ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.27 by root, Wed Nov 14 05:02:07 2007 UTC vs.
Revision 1.447 by root, Sat Jun 22 16:25:53 2019 UTC

		1	=encoding utf-8
		2
1	=head1 NAME	3	=head1 NAME
2		4
3	libev - a high performance full-featured event loop written in C	5	libev - a high performance full-featured event loop written in C
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
6		8
7	#include <ev.h>	9	#include <ev.h>
8		10
9	=head1 DESCRIPTION	11	=head2 EXAMPLE PROGRAM
		12
		13	// a single header file is required
		14	#include <ev.h>
		15
		16	#include <stdio.h> // for puts
		17
		18	// every watcher type has its own typedef'd struct
		19	// with the name ev_TYPE
		20	ev_io stdin_watcher;
		21	ev_timer timeout_watcher;
		22
		23	// all watcher callbacks have a similar signature
		24	// this callback is called when data is readable on stdin
		25	static void
		26	stdin_cb (EV_P_ ev_io *w, int revents)
		27	{
		28	puts ("stdin ready");
		29	// for one-shot events, one must manually stop the watcher
		30	// with its corresponding stop function.
		31	ev_io_stop (EV_A_ w);
		32
		33	// this causes all nested ev_run's to stop iterating
		34	ev_break (EV_A_ EVBREAK_ALL);
		35	}
		36
		37	// another callback, this time for a time-out
		38	static void
		39	timeout_cb (EV_P_ ev_timer *w, int revents)
		40	{
		41	puts ("timeout");
		42	// this causes the innermost ev_run to stop iterating
		43	ev_break (EV_A_ EVBREAK_ONE);
		44	}
		45
		46	int
		47	main (void)
		48	{
		49	// use the default event loop unless you have special needs
		50	struct ev_loop *loop = EV_DEFAULT;
		51
		52	// initialise an io watcher, then start it
		53	// this one will watch for stdin to become readable
		54	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		55	ev_io_start (loop, &stdin_watcher);
		56
		57	// initialise a timer watcher, then start it
		58	// simple non-repeating 5.5 second timeout
		59	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		60	ev_timer_start (loop, &timeout_watcher);
		61
		62	// now wait for events to arrive
		63	ev_run (loop, 0);
		64
		65	// break was called, so exit
		66	return 0;
		67	}
		68
		69	=head1 ABOUT THIS DOCUMENT
		70
		71	This document documents the libev software package.
		72
		73	The newest version of this document is also available as an html-formatted
		74	web page you might find easier to navigate when reading it for the first
		75	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		76
		77	While this document tries to be as complete as possible in documenting
		78	libev, its usage and the rationale behind its design, it is not a tutorial
		79	on event-based programming, nor will it introduce event-based programming
		80	with libev.
		81
		82	Familiarity with event based programming techniques in general is assumed
		83	throughout this document.
		84
		85	=head1 WHAT TO READ WHEN IN A HURRY
		86
		87	This manual tries to be very detailed, but unfortunately, this also makes
		88	it very long. If you just want to know the basics of libev, I suggest
		89	reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
		90	look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
		91	C<ev_timer> sections in L</WATCHER TYPES>.
		92
		93	=head1 ABOUT LIBEV
10		94
11	Libev is an event loop: you register interest in certain events (such as a	95	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	96	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	97	these event sources and provide your program with events.
14		98
15	To do this, it must take more or less complete control over your process	99	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	100	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	101	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	103	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	104	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	105	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	106	watcher.
23		107
24	=head1 FEATURES	108	=head2 FEATURES
25		109
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	110	Libev supports C<select>, C<poll>, the Linux-specific aio and C<epoll>
27	kqueue mechanisms for file descriptor events, relative timers, absolute	111	interfaces, the BSD-specific C<kqueue> and the Solaris-specific event port
28	timers with customised rescheduling, signal events, process status change	112	mechanisms for file descriptor events (C<ev_io>), the Linux C<inotify>
		113	interface (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		114	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		115	timers (C<ev_timer>), absolute timers with customised rescheduling
		116	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	117	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	118	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		119	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		120	limited support for fork events (C<ev_fork>).
		121
		122	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	123	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	124	for example).
33		125
34	=head1 CONVENTIONS	126	=head2 CONVENTIONS
35		127
36	Libev is very configurable. In this manual the default configuration	128	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	129	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	130	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	131	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	132	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	133	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	134	this argument.
43		135
44	=head1 TIME REPRESENTATION	136	=head2 TIME REPRESENTATION
45		137
46	Libev represents time as a single floating point number, representing the	138	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	139	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	140	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	141	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	142	too. It usually aliases to the C<double> type in C. When you need to do
		143	any calculations on it, you should treat it as some floating point value.
		144
		145	Unlike the name component C<stamp> might indicate, it is also used for
		146	time differences (e.g. delays) throughout libev.
		147
		148	=head1 ERROR HANDLING
		149
		150	Libev knows three classes of errors: operating system errors, usage errors
		151	and internal errors (bugs).
		152
		153	When libev catches an operating system error it cannot handle (for example
		154	a system call indicating a condition libev cannot fix), it calls the callback
		155	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		156	abort. The default is to print a diagnostic message and to call C<abort
		157	()>.
		158
		159	When libev detects a usage error such as a negative timer interval, then
		160	it will print a diagnostic message and abort (via the C<assert> mechanism,
		161	so C<NDEBUG> will disable this checking): these are programming errors in
		162	the libev caller and need to be fixed there.
		163
		164	Libev also has a few internal error-checking C<assert>ions, and also has
		165	extensive consistency checking code. These do not trigger under normal
		166	circumstances, as they indicate either a bug in libev or worse.
		167
51		168
52	=head1 GLOBAL FUNCTIONS	169	=head1 GLOBAL FUNCTIONS
53		170
54	These functions can be called anytime, even before initialising the	171	These functions can be called anytime, even before initialising the
55	library in any way.	172	library in any way.
…		…
58		175
59	=item ev_tstamp ev_time ()	176	=item ev_tstamp ev_time ()
60		177
61	Returns the current time as libev would use it. Please note that the	178	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	179	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	180	you actually want to know. Also interesting is the combination of
		181	C<ev_now_update> and C<ev_now>.
		182
		183	=item ev_sleep (ev_tstamp interval)
		184
		185	Sleep for the given interval: The current thread will be blocked
		186	until either it is interrupted or the given time interval has
		187	passed (approximately - it might return a bit earlier even if not
		188	interrupted). Returns immediately if C<< interval <= 0 >>.
		189
		190	Basically this is a sub-second-resolution C<sleep ()>.
		191
		192	The range of the C<interval> is limited - libev only guarantees to work
		193	with sleep times of up to one day (C<< interval <= 86400 >>).
64		194
65	=item int ev_version_major ()	195	=item int ev_version_major ()
66		196
67	=item int ev_version_minor ()	197	=item int ev_version_minor ()
68		198
69	You can find out the major and minor version numbers of the library	199	You can find out the major and minor ABI version numbers of the library
70	you linked against by calling the functions C<ev_version_major> and	200	you linked against by calling the functions C<ev_version_major> and
71	C<ev_version_minor>. If you want, you can compare against the global	201	C<ev_version_minor>. If you want, you can compare against the global
72	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	202	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
73	version of the library your program was compiled against.	203	version of the library your program was compiled against.
74		204
		205	These version numbers refer to the ABI version of the library, not the
		206	release version.
		207
75	Usually, it's a good idea to terminate if the major versions mismatch,	208	Usually, it's a good idea to terminate if the major versions mismatch,
76	as this indicates an incompatible change. Minor versions are usually	209	as this indicates an incompatible change. Minor versions are usually
77	compatible to older versions, so a larger minor version alone is usually	210	compatible to older versions, so a larger minor version alone is usually
78	not a problem.	211	not a problem.
79		212
		213	Example: Make sure we haven't accidentally been linked against the wrong
		214	version (note, however, that this will not detect other ABI mismatches,
		215	such as LFS or reentrancy).
		216
		217	assert (("libev version mismatch",
		218	ev_version_major () == EV_VERSION_MAJOR
		219	&& ev_version_minor () >= EV_VERSION_MINOR));
		220
		221	=item unsigned int ev_supported_backends ()
		222
		223	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		224	value) compiled into this binary of libev (independent of their
		225	availability on the system you are running on). See C<ev_default_loop> for
		226	a description of the set values.
		227
		228	Example: make sure we have the epoll method, because yeah this is cool and
		229	a must have and can we have a torrent of it please!!!11
		230
		231	assert (("sorry, no epoll, no sex",
		232	ev_supported_backends () & EVBACKEND_EPOLL));
		233
		234	=item unsigned int ev_recommended_backends ()
		235
		236	Return the set of all backends compiled into this binary of libev and
		237	also recommended for this platform, meaning it will work for most file
		238	descriptor types. This set is often smaller than the one returned by
		239	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		240	and will not be auto-detected unless you explicitly request it (assuming
		241	you know what you are doing). This is the set of backends that libev will
		242	probe for if you specify no backends explicitly.
		243
		244	=item unsigned int ev_embeddable_backends ()
		245
		246	Returns the set of backends that are embeddable in other event loops. This
		247	value is platform-specific but can include backends not available on the
		248	current system. To find which embeddable backends might be supported on
		249	the current system, you would need to look at C<ev_embeddable_backends ()
		250	& ev_supported_backends ()>, likewise for recommended ones.
		251
		252	See the description of C<ev_embed> watchers for more info.
		253
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	254	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
81		255
82	Sets the allocation function to use (the prototype is similar to the	256	Sets the allocation function to use (the prototype is similar - the
83	realloc C function, the semantics are identical). It is used to allocate	257	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
84	and free memory (no surprises here). If it returns zero when memory	258	used to allocate and free memory (no surprises here). If it returns zero
85	needs to be allocated, the library might abort or take some potentially	259	when memory needs to be allocated (C<size != 0>), the library might abort
86	destructive action. The default is your system realloc function.	260	or take some potentially destructive action.
		261
		262	Since some systems (at least OpenBSD and Darwin) fail to implement
		263	correct C<realloc> semantics, libev will use a wrapper around the system
		264	C<realloc> and C<free> functions by default.
87		265
88	You could override this function in high-availability programs to, say,	266	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	267	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	268	or even to sleep a while and retry until some memory is available.
91		269
		270	Example: The following is the C<realloc> function that libev itself uses
		271	which should work with C<realloc> and C<free> functions of all kinds and
		272	is probably a good basis for your own implementation.
		273
		274	static void *
		275	ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
		276	{
		277	if (size)
		278	return realloc (ptr, size);
		279
		280	free (ptr);
		281	return 0;
		282	}
		283
		284	Example: Replace the libev allocator with one that waits a bit and then
		285	retries.
		286
		287	static void *
		288	persistent_realloc (void *ptr, size_t size)
		289	{
		290	if (!size)
		291	{
		292	free (ptr);
		293	return 0;
		294	}
		295
		296	for (;;)
		297	{
		298	void *newptr = realloc (ptr, size);
		299
		300	if (newptr)
		301	return newptr;
		302
		303	sleep (60);
		304	}
		305	}
		306
		307	...
		308	ev_set_allocator (persistent_realloc);
		309
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	310	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
93		311
94	Set the callback function to call on a retryable syscall error (such	312	Set the callback function to call on a retryable system call error (such
95	as failed select, poll, epoll_wait). The message is a printable string	313	as failed select, poll, epoll_wait). The message is a printable string
96	indicating the system call or subsystem causing the problem. If this	314	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	315	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	316	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	317	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	318	(such as abort).
101		319
		320	Example: This is basically the same thing that libev does internally, too.
		321
		322	static void
		323	fatal_error (const char *msg)
		324	{
		325	perror (msg);
		326	abort ();
		327	}
		328
		329	...
		330	ev_set_syserr_cb (fatal_error);
		331
		332	=item ev_feed_signal (int signum)
		333
		334	This function can be used to "simulate" a signal receive. It is completely
		335	safe to call this function at any time, from any context, including signal
		336	handlers or random threads.
		337
		338	Its main use is to customise signal handling in your process, especially
		339	in the presence of threads. For example, you could block signals
		340	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		341	creating any loops), and in one thread, use C<sigwait> or any other
		342	mechanism to wait for signals, then "deliver" them to libev by calling
		343	C<ev_feed_signal>.
		344
102	=back	345	=back
103		346
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	347	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
105		348
106	An event loop is described by a C<struct ev_loop *>. The library knows two	349	An event loop is described by a C<struct ev_loop *> (the C<struct> is
107	types of such loops, the I<default> loop, which supports signals and child	350	I<not> optional in this case unless libev 3 compatibility is disabled, as
108	events, and dynamically created loops which do not.	351	libev 3 had an C<ev_loop> function colliding with the struct name).
109		352
110	If you use threads, a common model is to run the default event loop	353	The library knows two types of such loops, the I<default> loop, which
111	in your main thread (or in a separate thread) and for each thread you	354	supports child process events, and dynamically created event loops which
112	create, you also create another event loop. Libev itself does no locking	355	do not.
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		356
117	=over 4	357	=over 4
118		358
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	359	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		360
121	This will initialise the default event loop if it hasn't been initialised	361	This returns the "default" event loop object, which is what you should
122	yet and return it. If the default loop could not be initialised, returns	362	normally use when you just need "the event loop". Event loop objects and
123	false. If it already was initialised it simply returns it (and ignores the	363	the C<flags> parameter are described in more detail in the entry for
124	flags).	364	C<ev_loop_new>.
		365
		366	If the default loop is already initialised then this function simply
		367	returns it (and ignores the flags. If that is troubling you, check
		368	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		369	flags, which should almost always be C<0>, unless the caller is also the
		370	one calling C<ev_run> or otherwise qualifies as "the main program".
125		371
126	If you don't know what event loop to use, use the one returned from this	372	If you don't know what event loop to use, use the one returned from this
127	function.	373	function (or via the C<EV_DEFAULT> macro).
		374
		375	Note that this function is I<not> thread-safe, so if you want to use it
		376	from multiple threads, you have to employ some kind of mutex (note also
		377	that this case is unlikely, as loops cannot be shared easily between
		378	threads anyway).
		379
		380	The default loop is the only loop that can handle C<ev_child> watchers,
		381	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		382	a problem for your application you can either create a dynamic loop with
		383	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		384	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		385
		386	Example: This is the most typical usage.
		387
		388	if (!ev_default_loop (0))
		389	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		390
		391	Example: Restrict libev to the select and poll backends, and do not allow
		392	environment settings to be taken into account:
		393
		394	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		395
		396	=item struct ev_loop *ev_loop_new (unsigned int flags)
		397
		398	This will create and initialise a new event loop object. If the loop
		399	could not be initialised, returns false.
		400
		401	This function is thread-safe, and one common way to use libev with
		402	threads is indeed to create one loop per thread, and using the default
		403	loop in the "main" or "initial" thread.
128		404
129	The flags argument can be used to specify special behaviour or specific	405	The flags argument can be used to specify special behaviour or specific
130	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	406	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
131		407
132	It supports the following flags:	408	The following flags are supported:
133		409
134	=over 4	410	=over 4
135		411
136	=item C<EVFLAG_AUTO>	412	=item C<EVFLAG_AUTO>
137		413
138	The default flags value. Use this if you have no clue (it's the right	414	The default flags value. Use this if you have no clue (it's the right
139	thing, believe me).	415	thing, believe me).
