		4853	you use more.
		4854
		4855	There is an undocumented "workaround" for this - defining
		4856	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4857	work on OS/X.
		4858
		4859	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4860
		4861	=head3 C<errno> reentrancy
		4862
		4863	The default compile environment on Solaris is unfortunately so
		4864	thread-unsafe that you can't even use components/libraries compiled
		4865	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		4866	defined by default. A valid, if stupid, implementation choice.
		4867
		4868	If you want to use libev in threaded environments you have to make sure
		4869	it's compiled with C<_REENTRANT> defined.
		4870
		4871	=head3 Event port backend
		4872
		4873	The scalable event interface for Solaris is called "event
		4874	ports". Unfortunately, this mechanism is very buggy in all major
		4875	releases. If you run into high CPU usage, your program freezes or you get
		4876	a large number of spurious wakeups, make sure you have all the relevant
		4877	and latest kernel patches applied. No, I don't know which ones, but there
		4878	are multiple ones to apply, and afterwards, event ports actually work
		4879	great.
		4880
		4881	If you can't get it to work, you can try running the program by setting
		4882	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4883	C<select> backends.
		4884
		4885	=head2 AIX POLL BUG
		4886
		4887	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4888	this by trying to avoid the poll backend altogether (i.e. it's not even
		4889	compiled in), which normally isn't a big problem as C<select> works fine
		4890	with large bitsets on AIX, and AIX is dead anyway.
		4891
3346	=head1 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS	4892	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4893
		4894	=head3 General issues
3347		4895
3348	Win32 doesn't support any of the standards (e.g. POSIX) that libev	4896	Win32 doesn't support any of the standards (e.g. POSIX) that libev
3349	requires, and its I/O model is fundamentally incompatible with the POSIX	4897	requires, and its I/O model is fundamentally incompatible with the POSIX
3350	model. Libev still offers limited functionality on this platform in	4898	model. Libev still offers limited functionality on this platform in
3351	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	4899	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
3352	descriptors. This only applies when using Win32 natively, not when using	4900	descriptors. This only applies when using Win32 natively, not when using
3353	e.g. cygwin.	4901	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4902	as every compielr comes with a slightly differently broken/incompatible
		4903	environment.
3354		4904
3355	Lifting these limitations would basically require the full	4905	Lifting these limitations would basically require the full
3356	re-implementation of the I/O system. If you are into these kinds of	4906	re-implementation of the I/O system. If you are into this kind of thing,
3357	things, then note that glib does exactly that for you in a very portable	4907	then note that glib does exactly that for you in a very portable way (note
3358	way (note also that glib is the slowest event library known to man).	4908	also that glib is the slowest event library known to man).
3359		4909
3360	There is no supported compilation method available on windows except	4910	There is no supported compilation method available on windows except
3361	embedding it into other applications.	4911	embedding it into other applications.
		4912
		4913	Sensible signal handling is officially unsupported by Microsoft - libev
		4914	tries its best, but under most conditions, signals will simply not work.
3362		4915
3363	Not a libev limitation but worth mentioning: windows apparently doesn't	4916	Not a libev limitation but worth mentioning: windows apparently doesn't
3364	accept large writes: instead of resulting in a partial write, windows will	4917	accept large writes: instead of resulting in a partial write, windows will
3365	either accept everything or return C<ENOBUFS> if the buffer is too large,	4918	either accept everything or return C<ENOBUFS> if the buffer is too large,
3366	so make sure you only write small amounts into your sockets (less than a	4919	so make sure you only write small amounts into your sockets (less than a
3367	megabyte seems safe, but thsi apparently depends on the amount of memory	4920	megabyte seems safe, but this apparently depends on the amount of memory
3368	available).	4921	available).
3369		4922
3370	Due to the many, low, and arbitrary limits on the win32 platform and	4923	Due to the many, low, and arbitrary limits on the win32 platform and
3371	the abysmal performance of winsockets, using a large number of sockets	4924	the abysmal performance of winsockets, using a large number of sockets
3372	is not recommended (and not reasonable). If your program needs to use	4925	is not recommended (and not reasonable). If your program needs to use
3373	more than a hundred or so sockets, then likely it needs to use a totally	4926	more than a hundred or so sockets, then likely it needs to use a totally
3374	different implementation for windows, as libev offers the POSIX readiness	4927	different implementation for windows, as libev offers the POSIX readiness
3375	notification model, which cannot be implemented efficiently on windows	4928	notification model, which cannot be implemented efficiently on windows
3376	(Microsoft monopoly games).	4929	(due to Microsoft monopoly games).