140		416
141	=item C<EVFLAG_NOENV>	417	=item C<EVFLAG_NOENV>
142		418
143	If this flag bit is ored into the flag value (or the program runs setuid	419	If this flag bit is or'ed into the flag value (or the program runs setuid
144	or setgid) then libev will I<not> look at the environment variable	420	or setgid) then libev will I<not> look at the environment variable
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	421	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	override the flags completely if it is found in the environment. This is	422	override the flags completely if it is found in the environment. This is
147	useful to try out specific backends to test their performance, or to work	423	useful to try out specific backends to test their performance, to work
148	around bugs.	424	around bugs, or to make libev threadsafe (accessing environment variables
		425	cannot be done in a threadsafe way, but usually it works if no other
		426	thread modifies them).
149		427
		428	=item C<EVFLAG_FORKCHECK>
		429
		430	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		431	make libev check for a fork in each iteration by enabling this flag.
		432
		433	This works by calling C<getpid ()> on every iteration of the loop,
		434	and thus this might slow down your event loop if you do a lot of loop
		435	iterations and little real work, but is usually not noticeable (on my
		436	GNU/Linux system for example, C<getpid> is actually a simple 5-insn
		437	sequence without a system call and thus I<very> fast, but my GNU/Linux
		438	system also has C<pthread_atfork> which is even faster). (Update: glibc
		439	versions 2.25 apparently removed the C<getpid> optimisation again).
		440
		441	The big advantage of this flag is that you can forget about fork (and
		442	forget about forgetting to tell libev about forking, although you still
		443	have to ignore C<SIGPIPE>) when you use this flag.
		444
		445	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		446	environment variable.
		447
		448	=item C<EVFLAG_NOINOTIFY>
		449
		450	When this flag is specified, then libev will not attempt to use the
		451	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		452	testing, this flag can be useful to conserve inotify file descriptors, as
		453	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		454
		455	=item C<EVFLAG_SIGNALFD>
		456
		457	When this flag is specified, then libev will attempt to use the
		458	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		459	delivers signals synchronously, which makes it both faster and might make
		460	it possible to get the queued signal data. It can also simplify signal
		461	handling with threads, as long as you properly block signals in your
		462	threads that are not interested in handling them.
		463
		464	Signalfd will not be used by default as this changes your signal mask, and
		465	there are a lot of shoddy libraries and programs (glib's threadpool for
		466	example) that can't properly initialise their signal masks.
		467
		468	=item C<EVFLAG_NOSIGMASK>
		469
		470	When this flag is specified, then libev will avoid to modify the signal
		471	mask. Specifically, this means you have to make sure signals are unblocked
		472	when you want to receive them.
		473
		474	This behaviour is useful when you want to do your own signal handling, or
		475	want to handle signals only in specific threads and want to avoid libev
		476	unblocking the signals.
		477
		478	It's also required by POSIX in a threaded program, as libev calls
		479	C<sigprocmask>, whose behaviour is officially unspecified.
		480
		481	This flag's behaviour will become the default in future versions of libev.
		482
150	=item C<EVMETHOD_SELECT> (portable select backend)	483	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		484
		485	This is your standard select(2) backend. Not I<completely> standard, as
		486	libev tries to roll its own fd_set with no limits on the number of fds,
		487	but if that fails, expect a fairly low limit on the number of fds when
		488	using this backend. It doesn't scale too well (O(highest_fd)), but its
		489	usually the fastest backend for a low number of (low-numbered :) fds.
		490
		491	To get good performance out of this backend you need a high amount of
		492	parallelism (most of the file descriptors should be busy). If you are
		493	writing a server, you should C<accept ()> in a loop to accept as many
		494	connections as possible during one iteration. You might also want to have
		495	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		496	readiness notifications you get per iteration.
		497
		498	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		499	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		500	C<exceptfds> set on that platform).
		501
152	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	502	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
153		503
154	=item C<EVMETHOD_EPOLL> (linux only)	504	And this is your standard poll(2) backend. It's more complicated
		505	than select, but handles sparse fds better and has no artificial
		506	limit on the number of fds you can use (except it will slow down
		507	considerably with a lot of inactive fds). It scales similarly to select,
		508	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		509	performance tips.
155		510
156	=item C<EVMETHOD_KQUEUE> (some bsds only)	511	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		512	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157		513
158	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	514	=item C<EVBACKEND_EPOLL> (value 4, Linux)
159		515
160	=item C<EVMETHOD_PORT> (solaris 10 only)	516	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		517	kernels).
161		518
162	If one or more of these are ored into the flags value, then only these	519	For few fds, this backend is a bit little slower than poll and select, but
163	backends will be tried (in the reverse order as given here). If one are	520	it scales phenomenally better. While poll and select usually scale like
164	specified, any backend will do.	521	O(total_fds) where total_fds is the total number of fds (or the highest
		522	fd), epoll scales either O(1) or O(active_fds).
		523
		524	The epoll mechanism deserves honorable mention as the most misdesigned
		525	of the more advanced event mechanisms: mere annoyances include silently
		526	dropping file descriptors, requiring a system call per change per file
		527	descriptor (and unnecessary guessing of parameters), problems with dup,
		528	returning before the timeout value, resulting in additional iterations
		529	(and only giving 5ms accuracy while select on the same platform gives
		530	0.1ms) and so on. The biggest issue is fork races, however - if a program
		531	forks then I<both> parent and child process have to recreate the epoll
		532	set, which can take considerable time (one syscall per file descriptor)
		533	and is of course hard to detect.
		534
		535	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
		536	but of course I<doesn't>, and epoll just loves to report events for
		537	totally I<different> file descriptors (even already closed ones, so
		538	one cannot even remove them from the set) than registered in the set
		539	(especially on SMP systems). Libev tries to counter these spurious
		540	notifications by employing an additional generation counter and comparing
		541	that against the events to filter out spurious ones, recreating the set
		542	when required. Epoll also erroneously rounds down timeouts, but gives you
		543	no way to know when and by how much, so sometimes you have to busy-wait
		544	because epoll returns immediately despite a nonzero timeout. And last
		545	not least, it also refuses to work with some file descriptors which work
		546	perfectly fine with C<select> (files, many character devices...).
		547
		548	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		549	cobbled together in a hurry, no thought to design or interaction with
		550	others. Oh, the pain, will it ever stop...
		551
		552	While stopping, setting and starting an I/O watcher in the same iteration
		553	will result in some caching, there is still a system call per such
		554	incident (because the same I<file descriptor> could point to a different
		555	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		556	file descriptors might not work very well if you register events for both
		557	file descriptors.
		558
		559	Best performance from this backend is achieved by not unregistering all
		560	watchers for a file descriptor until it has been closed, if possible,
		561	i.e. keep at least one watcher active per fd at all times. Stopping and
		562	starting a watcher (without re-setting it) also usually doesn't cause
		563	extra overhead. A fork can both result in spurious notifications as well
		564	as in libev having to destroy and recreate the epoll object, which can
		565	take considerable time and thus should be avoided.
		566
		567	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		568	faster than epoll for maybe up to a hundred file descriptors, depending on
		569	the usage. So sad.
		570
		571	While nominally embeddable in other event loops, this feature is broken in
		572	a lot of kernel revisions, but probably(!) works in current versions.
		573
		574	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		575	C<EVBACKEND_POLL>.
		576
		577	=item C<EVBACKEND_LINUXAIO> (value 64, Linux)
		578
		579	Use the linux-specific linux aio (I<not> C<< aio(7) >>) event interface
		580	available in post-4.18 kernels.
		581
		582	If this backend works for you (as of this writing, it was very
		583	experimental and only supports a subset of file types), it is the best
		584	event interface available on linux and might be well worth it enabling it
		585	- if it isn't available in your kernel this will be detected and another
		586	backend will be chosen.
		587
		588	This backend can batch oneshot requests and uses a user-space ring buffer
		589	to receive events. It also doesn't suffer from most of the design problems
		590	of epoll (such as not being able to remove event sources from the epoll
		591	set), and generally sounds too good to be true. Because, this being the
		592	linux kernel, of course it suffers from a whole new set of limitations.
		593
		594	For one, it is not easily embeddable (but probably could be done using
		595	an event fd at some extra overhead). It also is subject to various
		596	arbitrary limits that can be configured in F</proc/sys/fs/aio-max-nr>
		597	and F</proc/sys/fs/aio-nr>), which could lead to it being skipped during
		598	initialisation.
		599
		600	Most problematic in practise, however, is that, like kqueue, it requires
		601	special support from drivers, and, not surprisingly, not all drivers
		602	implement it. For example, in linux 4.19, tcp sockets, pipes, event fds,
		603	files, F</dev/null> and a few others are supported, but ttys are not, so
		604	this is not (yet?) a generic event polling interface but is probably still
		605	be very useful in a web server or similar program.
		606
		607	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		608	C<EVBACKEND_POLL>.
		609
		610	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		611
		612	Kqueue deserves special mention, as at the time of this writing, it
		613	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		614	with anything but sockets and pipes, except on Darwin, where of course
		615	it's completely useless). Unlike epoll, however, whose brokenness
		616	is by design, these kqueue bugs can (and eventually will) be fixed
		617	without API changes to existing programs. For this reason it's not being
		618	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		619	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		620	system like NetBSD.
		621
		622	You still can embed kqueue into a normal poll or select backend and use it
		623	only for sockets (after having made sure that sockets work with kqueue on
		624	the target platform). See C<ev_embed> watchers for more info.
		625
		626	It scales in the same way as the epoll backend, but the interface to the
		627	kernel is more efficient (which says nothing about its actual speed, of
		628	course). While stopping, setting and starting an I/O watcher does never
		629	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		630	two event changes per incident. Support for C<fork ()> is very bad (you
		631	might have to leak fd's on fork, but it's more sane than epoll) and it
		632	drops fds silently in similarly hard-to-detect cases.
		633
		634	This backend usually performs well under most conditions.
		635
		636	While nominally embeddable in other event loops, this doesn't work
		637	everywhere, so you might need to test for this. And since it is broken
		638	almost everywhere, you should only use it when you have a lot of sockets
		639	(for which it usually works), by embedding it into another event loop
		640	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		641	also broken on OS X)) and, did I mention it, using it only for sockets.
		642
		643	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		644	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		645	C<NOTE_EOF>.
		646
		647	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		648
		649	This is not implemented yet (and might never be, unless you send me an
		650	implementation). According to reports, C</dev/poll> only supports sockets
		651	and is not embeddable, which would limit the usefulness of this backend
		652	immensely.
		653
		654	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		655
		656	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		657	it's really slow, but it still scales very well (O(active_fds)).
		658
		659	While this backend scales well, it requires one system call per active
		660	file descriptor per loop iteration. For small and medium numbers of file
		661	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		662	might perform better.
		663
		664	On the positive side, this backend actually performed fully to
		665	specification in all tests and is fully embeddable, which is a rare feat
		666	among the OS-specific backends (I vastly prefer correctness over speed
		667	hacks).
		668
		669	On the negative side, the interface is I<bizarre> - so bizarre that
		670	even sun itself gets it wrong in their code examples: The event polling
		671	function sometimes returns events to the caller even though an error
		672	occurred, but with no indication whether it has done so or not (yes, it's
		673	even documented that way) - deadly for edge-triggered interfaces where you
		674	absolutely have to know whether an event occurred or not because you have
		675	to re-arm the watcher.
		676
		677	Fortunately libev seems to be able to work around these idiocies.
		678
		679	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		680	C<EVBACKEND_POLL>.
		681
		682	=item C<EVBACKEND_ALL>
		683
		684	Try all backends (even potentially broken ones that wouldn't be tried
		685	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		686	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		687
		688	It is definitely not recommended to use this flag, use whatever
		689	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		690	at all.
		691
		692	=item C<EVBACKEND_MASK>
		693
		694	Not a backend at all, but a mask to select all backend bits from a
		695	C<flags> value, in case you want to mask out any backends from a flags
		696	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
165		697
166	=back	698	=back
167		699
168	=item struct ev_loop *ev_loop_new (unsigned int flags)	700	If one or more of the backend flags are or'ed into the flags value,
		701	then only these backends will be tried (in the reverse order as listed
		702	here). If none are specified, all backends in C<ev_recommended_backends
		703	()> will be tried.
169		704
170	Similar to C<ev_default_loop>, but always creates a new event loop that is	705	Example: Try to create a event loop that uses epoll and nothing else.
171	always distinct from the default loop. Unlike the default loop, it cannot
172	handle signal and child watchers, and attempts to do so will be greeted by
173	undefined behaviour (or a failed assertion if assertions are enabled).
174		706
175	=item ev_default_destroy ()	707	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		708	if (!epoller)
		709	fatal ("no epoll found here, maybe it hides under your chair");
176		710
177	Destroys the default loop again (frees all memory and kernel state	711	Example: Use whatever libev has to offer, but make sure that kqueue is
178	etc.). This stops all registered event watchers (by not touching them in	712	used if available.
179	any way whatsoever, although you cannot rely on this :).	713
		714	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		715
		716	Example: Similarly, on linux, you mgiht want to take advantage of the
		717	linux aio backend if possible, but fall back to something else if that
		718	isn't available.
		719
		720	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
180		721
181	=item ev_loop_destroy (loop)	722	=item ev_loop_destroy (loop)
182		723
183	Like C<ev_default_destroy>, but destroys an event loop created by an	724	Destroys an event loop object (frees all memory and kernel state
184	earlier call to C<ev_loop_new>.	725	etc.). None of the active event watchers will be stopped in the normal
		726	sense, so e.g. C<ev_is_active> might still return true. It is your
		727	responsibility to either stop all watchers cleanly yourself I<before>
		728	calling this function, or cope with the fact afterwards (which is usually
		729	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		730	for example).
185		731
186	=item ev_default_fork ()	732	Note that certain global state, such as signal state (and installed signal
		733	handlers), will not be freed by this function, and related watchers (such
		734	as signal and child watchers) would need to be stopped manually.
187		735
		736	This function is normally used on loop objects allocated by
		737	C<ev_loop_new>, but it can also be used on the default loop returned by
		738	C<ev_default_loop>, in which case it is not thread-safe.
		739
		740	Note that it is not advisable to call this function on the default loop
		741	except in the rare occasion where you really need to free its resources.
		742	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		743	and C<ev_loop_destroy>.
		744
		745	=item ev_loop_fork (loop)
		746
		747	This function sets a flag that causes subsequent C<ev_run> iterations
188	This function reinitialises the kernel state for backends that have	748	to reinitialise the kernel state for backends that have one. Despite
189	one. Despite the name, you can call it anytime, but it makes most sense	749	the name, you can call it anytime you are allowed to start or stop
190	after forking, in either the parent or child process (or both, but that	750	watchers (except inside an C<ev_prepare> callback), but it makes most
191	again makes little sense).	751	sense after forking, in the child process. You I<must> call it (or use
		752	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
192		753
193	You I<must> call this function after forking if and only if you want to	754	In addition, if you want to reuse a loop (via this function or
194	use the event library in both processes. If you just fork+exec, you don't	755	C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
195	have to call it.	756
		757	Again, you I<have> to call it on I<any> loop that you want to re-use after
		758	a fork, I<even if you do not plan to use the loop in the parent>. This is
		759	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		760	during fork.
		761
		762	On the other hand, you only need to call this function in the child
		763	process if and only if you want to use the event loop in the child. If
		764	you just fork+exec or create a new loop in the child, you don't have to
		765	call it at all (in fact, C<epoll> is so badly broken that it makes a
		766	difference, but libev will usually detect this case on its own and do a
		767	costly reset of the backend).
196		768
197	The function itself is quite fast and it's usually not a problem to call	769	The function itself is quite fast and it's usually not a problem to call
198	it just in case after a fork. To make this easy, the function will fit in	770	it just in case after a fork.
199	quite nicely into a call to C<pthread_atfork>:
200		771
		772	Example: Automate calling C<ev_loop_fork> on the default loop when
		773	using pthreads.
		774
		775	static void
		776	post_fork_child (void)
		777	{
		778	ev_loop_fork (EV_DEFAULT);
		779	}
		780
		781	...
201	pthread_atfork (0, 0, ev_default_fork);	782	pthread_atfork (0, 0, post_fork_child);
202		783
203	=item ev_loop_fork (loop)	784	=item int ev_is_default_loop (loop)
204		785
205	Like C<ev_default_fork>, but acts on an event loop created by	786	Returns true when the given loop is, in fact, the default loop, and false
206	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	787	otherwise.
207	after fork, and how you do this is entirely your own problem.
208		788
		789	=item unsigned int ev_iteration (loop)
		790
		791	Returns the current iteration count for the event loop, which is identical
		792	to the number of times libev did poll for new events. It starts at C<0>
		793	and happily wraps around with enough iterations.
		794
		795	This value can sometimes be useful as a generation counter of sorts (it
		796	"ticks" the number of loop iterations), as it roughly corresponds with
		797	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		798	prepare and check phases.
		799
209	=item unsigned int ev_method (loop)	800	=item unsigned int ev_depth (loop)
210		801
		802	Returns the number of times C<ev_run> was entered minus the number of
		803	times C<ev_run> was exited normally, in other words, the recursion depth.
		804
		805	Outside C<ev_run>, this number is zero. In a callback, this number is
		806	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		807	in which case it is higher.
		808
		809	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		810	throwing an exception etc.), doesn't count as "exit" - consider this
		811	as a hint to avoid such ungentleman-like behaviour unless it's really
		812	convenient, in which case it is fully supported.
		813
		814	=item unsigned int ev_backend (loop)
		815
211	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	816	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
212	use.	817	use.
213		818
214	=item ev_tstamp ev_now (loop)	819	=item ev_tstamp ev_now (loop)
215		820
216	Returns the current "event loop time", which is the time the event loop	821	Returns the current "event loop time", which is the time the event loop
217	got events and started processing them. This timestamp does not change	822	received events and started processing them. This timestamp does not
218	as long as callbacks are being processed, and this is also the base time	823	change as long as callbacks are being processed, and this is also the base
219	used for relative timers. You can treat it as the timestamp of the event	824	time used for relative timers. You can treat it as the timestamp of the
220	occuring (or more correctly, the mainloop finding out about it).	825	event occurring (or more correctly, libev finding out about it).
221		826
		827	=item ev_now_update (loop)
		828
		829	Establishes the current time by querying the kernel, updating the time
		830	returned by C<ev_now ()> in the progress. This is a costly operation and
		831	is usually done automatically within C<ev_run ()>.
		832
		833	This function is rarely useful, but when some event callback runs for a
		834	very long time without entering the event loop, updating libev's idea of
		835	the current time is a good idea.
		836
		837	See also L</The special problem of time updates> in the C<ev_timer> section.
		838
		839	=item ev_suspend (loop)
		840
		841	=item ev_resume (loop)
		842
		843	These two functions suspend and resume an event loop, for use when the
		844	loop is not used for a while and timeouts should not be processed.
		845
		846	A typical use case would be an interactive program such as a game: When
		847	the user presses C<^Z> to suspend the game and resumes it an hour later it
		848	would be best to handle timeouts as if no time had actually passed while
		849	the program was suspended. This can be achieved by calling C<ev_suspend>
		850	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		851	C<ev_resume> directly afterwards to resume timer processing.
		852
		853	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		854	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		855	will be rescheduled (that is, they will lose any events that would have
		856	occurred while suspended).
		857
		858	After calling C<ev_suspend> you B<must not> call I<any> function on the
		859	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		860	without a previous call to C<ev_suspend>.
		861
		862	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		863	event loop time (see C<ev_now_update>).
		864
222	=item ev_loop (loop, int flags)	865	=item bool ev_run (loop, int flags)
223		866
224	Finally, this is it, the event handler. This function usually is called	867	Finally, this is it, the event handler. This function usually is called
225	after you initialised all your watchers and you want to start handling	868	after you have initialised all your watchers and you want to start
226	events.	869	handling events. It will ask the operating system for any new events, call
		870	the watcher callbacks, and then repeat the whole process indefinitely: This
		871	is why event loops are called I<loops>.
227		872
228	If the flags argument is specified as 0, it will not return until either	873	If the flags argument is specified as C<0>, it will keep handling events
229	no event watchers are active anymore or C<ev_unloop> was called.	874	until either no event watchers are active anymore or C<ev_break> was
		875	called.
230		876
		877	The return value is false if there are no more active watchers (which
		878	usually means "all jobs done" or "deadlock"), and true in all other cases
		879	(which usually means " you should call C<ev_run> again").
		880
		881	Please note that an explicit C<ev_break> is usually better than
		882	relying on all watchers to be stopped when deciding when a program has
		883	finished (especially in interactive programs), but having a program
		884	that automatically loops as long as it has to and no longer by virtue
		885	of relying on its watchers stopping correctly, that is truly a thing of
		886	beauty.
		887
		888	This function is I<mostly> exception-safe - you can break out of a
		889	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		890	exception and so on. This does not decrement the C<ev_depth> value, nor
		891	will it clear any outstanding C<EVBREAK_ONE> breaks.
		892
231	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	893	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
232	those events and any outstanding ones, but will not block your process in	894	those events and any already outstanding ones, but will not wait and
233	case there are no events and will return after one iteration of the loop.	895	block your process in case there are no events and will return after one
		896	iteration of the loop. This is sometimes useful to poll and handle new
		897	events while doing lengthy calculations, to keep the program responsive.
234		898
235	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	899	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
236	neccessary) and will handle those and any outstanding ones. It will block	900	necessary) and will handle those and any already outstanding ones. It
237	your process until at least one new event arrives, and will return after	901	will block your process until at least one new event arrives (which could
		902	be an event internal to libev itself, so there is no guarantee that a
		903	user-registered callback will be called), and will return after one
238	one iteration of the loop.	904	iteration of the loop.
239		905
240	This flags value could be used to implement alternative looping	906	This is useful if you are waiting for some external event in conjunction
241	constructs, but the C<prepare> and C<check> watchers provide a better and	907	with something not expressible using other libev watchers (i.e. "roll your
242	more generic mechanism.	908	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		909	usually a better approach for this kind of thing.
243		910
244	Here are the gory details of what ev_loop does:	911	Here are the gory details of what C<ev_run> does (this is for your
		912	understanding, not a guarantee that things will work exactly like this in
		913	future versions):
245		914
246	1. If there are no active watchers (reference count is zero), return.	915	- Increment loop depth.
		916	- Reset the ev_break status.
		917	- Before the first iteration, call any pending watchers.
		918	LOOP:
		919	- If EVFLAG_FORKCHECK was used, check for a fork.
		920	- If a fork was detected (by any means), queue and call all fork watchers.
247	2. Queue and immediately call all prepare watchers.	921	- Queue and call all prepare watchers.
		922	- If ev_break was called, goto FINISH.
248	3. If we have been forked, recreate the kernel state.	923	- If we have been forked, detach and recreate the kernel state
		924	as to not disturb the other process.
249	4. Update the kernel state with all outstanding changes.	925	- Update the kernel state with all outstanding changes.
250	5. Update the "event loop time".	926	- Update the "event loop time" (ev_now ()).
251	6. Calculate for how long to block.	927	- Calculate for how long to sleep or block, if at all
		928	(active idle watchers, EVRUN_NOWAIT or not having
		929	any active watchers at all will result in not sleeping).
		930	- Sleep if the I/O and timer collect interval say so.
		931	- Increment loop iteration counter.
252	7. Block the process, waiting for events.	932	- Block the process, waiting for any events.
		933	- Queue all outstanding I/O (fd) events.
253	8. Update the "event loop time" and do time jump handling.	934	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
254	9. Queue all outstanding timers.	935	- Queue all expired timers.
255	10. Queue all outstanding periodics.	936	- Queue all expired periodics.
256	11. If no events are pending now, queue all idle watchers.	937	- Queue all idle watchers with priority higher than that of pending events.
257	12. Queue all check watchers.	938	- Queue all check watchers.
258	13. Call all queued watchers in reverse order (i.e. check watchers first).	939	- Call all queued watchers in reverse order (i.e. check watchers first).
259	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	940	Signals and child watchers are implemented as I/O watchers, and will
260	was used, return, otherwise continue with step #1.	941	be handled here by queueing them when their watcher gets executed.
		942	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		943	were used, or there are no active watchers, goto FINISH, otherwise
		944	continue with step LOOP.
		945	FINISH:
		946	- Reset the ev_break status iff it was EVBREAK_ONE.
		947	- Decrement the loop depth.
		948	- Return.
261		949
		950	Example: Queue some jobs and then loop until no events are outstanding
		951	anymore.
		952
		953	... queue jobs here, make sure they register event watchers as long
		954	... as they still have work to do (even an idle watcher will do..)
		955	ev_run (my_loop, 0);
		956	... jobs done or somebody called break. yeah!
		957
262	=item ev_unloop (loop, how)	958	=item ev_break (loop, how)
263		959
264	Can be used to make a call to C<ev_loop> return early (but only after it	960	Can be used to make a call to C<ev_run> return early (but only after it
265	has processed all outstanding events). The C<how> argument must be either	961	has processed all outstanding events). The C<how> argument must be either
266	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	962	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
267	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	963	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		964
		965	This "break state" will be cleared on the next call to C<ev_run>.
		966
		967	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		968	which case it will have no effect.
268		969
269	=item ev_ref (loop)	970	=item ev_ref (loop)
270		971
271	=item ev_unref (loop)	972	=item ev_unref (loop)
272		973
273	Ref/unref can be used to add or remove a reference count on the event	974	Ref/unref can be used to add or remove a reference count on the event
274	loop: Every watcher keeps one reference, and as long as the reference	975	loop: Every watcher keeps one reference, and as long as the reference
275	count is nonzero, C<ev_loop> will not return on its own. If you have	976	count is nonzero, C<ev_run> will not return on its own.
276	a watcher you never unregister that should not keep C<ev_loop> from	977
277	returning, ev_unref() after starting, and ev_ref() before stopping it. For	978	This is useful when you have a watcher that you never intend to
		979	unregister, but that nevertheless should not keep C<ev_run> from
		980	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		981	before stopping it.
		982
278	example, libev itself uses this for its internal signal pipe: It is not	983	As an example, libev itself uses this for its internal signal pipe: It
279	visible to the libev user and should not keep C<ev_loop> from exiting if	984	is not visible to the libev user and should not keep C<ev_run> from
280	no event watchers registered by it are active. It is also an excellent	985	exiting if no event watchers registered by it are active. It is also an
281	way to do this for generic recurring timers or from within third-party	986	excellent way to do this for generic recurring timers or from within
282	libraries. Just remember to I<unref after start> and I<ref before stop>.	987	third-party libraries. Just remember to I<unref after start> and I<ref
		988	before stop> (but only if the watcher wasn't active before, or was active
		989	before, respectively. Note also that libev might stop watchers itself
		990	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		991	in the callback).
		992
		993	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		994	running when nothing else is active.
		995
		996	ev_signal exitsig;
		997	ev_signal_init (&exitsig, sig_cb, SIGINT);
		998	ev_signal_start (loop, &exitsig);
		999	ev_unref (loop);
		1000
		1001	Example: For some weird reason, unregister the above signal handler again.
		1002
		1003	ev_ref (loop);
		1004	ev_signal_stop (loop, &exitsig);
		1005
		1006	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		1007
		1008	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		1009
		1010	These advanced functions influence the time that libev will spend waiting
		1011	for events. Both time intervals are by default C<0>, meaning that libev
		1012	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		1013	latency.
		1014
		1015	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		1016	allows libev to delay invocation of I/O and timer/periodic callbacks
		1017	to increase efficiency of loop iterations (or to increase power-saving
		1018	opportunities).
		1019
		1020	The idea is that sometimes your program runs just fast enough to handle
		1021	one (or very few) event(s) per loop iteration. While this makes the
		1022	program responsive, it also wastes a lot of CPU time to poll for new
		1023	events, especially with backends like C<select ()> which have a high
		1024	overhead for the actual polling but can deliver many events at once.
		1025
		1026	By setting a higher I<io collect interval> you allow libev to spend more
		1027	time collecting I/O events, so you can handle more events per iteration,
		1028	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		1029	C<ev_timer>) will not be affected. Setting this to a non-null value will
		1030	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		1031	sleep time ensures that libev will not poll for I/O events more often then
		1032	once per this interval, on average (as long as the host time resolution is
		1033	good enough).
		1034
		1035	Likewise, by setting a higher I<timeout collect interval> you allow libev
		1036	to spend more time collecting timeouts, at the expense of increased
		1037	latency/jitter/inexactness (the watcher callback will be called
		1038	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		1039	value will not introduce any overhead in libev.
		1040
		1041	Many (busy) programs can usually benefit by setting the I/O collect
		1042	interval to a value near C<0.1> or so, which is often enough for
		1043	interactive servers (of course not for games), likewise for timeouts. It
		1044	usually doesn't make much sense to set it to a lower value than C<0.01>,
		1045	as this approaches the timing granularity of most systems. Note that if
		1046	you do transactions with the outside world and you can't increase the
		1047	parallelity, then this setting will limit your transaction rate (if you
		1048	need to poll once per transaction and the I/O collect interval is 0.01,
		1049	then you can't do more than 100 transactions per second).
		1050
		1051	Setting the I<timeout collect interval> can improve the opportunity for
		1052	saving power, as the program will "bundle" timer callback invocations that
		1053	are "near" in time together, by delaying some, thus reducing the number of
		1054	times the process sleeps and wakes up again. Another useful technique to
		1055	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		1056	they fire on, say, one-second boundaries only.
		1057
		1058	Example: we only need 0.1s timeout granularity, and we wish not to poll
		1059	more often than 100 times per second:
		1060
		1061	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		1062	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		1063
		1064	=item ev_invoke_pending (loop)
		1065
		1066	This call will simply invoke all pending watchers while resetting their
		1067	pending state. Normally, C<ev_run> does this automatically when required,
		1068	but when overriding the invoke callback this call comes handy. This
		1069	function can be invoked from a watcher - this can be useful for example
		1070	when you want to do some lengthy calculation and want to pass further
		1071	event handling to another thread (you still have to make sure only one
		1072	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		1073
		1074	=item int ev_pending_count (loop)
		1075
		1076	Returns the number of pending watchers - zero indicates that no watchers
		1077	are pending.
		1078
		1079	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1080
		1081	This overrides the invoke pending functionality of the loop: Instead of
		1082	invoking all pending watchers when there are any, C<ev_run> will call
		1083	this callback instead. This is useful, for example, when you want to
		1084	invoke the actual watchers inside another context (another thread etc.).
		1085
		1086	If you want to reset the callback, use C<ev_invoke_pending> as new
		1087	callback.