3377		4930
3378	A typical way to use libev under windows is to embed it (see the embedding	4931	A typical way to use libev under windows is to embed it (see the embedding
3379	section for details) and use the following F<evwrap.h> header file instead	4932	section for details) and use the following F<evwrap.h> header file instead
3380	of F<ev.h>:	4933	of F<ev.h>:
3381		4934
…		…
3383	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */	4936	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
3384		4937
3385	#include "ev.h"	4938	#include "ev.h"
3386		4939
3387	And compile the following F<evwrap.c> file into your project (make sure	4940	And compile the following F<evwrap.c> file into your project (make sure
3388	you do I<not> compile the F<ev.c> or any other embedded soruce files!):	4941	you do I<not> compile the F<ev.c> or any other embedded source files!):
3389		4942
3390	#include "evwrap.h"	4943	#include "evwrap.h"
3391	#include "ev.c"	4944	#include "ev.c"
3392		4945
3393	=over 4
3394
3395	=item The winsocket select function	4946	=head3 The winsocket C<select> function
3396		4947
3397	The winsocket C<select> function doesn't follow POSIX in that it	4948	The winsocket C<select> function doesn't follow POSIX in that it
3398	requires socket I<handles> and not socket I<file descriptors> (it is	4949	requires socket I<handles> and not socket I<file descriptors> (it is
3399	also extremely buggy). This makes select very inefficient, and also	4950	also extremely buggy). This makes select very inefficient, and also
3400	requires a mapping from file descriptors to socket handles (the Microsoft	4951	requires a mapping from file descriptors to socket handles (the Microsoft
…		…
3409	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */	4960	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
3410		4961
3411	Note that winsockets handling of fd sets is O(n), so you can easily get a	4962	Note that winsockets handling of fd sets is O(n), so you can easily get a
3412	complexity in the O(n²) range when using win32.	4963	complexity in the O(n²) range when using win32.
3413		4964
3414	=item Limited number of file descriptors	4965	=head3 Limited number of file descriptors
3415		4966
3416	Windows has numerous arbitrary (and low) limits on things.	4967	Windows has numerous arbitrary (and low) limits on things.
3417		4968
3418	Early versions of winsocket's select only supported waiting for a maximum	4969	Early versions of winsocket's select only supported waiting for a maximum
3419	of C<64> handles (probably owning to the fact that all windows kernels	4970	of C<64> handles (probably owning to the fact that all windows kernels
3420	can only wait for C<64> things at the same time internally; Microsoft	4971	can only wait for C<64> things at the same time internally; Microsoft
3421	recommends spawning a chain of threads and wait for 63 handles and the	4972	recommends spawning a chain of threads and wait for 63 handles and the
3422	previous thread in each. Great).	4973	previous thread in each. Sounds great!).
3423		4974
3424	Newer versions support more handles, but you need to define C<FD_SETSIZE>	4975	Newer versions support more handles, but you need to define C<FD_SETSIZE>
3425	to some high number (e.g. C<2048>) before compiling the winsocket select	4976	to some high number (e.g. C<2048>) before compiling the winsocket select
3426	call (which might be in libev or elsewhere, for example, perl does its own	4977	call (which might be in libev or elsewhere, for example, perl and many
3427	select emulation on windows).	4978	other interpreters do their own select emulation on windows).
3428		4979
3429	Another limit is the number of file descriptors in the Microsoft runtime	4980	Another limit is the number of file descriptors in the Microsoft runtime
3430	libraries, which by default is C<64> (there must be a hidden I<64> fetish	4981	libraries, which by default is C<64> (there must be a hidden I<64>
3431	or something like this inside Microsoft). You can increase this by calling	4982	fetish or something like this inside Microsoft). You can increase this
3432	C<_setmaxstdio>, which can increase this limit to C<2048> (another	4983	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
3433	arbitrary limit), but is broken in many versions of the Microsoft runtime	4984	(another arbitrary limit), but is broken in many versions of the Microsoft
3434	libraries.