		1088
		1089	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
		1090
		1091	Sometimes you want to share the same loop between multiple threads. This
		1092	can be done relatively simply by putting mutex_lock/unlock calls around
		1093	each call to a libev function.
		1094
		1095	However, C<ev_run> can run an indefinite time, so it is not feasible
		1096	to wait for it to return. One way around this is to wake up the event
		1097	loop via C<ev_break> and C<ev_async_send>, another way is to set these
		1098	I<release> and I<acquire> callbacks on the loop.
		1099
		1100	When set, then C<release> will be called just before the thread is
		1101	suspended waiting for new events, and C<acquire> is called just
		1102	afterwards.
		1103
		1104	Ideally, C<release> will just call your mutex_unlock function, and
		1105	C<acquire> will just call the mutex_lock function again.
		1106
		1107	While event loop modifications are allowed between invocations of
		1108	C<release> and C<acquire> (that's their only purpose after all), no
		1109	modifications done will affect the event loop, i.e. adding watchers will
		1110	have no effect on the set of file descriptors being watched, or the time
		1111	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1112	to take note of any changes you made.
		1113
		1114	In theory, threads executing C<ev_run> will be async-cancel safe between
		1115	invocations of C<release> and C<acquire>.
		1116
		1117	See also the locking example in the C<THREADS> section later in this
		1118	document.
		1119
		1120	=item ev_set_userdata (loop, void *data)
		1121
		1122	=item void *ev_userdata (loop)
		1123
		1124	Set and retrieve a single C<void *> associated with a loop. When
		1125	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1126	C<0>.
		1127
		1128	These two functions can be used to associate arbitrary data with a loop,
		1129	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1130	C<acquire> callbacks described above, but of course can be (ab-)used for
		1131	any other purpose as well.
		1132
		1133	=item ev_verify (loop)
		1134
		1135	This function only does something when C<EV_VERIFY> support has been
		1136	compiled in, which is the default for non-minimal builds. It tries to go
		1137	through all internal structures and checks them for validity. If anything
		1138	is found to be inconsistent, it will print an error message to standard
		1139	error and call C<abort ()>.
		1140
		1141	This can be used to catch bugs inside libev itself: under normal
		1142	circumstances, this function will never abort as of course libev keeps its
		1143	data structures consistent.
283		1144
284	=back	1145	=back
285		1146
		1147
286	=head1 ANATOMY OF A WATCHER	1148	=head1 ANATOMY OF A WATCHER
287		1149
		1150	In the following description, uppercase C<TYPE> in names stands for the
		1151	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1152	watchers and C<ev_io_start> for I/O watchers.
		1153
288	A watcher is a structure that you create and register to record your	1154	A watcher is an opaque structure that you allocate and register to record
289	interest in some event. For instance, if you want to wait for STDIN to	1155	your interest in some event. To make a concrete example, imagine you want
290	become readable, you would create an C<ev_io> watcher for that:	1156	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1157	for that:
291		1158
292	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1159	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
293	{	1160	{
294	ev_io_stop (w);	1161	ev_io_stop (w);
295	ev_unloop (loop, EVUNLOOP_ALL);	1162	ev_break (loop, EVBREAK_ALL);
296	}	1163	}
297		1164
298	struct ev_loop *loop = ev_default_loop (0);	1165	struct ev_loop *loop = ev_default_loop (0);
		1166
299	struct ev_io stdin_watcher;	1167	ev_io stdin_watcher;
		1168
300	ev_init (&stdin_watcher, my_cb);	1169	ev_init (&stdin_watcher, my_cb);
301	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1170	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
302	ev_io_start (loop, &stdin_watcher);	1171	ev_io_start (loop, &stdin_watcher);
		1172
303	ev_loop (loop, 0);	1173	ev_run (loop, 0);
304		1174
305	As you can see, you are responsible for allocating the memory for your	1175	As you can see, you are responsible for allocating the memory for your
306	watcher structures (and it is usually a bad idea to do this on the stack,	1176	watcher structures (and it is I<usually> a bad idea to do this on the
307	although this can sometimes be quite valid).	1177	stack).
308		1178
		1179	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1180	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1181
309	Each watcher structure must be initialised by a call to C<ev_init	1182	Each watcher structure must be initialised by a call to C<ev_init (watcher
310	(watcher *, callback)>, which expects a callback to be provided. This	1183	*, callback)>, which expects a callback to be provided. This callback is
311	callback gets invoked each time the event occurs (or, in the case of io	1184	invoked each time the event occurs (or, in the case of I/O watchers, each
312	watchers, each time the event loop detects that the file descriptor given	1185	time the event loop detects that the file descriptor given is readable
313	is readable and/or writable).	1186	and/or writable).
314		1187
315	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1188	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
316	with arguments specific to this watcher type. There is also a macro	1189	macro to configure it, with arguments specific to the watcher type. There
317	to combine initialisation and setting in one call: C<< ev_<type>_init	1190	is also a macro to combine initialisation and setting in one call: C<<
318	(watcher *, callback, ...) >>.	1191	ev_TYPE_init (watcher *, callback, ...) >>.
319		1192
320	To make the watcher actually watch out for events, you have to start it	1193	To make the watcher actually watch out for events, you have to start it
321	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1194	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
322	*) >>), and you can stop watching for events at any time by calling the	1195	*) >>), and you can stop watching for events at any time by calling the
323	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1196	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
324		1197
325	As long as your watcher is active (has been started but not stopped) you	1198	As long as your watcher is active (has been started but not stopped) you
326	must not touch the values stored in it. Most specifically you must never	1199	must not touch the values stored in it. Most specifically you must never
327	reinitialise it or call its set method.	1200	reinitialise it or call its C<ev_TYPE_set> macro.
328
329	You can check whether an event is active by calling the C<ev_is_active
330	(watcher *)> macro. To see whether an event is outstanding (but the
331	callback for it has not been called yet) you can use the C<ev_is_pending
332	(watcher *)> macro.
333		1201
334	Each and every callback receives the event loop pointer as first, the	1202	Each and every callback receives the event loop pointer as first, the
335	registered watcher structure as second, and a bitset of received events as	1203	registered watcher structure as second, and a bitset of received events as
336	third argument.	1204	third argument.
337		1205
…		…
346	=item C<EV_WRITE>	1214	=item C<EV_WRITE>
347		1215
348	The file descriptor in the C<ev_io> watcher has become readable and/or	1216	The file descriptor in the C<ev_io> watcher has become readable and/or
349	writable.	1217	writable.
350		1218
351	=item C<EV_TIMEOUT>	1219	=item C<EV_TIMER>
352		1220
353	The C<ev_timer> watcher has timed out.	1221	The C<ev_timer> watcher has timed out.
354		1222
355	=item C<EV_PERIODIC>	1223	=item C<EV_PERIODIC>
356		1224
…		…
362		1230
363	=item C<EV_CHILD>	1231	=item C<EV_CHILD>
364		1232
365	The pid specified in the C<ev_child> watcher has received a status change.	1233	The pid specified in the C<ev_child> watcher has received a status change.
366		1234
		1235	=item C<EV_STAT>
		1236
		1237	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1238
367	=item C<EV_IDLE>	1239	=item C<EV_IDLE>
368		1240
369	The C<ev_idle> watcher has determined that you have nothing better to do.	1241	The C<ev_idle> watcher has determined that you have nothing better to do.
370		1242
371	=item C<EV_PREPARE>	1243	=item C<EV_PREPARE>
372		1244
373	=item C<EV_CHECK>	1245	=item C<EV_CHECK>
374		1246
375	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1247	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
376	to gather new events, and all C<ev_check> watchers are invoked just after	1248	gather new events, and all C<ev_check> watchers are queued (not invoked)
377	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1249	just after C<ev_run> has gathered them, but before it queues any callbacks
		1250	for any received events. That means C<ev_prepare> watchers are the last
		1251	watchers invoked before the event loop sleeps or polls for new events, and
		1252	C<ev_check> watchers will be invoked before any other watchers of the same
		1253	or lower priority within an event loop iteration.
		1254
378	received events. Callbacks of both watcher types can start and stop as	1255	Callbacks of both watcher types can start and stop as many watchers as
379	many watchers as they want, and all of them will be taken into account	1256	they want, and all of them will be taken into account (for example, a
380	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1257	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
381	C<ev_loop> from blocking).	1258	blocking).
		1259
		1260	=item C<EV_EMBED>
		1261
		1262	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1263
		1264	=item C<EV_FORK>
		1265
		1266	The event loop has been resumed in the child process after fork (see
		1267	C<ev_fork>).
		1268
		1269	=item C<EV_CLEANUP>
		1270
		1271	The event loop is about to be destroyed (see C<ev_cleanup>).
		1272
		1273	=item C<EV_ASYNC>
		1274
		1275	The given async watcher has been asynchronously notified (see C<ev_async>).
		1276
		1277	=item C<EV_CUSTOM>
		1278
		1279	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1280	by libev users to signal watchers (e.g. via C<ev_feed_event>).
382		1281
383	=item C<EV_ERROR>	1282	=item C<EV_ERROR>
384		1283
385	An unspecified error has occured, the watcher has been stopped. This might	1284	An unspecified error has occurred, the watcher has been stopped. This might
386	happen because the watcher could not be properly started because libev	1285	happen because the watcher could not be properly started because libev
387	ran out of memory, a file descriptor was found to be closed or any other	1286	ran out of memory, a file descriptor was found to be closed or any other
		1287	problem. Libev considers these application bugs.
		1288
388	problem. You best act on it by reporting the problem and somehow coping	1289	You best act on it by reporting the problem and somehow coping with the
389	with the watcher being stopped.	1290	watcher being stopped. Note that well-written programs should not receive
		1291	an error ever, so when your watcher receives it, this usually indicates a
		1292	bug in your program.
390		1293
391	Libev will usually signal a few "dummy" events together with an error,	1294	Libev will usually signal a few "dummy" events together with an error, for
392	for example it might indicate that a fd is readable or writable, and if	1295	example it might indicate that a fd is readable or writable, and if your
393	your callbacks is well-written it can just attempt the operation and cope	1296	callbacks is well-written it can just attempt the operation and cope with
394	with the error from read() or write(). This will not work in multithreaded	1297	the error from read() or write(). This will not work in multi-threaded
395	programs, though, so beware.	1298	programs, though, as the fd could already be closed and reused for another
		1299	thing, so beware.
396		1300
397	=back	1301	=back
398		1302
399	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1303	=head2 GENERIC WATCHER FUNCTIONS
400		1304
401	Each watcher has, by default, a member C<void *data> that you can change	1305	=over 4
402	and read at any time, libev will completely ignore it. This can be used
403	to associate arbitrary data with your watcher. If you need more data and
404	don't want to allocate memory and store a pointer to it in that data
405	member, you can also "subclass" the watcher type and provide your own
406	data:
407		1306
408	struct my_io	1307	=item C<ev_init> (ev_TYPE *watcher, callback)
		1308
		1309	This macro initialises the generic portion of a watcher. The contents
		1310	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1311	the generic parts of the watcher are initialised, you I<need> to call
		1312	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1313	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1314	which rolls both calls into one.
		1315
		1316	You can reinitialise a watcher at any time as long as it has been stopped
		1317	(or never started) and there are no pending events outstanding.
		1318
		1319	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1320	int revents)>.
		1321
		1322	Example: Initialise an C<ev_io> watcher in two steps.
		1323
		1324	ev_io w;
		1325	ev_init (&w, my_cb);
		1326	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1327
		1328	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1329
		1330	This macro initialises the type-specific parts of a watcher. You need to
		1331	call C<ev_init> at least once before you call this macro, but you can
		1332	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1333	macro on a watcher that is active (it can be pending, however, which is a
		1334	difference to the C<ev_init> macro).
		1335
		1336	Although some watcher types do not have type-specific arguments
		1337	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1338
		1339	See C<ev_init>, above, for an example.
		1340
		1341	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1342
		1343	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1344	calls into a single call. This is the most convenient method to initialise
		1345	a watcher. The same limitations apply, of course.
		1346
		1347	Example: Initialise and set an C<ev_io> watcher in one step.
		1348
		1349	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1350
		1351	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1352
		1353	Starts (activates) the given watcher. Only active watchers will receive
		1354	events. If the watcher is already active nothing will happen.
		1355
		1356	Example: Start the C<ev_io> watcher that is being abused as example in this
		1357	whole section.
		1358
		1359	ev_io_start (EV_DEFAULT_UC, &w);
		1360
		1361	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1362
		1363	Stops the given watcher if active, and clears the pending status (whether
		1364	the watcher was active or not).
		1365
		1366	It is possible that stopped watchers are pending - for example,
		1367	non-repeating timers are being stopped when they become pending - but
		1368	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1369	pending. If you want to free or reuse the memory used by the watcher it is
		1370	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1371
		1372	=item bool ev_is_active (ev_TYPE *watcher)
		1373
		1374	Returns a true value iff the watcher is active (i.e. it has been started
		1375	and not yet been stopped). As long as a watcher is active you must not modify
		1376	it.
		1377
		1378	=item bool ev_is_pending (ev_TYPE *watcher)
		1379
		1380	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1381	events but its callback has not yet been invoked). As long as a watcher
		1382	is pending (but not active) you must not call an init function on it (but
		1383	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1384	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1385	it).
		1386
		1387	=item callback ev_cb (ev_TYPE *watcher)
		1388
		1389	Returns the callback currently set on the watcher.
		1390
		1391	=item ev_set_cb (ev_TYPE *watcher, callback)
		1392
		1393	Change the callback. You can change the callback at virtually any time
		1394	(modulo threads).
		1395
		1396	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1397
		1398	=item int ev_priority (ev_TYPE *watcher)
		1399
		1400	Set and query the priority of the watcher. The priority is a small
		1401	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1402	(default: C<-2>). Pending watchers with higher priority will be invoked
		1403	before watchers with lower priority, but priority will not keep watchers
		1404	from being executed (except for C<ev_idle> watchers).
		1405
		1406	If you need to suppress invocation when higher priority events are pending
		1407	you need to look at C<ev_idle> watchers, which provide this functionality.
		1408
		1409	You I<must not> change the priority of a watcher as long as it is active or
		1410	pending.
		1411
		1412	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1413	fine, as long as you do not mind that the priority value you query might
		1414	or might not have been clamped to the valid range.
		1415
		1416	The default priority used by watchers when no priority has been set is
		1417	always C<0>, which is supposed to not be too high and not be too low :).
		1418
		1419	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1420	priorities.
		1421
		1422	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1423
		1424	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1425	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1426	can deal with that fact, as both are simply passed through to the
		1427	callback.
		1428
		1429	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1430
		1431	If the watcher is pending, this function clears its pending status and
		1432	returns its C<revents> bitset (as if its callback was invoked). If the
		1433	watcher isn't pending it does nothing and returns C<0>.
		1434
		1435	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1436	callback to be invoked, which can be accomplished with this function.
		1437
		1438	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1439
		1440	Feeds the given event set into the event loop, as if the specified event
		1441	had happened for the specified watcher (which must be a pointer to an
		1442	initialised but not necessarily started event watcher). Obviously you must
		1443	not free the watcher as long as it has pending events.
		1444
		1445	Stopping the watcher, letting libev invoke it, or calling
		1446	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1447	not started in the first place.
		1448
		1449	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1450	functions that do not need a watcher.
		1451
		1452	=back
		1453
		1454	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
		1455	OWN COMPOSITE WATCHERS> idioms.
		1456
		1457	=head2 WATCHER STATES
		1458
		1459	There are various watcher states mentioned throughout this manual -
		1460	active, pending and so on. In this section these states and the rules to
		1461	transition between them will be described in more detail - and while these
		1462	rules might look complicated, they usually do "the right thing".
		1463
		1464	=over 4
		1465
		1466	=item initialised
		1467
		1468	Before a watcher can be registered with the event loop it has to be
		1469	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1470	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1471
		1472	In this state it is simply some block of memory that is suitable for
		1473	use in an event loop. It can be moved around, freed, reused etc. at
		1474	will - as long as you either keep the memory contents intact, or call
		1475	C<ev_TYPE_init> again.
		1476
		1477	=item started/running/active
		1478
		1479	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1480	property of the event loop, and is actively waiting for events. While in
		1481	this state it cannot be accessed (except in a few documented ways), moved,
		1482	freed or anything else - the only legal thing is to keep a pointer to it,
		1483	and call libev functions on it that are documented to work on active watchers.
		1484
		1485	=item pending
		1486
		1487	If a watcher is active and libev determines that an event it is interested
		1488	in has occurred (such as a timer expiring), it will become pending. It will
		1489	stay in this pending state until either it is stopped or its callback is
		1490	about to be invoked, so it is not normally pending inside the watcher
		1491	callback.
		1492
		1493	The watcher might or might not be active while it is pending (for example,
		1494	an expired non-repeating timer can be pending but no longer active). If it
		1495	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1496	but it is still property of the event loop at this time, so cannot be
		1497	moved, freed or reused. And if it is active the rules described in the
		1498	previous item still apply.
		1499
		1500	It is also possible to feed an event on a watcher that is not active (e.g.
		1501	via C<ev_feed_event>), in which case it becomes pending without being
		1502	active.
		1503
		1504	=item stopped
		1505
		1506	A watcher can be stopped implicitly by libev (in which case it might still
		1507	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1508	latter will clear any pending state the watcher might be in, regardless
		1509	of whether it was active or not, so stopping a watcher explicitly before
		1510	freeing it is often a good idea.
		1511
		1512	While stopped (and not pending) the watcher is essentially in the
		1513	initialised state, that is, it can be reused, moved, modified in any way
		1514	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1515	it again).
		1516
		1517	=back
		1518
		1519	=head2 WATCHER PRIORITY MODELS
		1520
		1521	Many event loops support I<watcher priorities>, which are usually small
		1522	integers that influence the ordering of event callback invocation
		1523	between watchers in some way, all else being equal.
		1524
		1525	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1526	description for the more technical details such as the actual priority
		1527	range.
		1528
		1529	There are two common ways how these these priorities are being interpreted
		1530	by event loops:
		1531
		1532	In the more common lock-out model, higher priorities "lock out" invocation
		1533	of lower priority watchers, which means as long as higher priority
		1534	watchers receive events, lower priority watchers are not being invoked.
		1535
		1536	The less common only-for-ordering model uses priorities solely to order
		1537	callback invocation within a single event loop iteration: Higher priority
		1538	watchers are invoked before lower priority ones, but they all get invoked
		1539	before polling for new events.
		1540
		1541	Libev uses the second (only-for-ordering) model for all its watchers
		1542	except for idle watchers (which use the lock-out model).
		1543
		1544	The rationale behind this is that implementing the lock-out model for
		1545	watchers is not well supported by most kernel interfaces, and most event
		1546	libraries will just poll for the same events again and again as long as
		1547	their callbacks have not been executed, which is very inefficient in the
		1548	common case of one high-priority watcher locking out a mass of lower
		1549	priority ones.
		1550
		1551	Static (ordering) priorities are most useful when you have two or more
		1552	watchers handling the same resource: a typical usage example is having an
		1553	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1554	timeouts. Under load, data might be received while the program handles
		1555	other jobs, but since timers normally get invoked first, the timeout
		1556	handler will be executed before checking for data. In that case, giving
		1557	the timer a lower priority than the I/O watcher ensures that I/O will be
		1558	handled first even under adverse conditions (which is usually, but not
		1559	always, what you want).
		1560
		1561	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1562	will only be executed when no same or higher priority watchers have
		1563	received events, they can be used to implement the "lock-out" model when
		1564	required.
		1565
		1566	For example, to emulate how many other event libraries handle priorities,
		1567	you can associate an C<ev_idle> watcher to each such watcher, and in
		1568	the normal watcher callback, you just start the idle watcher. The real
		1569	processing is done in the idle watcher callback. This causes libev to
		1570	continuously poll and process kernel event data for the watcher, but when
		1571	the lock-out case is known to be rare (which in turn is rare :), this is
		1572	workable.
		1573
		1574	Usually, however, the lock-out model implemented that way will perform
		1575	miserably under the type of load it was designed to handle. In that case,
		1576	it might be preferable to stop the real watcher before starting the
		1577	idle watcher, so the kernel will not have to process the event in case
		1578	the actual processing will be delayed for considerable time.
		1579
		1580	Here is an example of an I/O watcher that should run at a strictly lower
		1581	priority than the default, and which should only process data when no
		1582	other events are pending:
		1583
		1584	ev_idle idle; // actual processing watcher
		1585	ev_io io; // actual event watcher
		1586
		1587	static void
		1588	io_cb (EV_P_ ev_io *w, int revents)
409	{	1589	{
410	struct ev_io io;	1590	// stop the I/O watcher, we received the event, but
411	int otherfd;	1591	// are not yet ready to handle it.
412	void *somedata;	1592	ev_io_stop (EV_A_ w);
413	struct whatever *mostinteresting;	1593
		1594	// start the idle watcher to handle the actual event.
		1595	// it will not be executed as long as other watchers
		1596	// with the default priority are receiving events.
		1597	ev_idle_start (EV_A_ &idle);
414	}	1598	}
415		1599
416	And since your callback will be called with a pointer to the watcher, you	1600	static void
417	can cast it back to your own type:	1601	idle_cb (EV_P_ ev_idle *w, int revents)
418
419	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
420	{	1602	{
421	struct my_io *w = (struct my_io *)w_;	1603	// actual processing
422	...	1604	read (STDIN_FILENO, ...);
		1605
		1606	// have to start the I/O watcher again, as
		1607	// we have handled the event
		1608	ev_io_start (EV_P_ &io);
423	}	1609	}
424		1610
425	More interesting and less C-conformant ways of catsing your callback type	1611	// initialisation
426	have been omitted....	1612	ev_idle_init (&idle, idle_cb);
		1613	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1614	ev_io_start (EV_DEFAULT_ &io);
		1615
		1616	In the "real" world, it might also be beneficial to start a timer, so that
		1617	low-priority connections can not be locked out forever under load. This
		1618	enables your program to keep a lower latency for important connections
		1619	during short periods of high load, while not completely locking out less
		1620	important ones.
427		1621
428		1622
429	=head1 WATCHER TYPES	1623	=head1 WATCHER TYPES
430		1624
431	This section describes each watcher in detail, but will not repeat	1625	This section describes each watcher in detail, but will not repeat
432	information given in the last section.	1626	information given in the last section. Any initialisation/set macros,
		1627	functions and members specific to the watcher type are explained.
433		1628
		1629	Members are additionally marked with either I<[read-only]>, meaning that,
		1630	while the watcher is active, you can look at the member and expect some
		1631	sensible content, but you must not modify it (you can modify it while the
		1632	watcher is stopped to your hearts content), or I<[read-write]>, which
		1633	means you can expect it to have some sensible content while the watcher
		1634	is active, but you can also modify it. Modifying it may not do something
		1635	sensible or take immediate effect (or do anything at all), but libev will
		1636	not crash or malfunction in any way.
		1637
		1638
434	=head2 C<ev_io> - is this file descriptor readable or writable	1639	=head2 C<ev_io> - is this file descriptor readable or writable?
435		1640
436	I/O watchers check whether a file descriptor is readable or writable	1641	I/O watchers check whether a file descriptor is readable or writable
437	in each iteration of the event loop (This behaviour is called	1642	in each iteration of the event loop, or, more precisely, when reading
438	level-triggering because you keep receiving events as long as the	1643	would not block the process and writing would at least be able to write
439	condition persists. Remember you can stop the watcher if you don't want to	1644	some data. This behaviour is called level-triggering because you keep
440	act on the event and neither want to receive future events).	1645	receiving events as long as the condition persists. Remember you can stop
		1646	the watcher if you don't want to act on the event and neither want to
		1647	receive future events.
441		1648
442	In general you can register as many read and/or write event watchers per	1649	In general you can register as many read and/or write event watchers per
443	fd as you want (as long as you don't confuse yourself). Setting all file	1650	fd as you want (as long as you don't confuse yourself). Setting all file
444	descriptors to non-blocking mode is also usually a good idea (but not	1651	descriptors to non-blocking mode is also usually a good idea (but not
445	required if you know what you are doing).	1652	required if you know what you are doing).
446		1653
447	You have to be careful with dup'ed file descriptors, though. Some backends	1654	Another thing you have to watch out for is that it is quite easy to
448	(the linux epoll backend is a notable example) cannot handle dup'ed file	1655	receive "spurious" readiness notifications, that is, your callback might
449	descriptors correctly if you register interest in two or more fds pointing	1656	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
450	to the same underlying file/socket etc. description (that is, they share	1657	because there is no data. It is very easy to get into this situation even
451	the same underlying "file open").	1658	with a relatively standard program structure. Thus it is best to always
		1659	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
		1660	preferable to a program hanging until some data arrives.
452		1661
453	If you must do this, then force the use of a known-to-be-good backend	1662	If you cannot run the fd in non-blocking mode (for example you should
454	(at the time of this writing, this includes only EVMETHOD_SELECT and	1663	not play around with an Xlib connection), then you have to separately
455	EVMETHOD_POLL).	1664	re-test whether a file descriptor is really ready with a known-to-be good
		1665	interface such as poll (fortunately in the case of Xlib, it already does
		1666	this on its own, so its quite safe to use). Some people additionally
		1667	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1668	indefinitely.
		1669
		1670	But really, best use non-blocking mode.
		1671
		1672	=head3 The special problem of disappearing file descriptors
		1673
		1674	Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
		1675	a file descriptor (either due to calling C<close> explicitly or any other
		1676	means, such as C<dup2>). The reason is that you register interest in some
		1677	file descriptor, but when it goes away, the operating system will silently
		1678	drop this interest. If another file descriptor with the same number then
		1679	is registered with libev, there is no efficient way to see that this is,
		1680	in fact, a different file descriptor.
		1681
		1682	To avoid having to explicitly tell libev about such cases, libev follows
		1683	the following policy: Each time C<ev_io_set> is being called, libev
		1684	will assume that this is potentially a new file descriptor, otherwise
		1685	it is assumed that the file descriptor stays the same. That means that
		1686	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1687	descriptor even if the file descriptor number itself did not change.
		1688
		1689	This is how one would do it normally anyway, the important point is that
		1690	the libev application should not optimise around libev but should leave
		1691	optimisations to libev.
		1692
		1693	=head3 The special problem of dup'ed file descriptors
		1694
		1695	Some backends (e.g. epoll), cannot register events for file descriptors,
		1696	but only events for the underlying file descriptions. That means when you
		1697	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1698	events for them, only one file descriptor might actually receive events.
		1699
		1700	There is no workaround possible except not registering events
		1701	for potentially C<dup ()>'ed file descriptors, or to resort to
		1702	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1703
		1704	=head3 The special problem of files
		1705
		1706	Many people try to use C<select> (or libev) on file descriptors
		1707	representing files, and expect it to become ready when their program
		1708	doesn't block on disk accesses (which can take a long time on their own).
		1709
		1710	However, this cannot ever work in the "expected" way - you get a readiness
		1711	notification as soon as the kernel knows whether and how much data is
		1712	there, and in the case of open files, that's always the case, so you
		1713	always get a readiness notification instantly, and your read (or possibly
		1714	write) will still block on the disk I/O.
		1715
		1716	Another way to view it is that in the case of sockets, pipes, character
		1717	devices and so on, there is another party (the sender) that delivers data
		1718	on its own, but in the case of files, there is no such thing: the disk
		1719	will not send data on its own, simply because it doesn't know what you
		1720	wish to read - you would first have to request some data.
		1721
		1722	Since files are typically not-so-well supported by advanced notification
		1723	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1724	to files, even though you should not use it. The reason for this is
		1725	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1726	usually a tty, often a pipe, but also sometimes files or special devices
		1727	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1728	F</dev/urandom>), and even though the file might better be served with
		1729	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1730	it "just works" instead of freezing.
		1731
		1732	So avoid file descriptors pointing to files when you know it (e.g. use
		1733	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1734	when you rarely read from a file instead of from a socket, and want to
		1735	reuse the same code path.
		1736
		1737	=head3 The special problem of fork
		1738
		1739	Some backends (epoll, kqueue, probably linuxaio) do not support C<fork ()>
		1740	at all or exhibit useless behaviour. Libev fully supports fork, but needs
		1741	to be told about it in the child if you want to continue to use it in the
		1742	child.
		1743
		1744	To support fork in your child processes, you have to call C<ev_loop_fork
		1745	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
		1746	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1747
		1748	=head3 The special problem of SIGPIPE
		1749
		1750	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1751	when writing to a pipe whose other end has been closed, your program gets
		1752	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1753	this is sensible behaviour, for daemons, this is usually undesirable.
		1754
		1755	So when you encounter spurious, unexplained daemon exits, make sure you
		1756	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1757	somewhere, as that would have given you a big clue).
		1758
		1759	=head3 The special problem of accept()ing when you can't
		1760
		1761	Many implementations of the POSIX C<accept> function (for example,
		1762	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1763	connection from the pending queue in all error cases.
		1764
		1765	For example, larger servers often run out of file descriptors (because
		1766	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1767	rejecting the connection, leading to libev signalling readiness on
		1768	the next iteration again (the connection still exists after all), and
		1769	typically causing the program to loop at 100% CPU usage.
		1770
		1771	Unfortunately, the set of errors that cause this issue differs between
		1772	operating systems, there is usually little the app can do to remedy the
		1773	situation, and no known thread-safe method of removing the connection to
		1774	cope with overload is known (to me).
		1775
		1776	One of the easiest ways to handle this situation is to just ignore it
		1777	- when the program encounters an overload, it will just loop until the
		1778	situation is over. While this is a form of busy waiting, no OS offers an
		1779	event-based way to handle this situation, so it's the best one can do.
		1780
		1781	A better way to handle the situation is to log any errors other than
		1782	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1783	messages, and continue as usual, which at least gives the user an idea of
		1784	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1785	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1786	usage.
		1787
		1788	If your program is single-threaded, then you could also keep a dummy file
		1789	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1790	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1791	close that fd, and create a new dummy fd. This will gracefully refuse
		1792	clients under typical overload conditions.
		1793
		1794	The last way to handle it is to simply log the error and C<exit>, as
		1795	is often done with C<malloc> failures, but this results in an easy
		1796	opportunity for a DoS attack.