3435
3436	This might get you to about C<512> or C<2048> sockets (depending on	4985	runtime libraries. This might get you to about C<512> or C<2048> sockets
3437	windows version and/or the phase of the moon). To get more, you need to	4986	(depending on windows version and/or the phase of the moon). To get more,
3438	wrap all I/O functions and provide your own fd management, but the cost of	4987	you need to wrap all I/O functions and provide your own fd management, but
3439	calling select (O(n²)) will likely make this unworkable.	4988	the cost of calling select (O(n²)) will likely make this unworkable.
3440		4989
3441	=back
3442
3443
3444	=head1 PORTABILITY REQUIREMENTS	4990	=head2 PORTABILITY REQUIREMENTS
3445		4991
3446	In addition to a working ISO-C implementation, libev relies on a few	4992	In addition to a working ISO-C implementation and of course the
3447	additional extensions:	4993	backend-specific APIs, libev relies on a few additional extensions:
3448		4994
3449	=over 4	4995	=over 4
3450		4996
3451	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible	4997	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
3452	calling conventions regardless of C<ev_watcher_type *>.	4998	calling conventions regardless of C<ev_watcher_type *>.
…		…
3455	structure (guaranteed by POSIX but not by ISO C for example), but it also	5001	structure (guaranteed by POSIX but not by ISO C for example), but it also
3456	assumes that the same (machine) code can be used to call any watcher	5002	assumes that the same (machine) code can be used to call any watcher
3457	callback: The watcher callbacks have different type signatures, but libev	5003	callback: The watcher callbacks have different type signatures, but libev
3458	calls them using an C<ev_watcher *> internally.	5004	calls them using an C<ev_watcher *> internally.
3459		5005
		5006	=item pointer accesses must be thread-atomic
		5007
		5008	Accessing a pointer value must be atomic, it must both be readable and
		5009	writable in one piece - this is the case on all current architectures.
		5010
3460	=item C<sig_atomic_t volatile> must be thread-atomic as well	5011	=item C<sig_atomic_t volatile> must be thread-atomic as well
3461		5012
3462	The type C<sig_atomic_t volatile> (or whatever is defined as	5013	The type C<sig_atomic_t volatile> (or whatever is defined as
3463	C<EV_ATOMIC_T>) must be atomic w.r.t. accesses from different	5014	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
3464	threads. This is not part of the specification for C<sig_atomic_t>, but is	5015	threads. This is not part of the specification for C<sig_atomic_t>, but is
3465	believed to be sufficiently portable.	5016	believed to be sufficiently portable.
3466		5017
3467	=item C<sigprocmask> must work in a threaded environment	5018	=item C<sigprocmask> must work in a threaded environment
3468		5019
…		…
3477	except the initial one, and run the default loop in the initial thread as	5028	except the initial one, and run the default loop in the initial thread as
3478	well.	5029	well.
3479		5030
3480	=item C<long> must be large enough for common memory allocation sizes	5031	=item C<long> must be large enough for common memory allocation sizes
3481		5032
3482	To improve portability and simplify using libev, libev uses C<long>	5033	To improve portability and simplify its API, libev uses C<long> internally
3483	internally instead of C<size_t> when allocating its data structures. On	5034	instead of C<size_t> when allocating its data structures. On non-POSIX
3484	non-POSIX systems (Microsoft...) this might be unexpectedly low, but	5035	systems (Microsoft...) this might be unexpectedly low, but is still at
3485	is still at least 31 bits everywhere, which is enough for hundreds of	5036	least 31 bits everywhere, which is enough for hundreds of millions of
3486	millions of watchers.	5037	watchers.
3487		5038
3488	=item C<double> must hold a time value in seconds with enough accuracy	5039	=item C<double> must hold a time value in seconds with enough accuracy
3489		5040
3490	The type C<double> is used to represent timestamps. It is required to	5041	The type C<double> is used to represent timestamps. It is required to
3491	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	5042	have at least 51 bits of mantissa (and 9 bits of exponent), which is
3492	enough for at least into the year 4000. This requirement is fulfilled by	5043	good enough for at least into the year 4000 with millisecond accuracy
		5044	(the design goal for libev). This requirement is overfulfilled by
3493	implementations implementing IEEE 754 (basically all existing ones).	5045	implementations using IEEE 754, which is basically all existing ones. With
		5046	IEEE 754 doubles, you get microsecond accuracy until at least 2200.
3494		5047
3495	=back	5048	=back
3496		5049
3497	If you know of other additional requirements drop me a note.	5050	If you know of other additional requirements drop me a note.