		1797
		1798	=head3 Watcher-Specific Functions
456		1799
457	=over 4	1800	=over 4
458		1801
459	=item ev_io_init (ev_io *, callback, int fd, int events)	1802	=item ev_io_init (ev_io *, callback, int fd, int events)
460		1803
461	=item ev_io_set (ev_io *, int fd, int events)	1804	=item ev_io_set (ev_io *, int fd, int events)
462		1805
463	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1806	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
464	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1807	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
465	EV_WRITE> to receive the given events.	1808	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1809
		1810	=item int fd [read-only]
		1811
		1812	The file descriptor being watched.
		1813
		1814	=item int events [read-only]
		1815
		1816	The events being watched.
466		1817
467	=back	1818	=back
468		1819
		1820	=head3 Examples
		1821
		1822	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1823	readable, but only once. Since it is likely line-buffered, you could
		1824	attempt to read a whole line in the callback.
		1825
		1826	static void
		1827	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1828	{
		1829	ev_io_stop (loop, w);
		1830	.. read from stdin here (or from w->fd) and handle any I/O errors
		1831	}
		1832
		1833	...
		1834	struct ev_loop *loop = ev_default_init (0);
		1835	ev_io stdin_readable;
		1836	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1837	ev_io_start (loop, &stdin_readable);
		1838	ev_run (loop, 0);
		1839
		1840
469	=head2 C<ev_timer> - relative and optionally recurring timeouts	1841	=head2 C<ev_timer> - relative and optionally repeating timeouts
470		1842
471	Timer watchers are simple relative timers that generate an event after a	1843	Timer watchers are simple relative timers that generate an event after a
472	given time, and optionally repeating in regular intervals after that.	1844	given time, and optionally repeating in regular intervals after that.
473		1845
474	The timers are based on real time, that is, if you register an event that	1846	The timers are based on real time, that is, if you register an event that
475	times out after an hour and you reset your system clock to last years	1847	times out after an hour and you reset your system clock to January last
476	time, it will still time out after (roughly) and hour. "Roughly" because	1848	year, it will still time out after (roughly) one hour. "Roughly" because
477	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1849	detecting time jumps is hard, and some inaccuracies are unavoidable (the
478	monotonic clock option helps a lot here).	1850	monotonic clock option helps a lot here).
		1851
		1852	The callback is guaranteed to be invoked only I<after> its timeout has
		1853	passed (not I<at>, so on systems with very low-resolution clocks this
		1854	might introduce a small delay, see "the special problem of being too
		1855	early", below). If multiple timers become ready during the same loop
		1856	iteration then the ones with earlier time-out values are invoked before
		1857	ones of the same priority with later time-out values (but this is no
		1858	longer true when a callback calls C<ev_run> recursively).
		1859
		1860	=head3 Be smart about timeouts
		1861
		1862	Many real-world problems involve some kind of timeout, usually for error
		1863	recovery. A typical example is an HTTP request - if the other side hangs,
		1864	you want to raise some error after a while.
		1865
		1866	What follows are some ways to handle this problem, from obvious and
		1867	inefficient to smart and efficient.
		1868
		1869	In the following, a 60 second activity timeout is assumed - a timeout that
		1870	gets reset to 60 seconds each time there is activity (e.g. each time some
		1871	data or other life sign was received).
		1872
		1873	=over 4
		1874
		1875	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1876
		1877	This is the most obvious, but not the most simple way: In the beginning,
		1878	start the watcher:
		1879
		1880	ev_timer_init (timer, callback, 60., 0.);
		1881	ev_timer_start (loop, timer);
		1882
		1883	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1884	and start it again:
		1885
		1886	ev_timer_stop (loop, timer);
		1887	ev_timer_set (timer, 60., 0.);
		1888	ev_timer_start (loop, timer);
		1889
		1890	This is relatively simple to implement, but means that each time there is
		1891	some activity, libev will first have to remove the timer from its internal
		1892	data structure and then add it again. Libev tries to be fast, but it's
		1893	still not a constant-time operation.
		1894
		1895	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1896
		1897	This is the easiest way, and involves using C<ev_timer_again> instead of
		1898	C<ev_timer_start>.
		1899
		1900	To implement this, configure an C<ev_timer> with a C<repeat> value
		1901	of C<60> and then call C<ev_timer_again> at start and each time you
		1902	successfully read or write some data. If you go into an idle state where
		1903	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1904	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1905
		1906	That means you can ignore both the C<ev_timer_start> function and the
		1907	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1908	member and C<ev_timer_again>.
		1909
		1910	At start:
		1911
		1912	ev_init (timer, callback);
		1913	timer->repeat = 60.;
		1914	ev_timer_again (loop, timer);
		1915
		1916	Each time there is some activity:
		1917
		1918	ev_timer_again (loop, timer);
		1919
		1920	It is even possible to change the time-out on the fly, regardless of
		1921	whether the watcher is active or not:
		1922
		1923	timer->repeat = 30.;
		1924	ev_timer_again (loop, timer);
		1925
		1926	This is slightly more efficient then stopping/starting the timer each time
		1927	you want to modify its timeout value, as libev does not have to completely
		1928	remove and re-insert the timer from/into its internal data structure.
		1929
		1930	It is, however, even simpler than the "obvious" way to do it.
		1931
		1932	=item 3. Let the timer time out, but then re-arm it as required.
		1933
		1934	This method is more tricky, but usually most efficient: Most timeouts are
		1935	relatively long compared to the intervals between other activity - in
		1936	our example, within 60 seconds, there are usually many I/O events with
		1937	associated activity resets.
		1938
		1939	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1940	but remember the time of last activity, and check for a real timeout only
		1941	within the callback:
		1942
		1943	ev_tstamp timeout = 60.;
		1944	ev_tstamp last_activity; // time of last activity
		1945	ev_timer timer;
		1946
		1947	static void
		1948	callback (EV_P_ ev_timer *w, int revents)
		1949	{
		1950	// calculate when the timeout would happen
		1951	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
		1952
		1953	// if negative, it means we the timeout already occurred
		1954	if (after < 0.)
		1955	{
		1956	// timeout occurred, take action
		1957	}
		1958	else
		1959	{
		1960	// callback was invoked, but there was some recent
		1961	// activity. simply restart the timer to time out
		1962	// after "after" seconds, which is the earliest time
		1963	// the timeout can occur.
		1964	ev_timer_set (w, after, 0.);
		1965	ev_timer_start (EV_A_ w);
		1966	}
		1967	}
		1968
		1969	To summarise the callback: first calculate in how many seconds the
		1970	timeout will occur (by calculating the absolute time when it would occur,
		1971	C<last_activity + timeout>, and subtracting the current time, C<ev_now
		1972	(EV_A)> from that).
		1973
		1974	If this value is negative, then we are already past the timeout, i.e. we
		1975	timed out, and need to do whatever is needed in this case.
		1976
		1977	Otherwise, we now the earliest time at which the timeout would trigger,
		1978	and simply start the timer with this timeout value.
		1979
		1980	In other words, each time the callback is invoked it will check whether
		1981	the timeout occurred. If not, it will simply reschedule itself to check
		1982	again at the earliest time it could time out. Rinse. Repeat.
		1983
		1984	This scheme causes more callback invocations (about one every 60 seconds
		1985	minus half the average time between activity), but virtually no calls to
		1986	libev to change the timeout.
		1987
		1988	To start the machinery, simply initialise the watcher and set
		1989	C<last_activity> to the current time (meaning there was some activity just
		1990	now), then call the callback, which will "do the right thing" and start
		1991	the timer:
		1992
		1993	last_activity = ev_now (EV_A);
		1994	ev_init (&timer, callback);
		1995	callback (EV_A_ &timer, 0);
		1996
		1997	When there is some activity, simply store the current time in
		1998	C<last_activity>, no libev calls at all:
		1999
		2000	if (activity detected)
		2001	last_activity = ev_now (EV_A);
		2002
		2003	When your timeout value changes, then the timeout can be changed by simply
		2004	providing a new value, stopping the timer and calling the callback, which
		2005	will again do the right thing (for example, time out immediately :).
		2006
		2007	timeout = new_value;
		2008	ev_timer_stop (EV_A_ &timer);
		2009	callback (EV_A_ &timer, 0);
		2010
		2011	This technique is slightly more complex, but in most cases where the
		2012	time-out is unlikely to be triggered, much more efficient.
		2013
		2014	=item 4. Wee, just use a double-linked list for your timeouts.
		2015
		2016	If there is not one request, but many thousands (millions...), all
		2017	employing some kind of timeout with the same timeout value, then one can
		2018	do even better:
		2019
		2020	When starting the timeout, calculate the timeout value and put the timeout
		2021	at the I<end> of the list.
		2022
		2023	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		2024	the list is expected to fire (for example, using the technique #3).
		2025
		2026	When there is some activity, remove the timer from the list, recalculate
		2027	the timeout, append it to the end of the list again, and make sure to
		2028	update the C<ev_timer> if it was taken from the beginning of the list.
		2029
		2030	This way, one can manage an unlimited number of timeouts in O(1) time for
		2031	starting, stopping and updating the timers, at the expense of a major
		2032	complication, and having to use a constant timeout. The constant timeout
		2033	ensures that the list stays sorted.
		2034
		2035	=back
		2036
		2037	So which method the best?
		2038
		2039	Method #2 is a simple no-brain-required solution that is adequate in most
		2040	situations. Method #3 requires a bit more thinking, but handles many cases
		2041	better, and isn't very complicated either. In most case, choosing either
		2042	one is fine, with #3 being better in typical situations.
		2043
		2044	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		2045	rather complicated, but extremely efficient, something that really pays
		2046	off after the first million or so of active timers, i.e. it's usually
		2047	overkill :)
		2048
		2049	=head3 The special problem of being too early
		2050
		2051	If you ask a timer to call your callback after three seconds, then
		2052	you expect it to be invoked after three seconds - but of course, this
		2053	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		2054	guaranteed to any precision by libev - imagine somebody suspending the
		2055	process with a STOP signal for a few hours for example.
		2056
		2057	So, libev tries to invoke your callback as soon as possible I<after> the
		2058	delay has occurred, but cannot guarantee this.
		2059
		2060	A less obvious failure mode is calling your callback too early: many event
		2061	loops compare timestamps with a "elapsed delay >= requested delay", but
		2062	this can cause your callback to be invoked much earlier than you would
		2063	expect.
		2064
		2065	To see why, imagine a system with a clock that only offers full second
		2066	resolution (think windows if you can't come up with a broken enough OS
		2067	yourself). If you schedule a one-second timer at the time 500.9, then the
		2068	event loop will schedule your timeout to elapse at a system time of 500
		2069	(500.9 truncated to the resolution) + 1, or 501.
		2070
		2071	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2072	501" and invoke the callback 0.1s after it was started, even though a
		2073	one-second delay was requested - this is being "too early", despite best
		2074	intentions.
		2075
		2076	This is the reason why libev will never invoke the callback if the elapsed
		2077	delay equals the requested delay, but only when the elapsed delay is
		2078	larger than the requested delay. In the example above, libev would only invoke
		2079	the callback at system time 502, or 1.1s after the timer was started.
		2080
		2081	So, while libev cannot guarantee that your callback will be invoked
		2082	exactly when requested, it I<can> and I<does> guarantee that the requested
		2083	delay has actually elapsed, or in other words, it always errs on the "too
		2084	late" side of things.
		2085
		2086	=head3 The special problem of time updates
		2087
		2088	Establishing the current time is a costly operation (it usually takes
		2089	at least one system call): EV therefore updates its idea of the current
		2090	time only before and after C<ev_run> collects new events, which causes a
		2091	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		2092	lots of events in one iteration.
479		2093
480	The relative timeouts are calculated relative to the C<ev_now ()>	2094	The relative timeouts are calculated relative to the C<ev_now ()>
481	time. This is usually the right thing as this timestamp refers to the time	2095	time. This is usually the right thing as this timestamp refers to the time
482	of the event triggering whatever timeout you are modifying/starting. If	2096	of the event triggering whatever timeout you are modifying/starting. If
483	you suspect event processing to be delayed and you *need* to base the timeout	2097	you suspect event processing to be delayed and you I<need> to base the
484	on the current time, use something like this to adjust for this:	2098	timeout on the current time, use something like the following to adjust
		2099	for it:
485		2100
486	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2101	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
		2102
		2103	If the event loop is suspended for a long time, you can also force an
		2104	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		2105	()>, although that will push the event time of all outstanding events
		2106	further into the future.
		2107
		2108	=head3 The special problem of unsynchronised clocks
		2109
		2110	Modern systems have a variety of clocks - libev itself uses the normal
		2111	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2112	jumps).
		2113
		2114	Neither of these clocks is synchronised with each other or any other clock
		2115	on the system, so C<ev_time ()> might return a considerably different time
		2116	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2117	a call to C<gettimeofday> might return a second count that is one higher
		2118	than a directly following call to C<time>.
		2119
		2120	The moral of this is to only compare libev-related timestamps with
		2121	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2122	a second or so.
		2123
		2124	One more problem arises due to this lack of synchronisation: if libev uses
		2125	the system monotonic clock and you compare timestamps from C<ev_time>
		2126	or C<ev_now> from when you started your timer and when your callback is
		2127	invoked, you will find that sometimes the callback is a bit "early".
		2128
		2129	This is because C<ev_timer>s work in real time, not wall clock time, so
		2130	libev makes sure your callback is not invoked before the delay happened,
		2131	I<measured according to the real time>, not the system clock.
		2132
		2133	If your timeouts are based on a physical timescale (e.g. "time out this
		2134	connection after 100 seconds") then this shouldn't bother you as it is
		2135	exactly the right behaviour.
		2136
		2137	If you want to compare wall clock/system timestamps to your timers, then
		2138	you need to use C<ev_periodic>s, as these are based on the wall clock
		2139	time, where your comparisons will always generate correct results.
		2140
		2141	=head3 The special problems of suspended animation
		2142
		2143	When you leave the server world it is quite customary to hit machines that
		2144	can suspend/hibernate - what happens to the clocks during such a suspend?
		2145
		2146	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		2147	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2148	to run until the system is suspended, but they will not advance while the
		2149	system is suspended. That means, on resume, it will be as if the program
		2150	was frozen for a few seconds, but the suspend time will not be counted
		2151	towards C<ev_timer> when a monotonic clock source is used. The real time
		2152	clock advanced as expected, but if it is used as sole clocksource, then a
		2153	long suspend would be detected as a time jump by libev, and timers would
		2154	be adjusted accordingly.
		2155
		2156	I would not be surprised to see different behaviour in different between
		2157	operating systems, OS versions or even different hardware.
		2158
		2159	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2160	time jump in the monotonic clocks and the realtime clock. If the program
		2161	is suspended for a very long time, and monotonic clock sources are in use,
		2162	then you can expect C<ev_timer>s to expire as the full suspension time
		2163	will be counted towards the timers. When no monotonic clock source is in
		2164	use, then libev will again assume a timejump and adjust accordingly.
		2165
		2166	It might be beneficial for this latter case to call C<ev_suspend>
		2167	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2168	deterministic behaviour in this case (you can do nothing against
		2169	C<SIGSTOP>).
		2170
		2171	=head3 Watcher-Specific Functions and Data Members
487		2172
488	=over 4	2173	=over 4
489		2174
490	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2175	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
491		2176
492	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2177	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
493		2178
494	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2179	Configure the timer to trigger after C<after> seconds (fractional and
495	C<0.>, then it will automatically be stopped. If it is positive, then the	2180	negative values are supported). If C<repeat> is C<0.>, then it will
		2181	automatically be stopped once the timeout is reached. If it is positive,
496	timer will automatically be configured to trigger again C<repeat> seconds	2182	then the timer will automatically be configured to trigger again C<repeat>
497	later, again, and again, until stopped manually.	2183	seconds later, again, and again, until stopped manually.