3498		5051
3499		5052
3500	=head1 COMPILER WARNINGS	5053	=head1 ALGORITHMIC COMPLEXITIES
3501		5054
3502	Depending on your compiler and compiler settings, you might get no or a	5055	In this section the complexities of (many of) the algorithms used inside
3503	lot of warnings when compiling libev code. Some people are apparently	5056	libev will be documented. For complexity discussions about backends see
3504	scared by this.	5057	the documentation for C<ev_default_init>.
3505		5058
3506	However, these are unavoidable for many reasons. For one, each compiler	5059	All of the following are about amortised time: If an array needs to be
3507	has different warnings, and each user has different tastes regarding	5060	extended, libev needs to realloc and move the whole array, but this
3508	warning options. "Warn-free" code therefore cannot be a goal except when	5061	happens asymptotically rarer with higher number of elements, so O(1) might
3509	targeting a specific compiler and compiler-version.	5062	mean that libev does a lengthy realloc operation in rare cases, but on
		5063	average it is much faster and asymptotically approaches constant time.
3510		5064
3511	Another reason is that some compiler warnings require elaborate	5065	=over 4
3512	workarounds, or other changes to the code that make it less clear and less
3513	maintainable.
3514		5066
3515	And of course, some compiler warnings are just plain stupid, or simply	5067	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
3516	wrong (because they don't actually warn about the condition their message
3517	seems to warn about).
3518		5068
3519	While libev is written to generate as few warnings as possible,	5069	This means that, when you have a watcher that triggers in one hour and
3520	"warn-free" code is not a goal, and it is recommended not to build libev	5070	there are 100 watchers that would trigger before that, then inserting will
3521	with any compiler warnings enabled unless you are prepared to cope with	5071	have to skip roughly seven (C<ld 100>) of these watchers.
3522	them (e.g. by ignoring them). Remember that warnings are just that:
3523	warnings, not errors, or proof of bugs.
3524		5072
		5073	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
3525		5074
3526	=head1 VALGRIND	5075	That means that changing a timer costs less than removing/adding them,
		5076	as only the relative motion in the event queue has to be paid for.
3527		5077
3528	Valgrind has a special section here because it is a popular tool that is	5078	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
3529	highly useful, but valgrind reports are very hard to interpret.
3530		5079
3531	If you think you found a bug (memory leak, uninitialised data access etc.)	5080	These just add the watcher into an array or at the head of a list.
3532	in libev, then check twice: If valgrind reports something like:
3533		5081
3534	==2274== definitely lost: 0 bytes in 0 blocks.	5082	=item Stopping check/prepare/idle/fork/async watchers: O(1)
3535	==2274== possibly lost: 0 bytes in 0 blocks.
3536	==2274== still reachable: 256 bytes in 1 blocks.
3537		5083
3538	Then there is no memory leak. Similarly, under some circumstances,	5084	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
3539	valgrind might report kernel bugs as if it were a bug in libev, or it
3540	might be confused (it is a very good tool, but only a tool).
3541		5085
3542	If you are unsure about something, feel free to contact the mailing list	5086	These watchers are stored in lists, so they need to be walked to find the
3543	with the full valgrind report and an explanation on why you think this is	5087	correct watcher to remove. The lists are usually short (you don't usually
3544	a bug in libev. However, don't be annoyed when you get a brisk "this is	5088	have many watchers waiting for the same fd or signal: one is typical, two
3545	no bug" answer and take the chance of learning how to interpret valgrind	5089	is rare).
3546	properly.
3547		5090
3548	If you need, for some reason, empty reports from valgrind for your project	5091	=item Finding the next timer in each loop iteration: O(1)
3549	I suggest using suppression lists.
3550		5092
		5093	By virtue of using a binary or 4-heap, the next timer is always found at a
		5094	fixed position in the storage array.
		5095
		5096	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5097
		5098	A change means an I/O watcher gets started or stopped, which requires
		5099	libev to recalculate its status (and possibly tell the kernel, depending
		5100	on backend and whether C<ev_io_set> was used).
		5101
		5102	=item Activating one watcher (putting it into the pending state): O(1)
		5103
		5104	=item Priority handling: O(number_of_priorities)
		5105
		5106	Priorities are implemented by allocating some space for each
		5107	priority. When doing priority-based operations, libev usually has to
		5108	linearly search all the priorities, but starting/stopping and activating
		5109	watchers becomes O(1) with respect to priority handling.