498		2184
499	The timer itself will do a best-effort at avoiding drift, that is, if you	2185	The timer itself will do a best-effort at avoiding drift, that is, if
500	configure a timer to trigger every 10 seconds, then it will trigger at	2186	you configure a timer to trigger every 10 seconds, then it will normally
501	exactly 10 second intervals. If, however, your program cannot keep up with	2187	trigger at exactly 10 second intervals. If, however, your program cannot
502	the timer (because it takes longer than those 10 seconds to do stuff) the	2188	keep up with the timer (because it takes longer than those 10 seconds to
503	timer will not fire more than once per event loop iteration.	2189	do stuff) the timer will not fire more than once per event loop iteration.
504		2190
505	=item ev_timer_again (loop)	2191	=item ev_timer_again (loop, ev_timer *)
506		2192
507	This will act as if the timer timed out and restart it again if it is	2193	This will act as if the timer timed out, and restarts it again if it is
508	repeating. The exact semantics are:	2194	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2195	timeout to the C<repeat> value and calling C<ev_timer_start>.
509		2196
		2197	The exact semantics are as in the following rules, all of which will be
		2198	applied to the watcher:
		2199
		2200	=over 4
		2201
		2202	=item If the timer is pending, the pending status is always cleared.
		2203
510	If the timer is started but nonrepeating, stop it.	2204	=item If the timer is started but non-repeating, stop it (as if it timed
		2205	out, without invoking it).
511		2206
512	If the timer is repeating, either start it if necessary (with the repeat	2207	=item If the timer is repeating, make the C<repeat> value the new timeout
513	value), or reset the running timer to the repeat value.	2208	and start the timer, if necessary.
514
515	This sounds a bit complicated, but here is a useful and typical
516	example: Imagine you have a tcp connection and you want a so-called idle
517	timeout, that is, you want to be called when there have been, say, 60
518	seconds of inactivity on the socket. The easiest way to do this is to
519	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each
520	time you successfully read or write some data. If you go into an idle
521	state where you do not expect data to travel on the socket, you can stop
522	the timer, and again will automatically restart it if need be.
523		2209
524	=back	2210	=back
525		2211
		2212	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
		2213	usage example.
		2214
		2215	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
		2216
		2217	Returns the remaining time until a timer fires. If the timer is active,
		2218	then this time is relative to the current event loop time, otherwise it's
		2219	the timeout value currently configured.
		2220
		2221	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2222	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2223	will return C<4>. When the timer expires and is restarted, it will return
		2224	roughly C<7> (likely slightly less as callback invocation takes some time,
		2225	too), and so on.
		2226
		2227	=item ev_tstamp repeat [read-write]
		2228
		2229	The current C<repeat> value. Will be used each time the watcher times out
		2230	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2231	which is also when any modifications are taken into account.
		2232
		2233	=back
		2234
		2235	=head3 Examples
		2236
		2237	Example: Create a timer that fires after 60 seconds.
		2238
		2239	static void
		2240	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2241	{
		2242	.. one minute over, w is actually stopped right here
		2243	}
		2244
		2245	ev_timer mytimer;
		2246	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2247	ev_timer_start (loop, &mytimer);
		2248
		2249	Example: Create a timeout timer that times out after 10 seconds of
		2250	inactivity.
		2251
		2252	static void
		2253	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2254	{
		2255	.. ten seconds without any activity
		2256	}
		2257
		2258	ev_timer mytimer;
		2259	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2260	ev_timer_again (&mytimer); /* start timer */
		2261	ev_run (loop, 0);
		2262
		2263	// and in some piece of code that gets executed on any "activity":
		2264	// reset the timeout to start ticking again at 10 seconds
		2265	ev_timer_again (&mytimer);
		2266
		2267
526	=head2 C<ev_periodic> - to cron or not to cron	2268	=head2 C<ev_periodic> - to cron or not to cron?
527		2269
528	Periodic watchers are also timers of a kind, but they are very versatile	2270	Periodic watchers are also timers of a kind, but they are very versatile
529	(and unfortunately a bit complex).	2271	(and unfortunately a bit complex).
530		2272
531	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2273	Unlike C<ev_timer>, periodic watchers are not based on real time (or
532	but on wallclock time (absolute time). You can tell a periodic watcher	2274	relative time, the physical time that passes) but on wall clock time
533	to trigger "at" some specific point in time. For example, if you tell a	2275	(absolute time, the thing you can read on your calendar or clock). The
534	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2276	difference is that wall clock time can run faster or slower than real
535	+ 10.>) and then reset your system clock to the last year, then it will	2277	time, and time jumps are not uncommon (e.g. when you adjust your
536	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2278	wrist-watch).
537	roughly 10 seconds later and of course not if you reset your system time
538	again).
539		2279
540	They can also be used to implement vastly more complex timers, such as	2280	You can tell a periodic watcher to trigger after some specific point
		2281	in time: for example, if you tell a periodic watcher to trigger "in 10
		2282	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2283	not a delay) and then reset your system clock to January of the previous
		2284	year, then it will take a year or more to trigger the event (unlike an
		2285	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2286	it, as it uses a relative timeout).
		2287
		2288	C<ev_periodic> watchers can also be used to implement vastly more complex
541	triggering an event on eahc midnight, local time.	2289	timers, such as triggering an event on each "midnight, local time", or
		2290	other complicated rules. This cannot easily be done with C<ev_timer>
		2291	watchers, as those cannot react to time jumps.
		2292
		2293	As with timers, the callback is guaranteed to be invoked only when the
		2294	point in time where it is supposed to trigger has passed. If multiple
		2295	timers become ready during the same loop iteration then the ones with
		2296	earlier time-out values are invoked before ones with later time-out values
		2297	(but this is no longer true when a callback calls C<ev_run> recursively).
		2298
		2299	=head3 Watcher-Specific Functions and Data Members
542		2300
543	=over 4	2301	=over 4
544		2302
545	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2303	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
546		2304
547	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2305	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
548		2306
549	Lots of arguments, lets sort it out... There are basically three modes of	2307	Lots of arguments, let's sort it out... There are basically three modes of
550	operation, and we will explain them from simplest to complex:	2308	operation, and we will explain them from simplest to most complex:
551
552		2309
553	=over 4	2310	=over 4
554		2311
555	=item * absolute timer (interval = reschedule_cb = 0)	2312	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
556		2313
557	In this configuration the watcher triggers an event at the wallclock time	2314	In this configuration the watcher triggers an event after the wall clock
558	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2315	time C<offset> has passed. It will not repeat and will not adjust when a
559	that is, if it is to be run at January 1st 2011 then it will run when the	2316	time jump occurs, that is, if it is to be run at January 1st 2011 then it
560	system time reaches or surpasses this time.	2317	will be stopped and invoked when the system clock reaches or surpasses
		2318	this point in time.
561		2319
562	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2320	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
563		2321
564	In this mode the watcher will always be scheduled to time out at the next	2322	In this mode the watcher will always be scheduled to time out at the next
565	C<at + N * interval> time (for some integer N) and then repeat, regardless	2323	C<offset + N * interval> time (for some integer N, which can also be
566	of any time jumps.	2324	negative) and then repeat, regardless of any time jumps. The C<offset>
		2325	argument is merely an offset into the C<interval> periods.
567		2326
568	This can be used to create timers that do not drift with respect to system	2327	This can be used to create timers that do not drift with respect to the
569	time:	2328	system clock, for example, here is an C<ev_periodic> that triggers each
		2329	hour, on the hour (with respect to UTC):
570		2330
571	ev_periodic_set (&periodic, 0., 3600., 0);	2331	ev_periodic_set (&periodic, 0., 3600., 0);
572		2332
573	This doesn't mean there will always be 3600 seconds in between triggers,	2333	This doesn't mean there will always be 3600 seconds in between triggers,
574	but only that the the callback will be called when the system time shows a	2334	but only that the callback will be called when the system time shows a
575	full hour (UTC), or more correctly, when the system time is evenly divisible	2335	full hour (UTC), or more correctly, when the system time is evenly divisible
576	by 3600.	2336	by 3600.
577		2337
578	Another way to think about it (for the mathematically inclined) is that	2338	Another way to think about it (for the mathematically inclined) is that
579	C<ev_periodic> will try to run the callback in this mode at the next possible	2339	C<ev_periodic> will try to run the callback in this mode at the next possible
580	time where C<time = at (mod interval)>, regardless of any time jumps.	2340	time where C<time = offset (mod interval)>, regardless of any time jumps.
581		2341
		2342	The C<interval> I<MUST> be positive, and for numerical stability, the
		2343	interval value should be higher than C<1/8192> (which is around 100
		2344	microseconds) and C<offset> should be higher than C<0> and should have
		2345	at most a similar magnitude as the current time (say, within a factor of
		2346	ten). Typical values for offset are, in fact, C<0> or something between
		2347	C<0> and C<interval>, which is also the recommended range.
		2348
		2349	Note also that there is an upper limit to how often a timer can fire (CPU
		2350	speed for example), so if C<interval> is very small then timing stability
		2351	will of course deteriorate. Libev itself tries to be exact to be about one
		2352	millisecond (if the OS supports it and the machine is fast enough).
		2353
582	=item * manual reschedule mode (reschedule_cb = callback)	2354	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
583		2355
584	In this mode the values for C<interval> and C<at> are both being	2356	In this mode the values for C<interval> and C<offset> are both being
585	ignored. Instead, each time the periodic watcher gets scheduled, the	2357	ignored. Instead, each time the periodic watcher gets scheduled, the
586	reschedule callback will be called with the watcher as first, and the	2358	reschedule callback will be called with the watcher as first, and the
587	current time as second argument.	2359	current time as second argument.
588		2360
589	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2361	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
590	ever, or make any event loop modifications>. If you need to stop it,	2362	or make ANY other event loop modifications whatsoever, unless explicitly
591	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2363	allowed by documentation here>.
592	starting a prepare watcher).
593		2364
		2365	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2366	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2367	only event loop modification you are allowed to do).
		2368
594	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2369	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
595	ev_tstamp now)>, e.g.:	2370	*w, ev_tstamp now)>, e.g.:
596		2371
		2372	static ev_tstamp
597	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2373	my_rescheduler (ev_periodic *w, ev_tstamp now)
598	{	2374	{
599	return now + 60.;	2375	return now + 60.;
600	}	2376	}
601		2377
602	It must return the next time to trigger, based on the passed time value	2378	It must return the next time to trigger, based on the passed time value
603	(that is, the lowest time value larger than to the second argument). It	2379	(that is, the lowest time value larger than to the second argument). It
604	will usually be called just before the callback will be triggered, but	2380	will usually be called just before the callback will be triggered, but
605	might be called at other times, too.	2381	might be called at other times, too.
606		2382
607	NOTE: I<< This callback must always return a time that is later than the	2383	NOTE: I<< This callback must always return a time that is higher than or
608	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2384	equal to the passed C<now> value >>.
609		2385
610	This can be used to create very complex timers, such as a timer that	2386	This can be used to create very complex timers, such as a timer that
611	triggers on each midnight, local time. To do this, you would calculate the	2387	triggers on "next midnight, local time". To do this, you would calculate
612	next midnight after C<now> and return the timestamp value for this. How	2388	the next midnight after C<now> and return the timestamp value for
613	you do this is, again, up to you (but it is not trivial, which is the main	2389	this. Here is a (completely untested, no error checking) example on how to
614	reason I omitted it as an example).	2390	do this:
		2391
		2392	#include <time.h>
		2393
		2394	static ev_tstamp
		2395	my_rescheduler (ev_periodic *w, ev_tstamp now)
		2396	{
		2397	time_t tnow = (time_t)now;
		2398	struct tm tm;
		2399	localtime_r (&tnow, &tm);
		2400
		2401	tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
		2402	++tm.tm_mday; // midnight next day
		2403
		2404	return mktime (&tm);
		2405	}
		2406
		2407	Note: this code might run into trouble on days that have more then two
		2408	midnights (beginning and end).
615		2409
616	=back	2410	=back
617		2411
618	=item ev_periodic_again (loop, ev_periodic *)	2412	=item ev_periodic_again (loop, ev_periodic *)
619		2413
620	Simply stops and restarts the periodic watcher again. This is only useful	2414	Simply stops and restarts the periodic watcher again. This is only useful
621	when you changed some parameters or the reschedule callback would return	2415	when you changed some parameters or the reschedule callback would return
622	a different time than the last time it was called (e.g. in a crond like	2416	a different time than the last time it was called (e.g. in a crond like
623	program when the crontabs have changed).	2417	program when the crontabs have changed).
624		2418
		2419	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2420
		2421	When active, returns the absolute time that the watcher is supposed
		2422	to trigger next. This is not the same as the C<offset> argument to
		2423	C<ev_periodic_set>, but indeed works even in interval and manual
		2424	rescheduling modes.
		2425
		2426	=item ev_tstamp offset [read-write]
		2427
		2428	When repeating, this contains the offset value, otherwise this is the
		2429	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2430	although libev might modify this value for better numerical stability).
		2431
		2432	Can be modified any time, but changes only take effect when the periodic
		2433	timer fires or C<ev_periodic_again> is being called.
		2434
		2435	=item ev_tstamp interval [read-write]
		2436
		2437	The current interval value. Can be modified any time, but changes only
		2438	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2439	called.
		2440
		2441	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2442
		2443	The current reschedule callback, or C<0>, if this functionality is
		2444	switched off. Can be changed any time, but changes only take effect when
		2445	the periodic timer fires or C<ev_periodic_again> is being called.
		2446
625	=back	2447	=back
626		2448
		2449	=head3 Examples
		2450
		2451	Example: Call a callback every hour, or, more precisely, whenever the
		2452	system time is divisible by 3600. The callback invocation times have
		2453	potentially a lot of jitter, but good long-term stability.
		2454
		2455	static void
		2456	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2457	{
		2458	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2459	}
		2460
		2461	ev_periodic hourly_tick;
		2462	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2463	ev_periodic_start (loop, &hourly_tick);
		2464
		2465	Example: The same as above, but use a reschedule callback to do it:
		2466
		2467	#include <math.h>
		2468
		2469	static ev_tstamp
		2470	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2471	{
		2472	return now + (3600. - fmod (now, 3600.));
		2473	}
		2474
		2475	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2476
		2477	Example: Call a callback every hour, starting now:
		2478
		2479	ev_periodic hourly_tick;
		2480	ev_periodic_init (&hourly_tick, clock_cb,
		2481	fmod (ev_now (loop), 3600.), 3600., 0);
		2482	ev_periodic_start (loop, &hourly_tick);
		2483
		2484
627	=head2 C<ev_signal> - signal me when a signal gets signalled	2485	=head2 C<ev_signal> - signal me when a signal gets signalled!
628		2486
629	Signal watchers will trigger an event when the process receives a specific	2487	Signal watchers will trigger an event when the process receives a specific
630	signal one or more times. Even though signals are very asynchronous, libev	2488	signal one or more times. Even though signals are very asynchronous, libev
631	will try it's best to deliver signals synchronously, i.e. as part of the	2489	will try its best to deliver signals synchronously, i.e. as part of the
632	normal event processing, like any other event.	2490	normal event processing, like any other event.