		5110
		5111	=item Sending an ev_async: O(1)
		5112
		5113	=item Processing ev_async_send: O(number_of_async_watchers)
		5114
		5115	=item Processing signals: O(max_signal_number)
		5116
		5117	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5118	calls in the current loop iteration. Checking for async and signal events
		5119	involves iterating over all running async watchers or all signal numbers.
		5120
		5121	=back
		5122
		5123
		5124	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5125
		5126	The major version 4 introduced some incompatible changes to the API.
		5127
		5128	At the moment, the C<ev.h> header file provides compatibility definitions
		5129	for all changes, so most programs should still compile. The compatibility
		5130	layer might be removed in later versions of libev, so better update to the
		5131	new API early than late.
		5132
		5133	=over 4
		5134
		5135	=item C<EV_COMPAT3> backwards compatibility mechanism
		5136
		5137	The backward compatibility mechanism can be controlled by
		5138	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>
		5139	section.
		5140
		5141	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5142
		5143	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5144
		5145	ev_loop_destroy (EV_DEFAULT_UC);
		5146	ev_loop_fork (EV_DEFAULT);
		5147
		5148	=item function/symbol renames
		5149
		5150	A number of functions and symbols have been renamed:
		5151
		5152	ev_loop => ev_run
		5153	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5154	EVLOOP_ONESHOT => EVRUN_ONCE
		5155
		5156	ev_unloop => ev_break
		5157	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5158	EVUNLOOP_ONE => EVBREAK_ONE
		5159	EVUNLOOP_ALL => EVBREAK_ALL
		5160
		5161	EV_TIMEOUT => EV_TIMER
		5162
		5163	ev_loop_count => ev_iteration
		5164	ev_loop_depth => ev_depth
		5165	ev_loop_verify => ev_verify
		5166
		5167	Most functions working on C<struct ev_loop> objects don't have an
		5168	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5169	associated constants have been renamed to not collide with the C<struct
		5170	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5171	as all other watcher types. Note that C<ev_loop_fork> is still called
		5172	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5173	typedef.
		5174
		5175	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5176
		5177	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5178	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5179	and work, but the library code will of course be larger.
		5180
		5181	=back
		5182
		5183
		5184	=head1 GLOSSARY
		5185
		5186	=over 4
		5187
		5188	=item active
		5189
		5190	A watcher is active as long as it has been started and not yet stopped.
		5191	See L<WATCHER STATES> for details.
		5192
		5193	=item application
		5194
		5195	In this document, an application is whatever is using libev.
		5196
		5197	=item backend
		5198
		5199	The part of the code dealing with the operating system interfaces.
		5200
		5201	=item callback
		5202
		5203	The address of a function that is called when some event has been
		5204	detected. Callbacks are being passed the event loop, the watcher that
		5205	received the event, and the actual event bitset.
		5206
		5207	=item callback/watcher invocation
		5208
		5209	The act of calling the callback associated with a watcher.
		5210
		5211	=item event
		5212
		5213	A change of state of some external event, such as data now being available
		5214	for reading on a file descriptor, time having passed or simply not having
		5215	any other events happening anymore.
		5216
		5217	In libev, events are represented as single bits (such as C<EV_READ> or
		5218	C<EV_TIMER>).
		5219
		5220	=item event library
		5221
		5222	A software package implementing an event model and loop.
		5223
		5224	=item event loop
		5225
		5226	An entity that handles and processes external events and converts them
		5227	into callback invocations.
		5228
		5229	=item event model
		5230
		5231	The model used to describe how an event loop handles and processes
		5232	watchers and events.
		5233
		5234	=item pending
		5235
		5236	A watcher is pending as soon as the corresponding event has been
		5237	detected. See L<WATCHER STATES> for details.
		5238
		5239	=item real time
		5240
		5241	The physical time that is observed. It is apparently strictly monotonic :)
		5242
		5243	=item wall-clock time
		5244
		5245	The time and date as shown on clocks. Unlike real time, it can actually
		5246	be wrong and jump forwards and backwards, e.g. when you adjust your
		5247	clock.
		5248
		5249	=item watcher
		5250
		5251	A data structure that describes interest in certain events. Watchers need
		5252	to be started (attached to an event loop) before they can receive events.
		5253
		5254	=back
3551		5255
3552	=head1 AUTHOR	5256	=head1 AUTHOR
3553		5257
3554	Marc Lehmann <libev@schmorp.de>.	5258	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5259	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
3555		5260

Diff Legend

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