633		2491
		2492	If you want signals to be delivered truly asynchronously, just use
		2493	C<sigaction> as you would do without libev and forget about sharing
		2494	the signal. You can even use C<ev_async> from a signal handler to
		2495	synchronously wake up an event loop.
		2496
634	You can configure as many watchers as you like per signal. Only when the	2497	You can configure as many watchers as you like for the same signal, but
635	first watcher gets started will libev actually register a signal watcher	2498	only within the same loop, i.e. you can watch for C<SIGINT> in your
636	with the kernel (thus it coexists with your own signal handlers as long	2499	default loop and for C<SIGIO> in another loop, but you cannot watch for
637	as you don't register any with libev). Similarly, when the last signal	2500	C<SIGINT> in both the default loop and another loop at the same time. At
638	watcher for a signal is stopped libev will reset the signal handler to	2501	the moment, C<SIGCHLD> is permanently tied to the default loop.
639	SIG_DFL (regardless of what it was set to before).	2502
		2503	Only after the first watcher for a signal is started will libev actually
		2504	register something with the kernel. It thus coexists with your own signal
		2505	handlers as long as you don't register any with libev for the same signal.
		2506
		2507	If possible and supported, libev will install its handlers with
		2508	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2509	not be unduly interrupted. If you have a problem with system calls getting
		2510	interrupted by signals you can block all signals in an C<ev_check> watcher
		2511	and unblock them in an C<ev_prepare> watcher.
		2512
		2513	=head3 The special problem of inheritance over fork/execve/pthread_create
		2514
		2515	Both the signal mask (C<sigprocmask>) and the signal disposition
		2516	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2517	stopping it again), that is, libev might or might not block the signal,
		2518	and might or might not set or restore the installed signal handler (but
		2519	see C<EVFLAG_NOSIGMASK>).
		2520
		2521	While this does not matter for the signal disposition (libev never
		2522	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2523	C<execve>), this matters for the signal mask: many programs do not expect
		2524	certain signals to be blocked.
		2525
		2526	This means that before calling C<exec> (from the child) you should reset
		2527	the signal mask to whatever "default" you expect (all clear is a good
		2528	choice usually).
		2529
		2530	The simplest way to ensure that the signal mask is reset in the child is
		2531	to install a fork handler with C<pthread_atfork> that resets it. That will
		2532	catch fork calls done by libraries (such as the libc) as well.
		2533
		2534	In current versions of libev, the signal will not be blocked indefinitely
		2535	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2536	the window of opportunity for problems, it will not go away, as libev
		2537	I<has> to modify the signal mask, at least temporarily.
		2538
		2539	So I can't stress this enough: I<If you do not reset your signal mask when
		2540	you expect it to be empty, you have a race condition in your code>. This
		2541	is not a libev-specific thing, this is true for most event libraries.
		2542
		2543	=head3 The special problem of threads signal handling
		2544
		2545	POSIX threads has problematic signal handling semantics, specifically,
		2546	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2547	threads in a process block signals, which is hard to achieve.
		2548
		2549	When you want to use sigwait (or mix libev signal handling with your own
		2550	for the same signals), you can tackle this problem by globally blocking
		2551	all signals before creating any threads (or creating them with a fully set
		2552	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2553	loops. Then designate one thread as "signal receiver thread" which handles
		2554	these signals. You can pass on any signals that libev might be interested
		2555	in by calling C<ev_feed_signal>.
		2556
		2557	=head3 Watcher-Specific Functions and Data Members
640		2558
641	=over 4	2559	=over 4
642		2560
643	=item ev_signal_init (ev_signal *, callback, int signum)	2561	=item ev_signal_init (ev_signal *, callback, int signum)
644		2562
645	=item ev_signal_set (ev_signal *, int signum)	2563	=item ev_signal_set (ev_signal *, int signum)
646		2564
647	Configures the watcher to trigger on the given signal number (usually one	2565	Configures the watcher to trigger on the given signal number (usually one
648	of the C<SIGxxx> constants).	2566	of the C<SIGxxx> constants).
649		2567
		2568	=item int signum [read-only]
		2569
		2570	The signal the watcher watches out for.
		2571
650	=back	2572	=back
651		2573
		2574	=head3 Examples
		2575
		2576	Example: Try to exit cleanly on SIGINT.
		2577
		2578	static void
		2579	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2580	{
		2581	ev_break (loop, EVBREAK_ALL);
		2582	}
		2583
		2584	ev_signal signal_watcher;
		2585	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2586	ev_signal_start (loop, &signal_watcher);
		2587
		2588
652	=head2 C<ev_child> - wait for pid status changes	2589	=head2 C<ev_child> - watch out for process status changes
653		2590
654	Child watchers trigger when your process receives a SIGCHLD in response to	2591	Child watchers trigger when your process receives a SIGCHLD in response to
655	some child status changes (most typically when a child of yours dies).	2592	some child status changes (most typically when a child of yours dies or
		2593	exits). It is permissible to install a child watcher I<after> the child
		2594	has been forked (which implies it might have already exited), as long
		2595	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2596	forking and then immediately registering a watcher for the child is fine,
		2597	but forking and registering a watcher a few event loop iterations later or
		2598	in the next callback invocation is not.
		2599
		2600	Only the default event loop is capable of handling signals, and therefore
		2601	you can only register child watchers in the default event loop.
		2602
		2603	Due to some design glitches inside libev, child watchers will always be
		2604	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2605	libev)
		2606
		2607	=head3 Process Interaction
		2608
		2609	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2610	initialised. This is necessary to guarantee proper behaviour even if the
		2611	first child watcher is started after the child exits. The occurrence
		2612	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2613	synchronously as part of the event loop processing. Libev always reaps all
		2614	children, even ones not watched.
		2615
		2616	=head3 Overriding the Built-In Processing
		2617
		2618	Libev offers no special support for overriding the built-in child
		2619	processing, but if your application collides with libev's default child
		2620	handler, you can override it easily by installing your own handler for
		2621	C<SIGCHLD> after initialising the default loop, and making sure the
		2622	default loop never gets destroyed. You are encouraged, however, to use an
		2623	event-based approach to child reaping and thus use libev's support for
		2624	that, so other libev users can use C<ev_child> watchers freely.
		2625
		2626	=head3 Stopping the Child Watcher
		2627
		2628	Currently, the child watcher never gets stopped, even when the
		2629	child terminates, so normally one needs to stop the watcher in the
		2630	callback. Future versions of libev might stop the watcher automatically
		2631	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2632	problem).
		2633
		2634	=head3 Watcher-Specific Functions and Data Members
656		2635
657	=over 4	2636	=over 4
658		2637
659	=item ev_child_init (ev_child *, callback, int pid)	2638	=item ev_child_init (ev_child *, callback, int pid, int trace)
660		2639
661	=item ev_child_set (ev_child *, int pid)	2640	=item ev_child_set (ev_child *, int pid, int trace)
662		2641
663	Configures the watcher to wait for status changes of process C<pid> (or	2642	Configures the watcher to wait for status changes of process C<pid> (or
664	I<any> process if C<pid> is specified as C<0>). The callback can look	2643	I<any> process if C<pid> is specified as C<0>). The callback can look
665	at the C<rstatus> member of the C<ev_child> watcher structure to see	2644	at the C<rstatus> member of the C<ev_child> watcher structure to see
666	the status word (use the macros from C<sys/wait.h> and see your systems	2645	the status word (use the macros from C<sys/wait.h> and see your systems
667	C<waitpid> documentation). The C<rpid> member contains the pid of the	2646	C<waitpid> documentation). The C<rpid> member contains the pid of the
668	process causing the status change.	2647	process causing the status change. C<trace> must be either C<0> (only
		2648	activate the watcher when the process terminates) or C<1> (additionally
		2649	activate the watcher when the process is stopped or continued).
		2650
		2651	=item int pid [read-only]
		2652
		2653	The process id this watcher watches out for, or C<0>, meaning any process id.
		2654
		2655	=item int rpid [read-write]
		2656
		2657	The process id that detected a status change.
		2658
		2659	=item int rstatus [read-write]
		2660
		2661	The process exit/trace status caused by C<rpid> (see your systems
		2662	C<waitpid> and C<sys/wait.h> documentation for details).
669		2663
670	=back	2664	=back
671		2665
		2666	=head3 Examples
		2667
		2668	Example: C<fork()> a new process and install a child handler to wait for
		2669	its completion.
		2670
		2671	ev_child cw;
		2672
		2673	static void
		2674	child_cb (EV_P_ ev_child *w, int revents)
		2675	{
		2676	ev_child_stop (EV_A_ w);
		2677	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2678	}
		2679
		2680	pid_t pid = fork ();
		2681
		2682	if (pid < 0)
		2683	// error
		2684	else if (pid == 0)
		2685	{
		2686	// the forked child executes here
		2687	exit (1);
		2688	}
		2689	else
		2690	{
		2691	ev_child_init (&cw, child_cb, pid, 0);
		2692	ev_child_start (EV_DEFAULT_ &cw);
		2693	}
		2694
		2695
		2696	=head2 C<ev_stat> - did the file attributes just change?
		2697
		2698	This watches a file system path for attribute changes. That is, it calls
		2699	C<stat> on that path in regular intervals (or when the OS says it changed)
		2700	and sees if it changed compared to the last time, invoking the callback
		2701	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2702	happen after the watcher has been started will be reported.
		2703
		2704	The path does not need to exist: changing from "path exists" to "path does
		2705	not exist" is a status change like any other. The condition "path does not
		2706	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2707	C<st_nlink> field being zero (which is otherwise always forced to be at
		2708	least one) and all the other fields of the stat buffer having unspecified
		2709	contents.
		2710
		2711	The path I<must not> end in a slash or contain special components such as
		2712	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2713	your working directory changes, then the behaviour is undefined.
		2714
		2715	Since there is no portable change notification interface available, the
		2716	portable implementation simply calls C<stat(2)> regularly on the path
		2717	to see if it changed somehow. You can specify a recommended polling
		2718	interval for this case. If you specify a polling interval of C<0> (highly
		2719	recommended!) then a I<suitable, unspecified default> value will be used
		2720	(which you can expect to be around five seconds, although this might
		2721	change dynamically). Libev will also impose a minimum interval which is
		2722	currently around C<0.1>, but that's usually overkill.
		2723
		2724	This watcher type is not meant for massive numbers of stat watchers,
		2725	as even with OS-supported change notifications, this can be
		2726	resource-intensive.
		2727
		2728	At the time of this writing, the only OS-specific interface implemented
		2729	is the Linux inotify interface (implementing kqueue support is left as an
		2730	exercise for the reader. Note, however, that the author sees no way of
		2731	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2732
		2733	=head3 ABI Issues (Largefile Support)
		2734
		2735	Libev by default (unless the user overrides this) uses the default
		2736	compilation environment, which means that on systems with large file
		2737	support disabled by default, you get the 32 bit version of the stat
		2738	structure. When using the library from programs that change the ABI to
		2739	use 64 bit file offsets the programs will fail. In that case you have to
		2740	compile libev with the same flags to get binary compatibility. This is
		2741	obviously the case with any flags that change the ABI, but the problem is
		2742	most noticeably displayed with ev_stat and large file support.
		2743
		2744	The solution for this is to lobby your distribution maker to make large
		2745	file interfaces available by default (as e.g. FreeBSD does) and not
		2746	optional. Libev cannot simply switch on large file support because it has
		2747	to exchange stat structures with application programs compiled using the
		2748	default compilation environment.
		2749
		2750	=head3 Inotify and Kqueue
		2751
		2752	When C<inotify (7)> support has been compiled into libev and present at
		2753	runtime, it will be used to speed up change detection where possible. The
		2754	inotify descriptor will be created lazily when the first C<ev_stat>
		2755	watcher is being started.
		2756
		2757	Inotify presence does not change the semantics of C<ev_stat> watchers
		2758	except that changes might be detected earlier, and in some cases, to avoid
		2759	making regular C<stat> calls. Even in the presence of inotify support
		2760	there are many cases where libev has to resort to regular C<stat> polling,
		2761	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2762	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2763	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2764	xfs are fully working) libev usually gets away without polling.
		2765
		2766	There is no support for kqueue, as apparently it cannot be used to
		2767	implement this functionality, due to the requirement of having a file
		2768	descriptor open on the object at all times, and detecting renames, unlinks
		2769	etc. is difficult.
		2770
		2771	=head3 C<stat ()> is a synchronous operation
		2772
		2773	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2774	the process. The exception are C<ev_stat> watchers - those call C<stat
		2775	()>, which is a synchronous operation.
		2776
		2777	For local paths, this usually doesn't matter: unless the system is very
		2778	busy or the intervals between stat's are large, a stat call will be fast,
		2779	as the path data is usually in memory already (except when starting the
		2780	watcher).
		2781
		2782	For networked file systems, calling C<stat ()> can block an indefinite
		2783	time due to network issues, and even under good conditions, a stat call
		2784	often takes multiple milliseconds.
		2785
		2786	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2787	paths, although this is fully supported by libev.
		2788
		2789	=head3 The special problem of stat time resolution
		2790
		2791	The C<stat ()> system call only supports full-second resolution portably,
		2792	and even on systems where the resolution is higher, most file systems
		2793	still only support whole seconds.
		2794
		2795	That means that, if the time is the only thing that changes, you can
		2796	easily miss updates: on the first update, C<ev_stat> detects a change and
		2797	calls your callback, which does something. When there is another update
		2798	within the same second, C<ev_stat> will be unable to detect unless the
		2799	stat data does change in other ways (e.g. file size).
		2800
		2801	The solution to this is to delay acting on a change for slightly more
		2802	than a second (or till slightly after the next full second boundary), using
		2803	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2804	ev_timer_again (loop, w)>).
		2805
		2806	The C<.02> offset is added to work around small timing inconsistencies
		2807	of some operating systems (where the second counter of the current time
		2808	might be be delayed. One such system is the Linux kernel, where a call to
		2809	C<gettimeofday> might return a timestamp with a full second later than
		2810	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2811	update file times then there will be a small window where the kernel uses
		2812	the previous second to update file times but libev might already execute
		2813	the timer callback).
		2814
		2815	=head3 Watcher-Specific Functions and Data Members
		2816
		2817	=over 4
		2818
		2819	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2820
		2821	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2822
		2823	Configures the watcher to wait for status changes of the given
		2824	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2825	be detected and should normally be specified as C<0> to let libev choose
		2826	a suitable value. The memory pointed to by C<path> must point to the same
		2827	path for as long as the watcher is active.
		2828
		2829	The callback will receive an C<EV_STAT> event when a change was detected,
		2830	relative to the attributes at the time the watcher was started (or the
		2831	last change was detected).
		2832
		2833	=item ev_stat_stat (loop, ev_stat *)
		2834
		2835	Updates the stat buffer immediately with new values. If you change the
		2836	watched path in your callback, you could call this function to avoid
		2837	detecting this change (while introducing a race condition if you are not
		2838	the only one changing the path). Can also be useful simply to find out the
		2839	new values.
		2840
		2841	=item ev_statdata attr [read-only]
		2842
		2843	The most-recently detected attributes of the file. Although the type is
		2844	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2845	suitable for your system, but you can only rely on the POSIX-standardised
		2846	members to be present. If the C<st_nlink> member is C<0>, then there was
		2847	some error while C<stat>ing the file.
		2848
		2849	=item ev_statdata prev [read-only]
		2850
		2851	The previous attributes of the file. The callback gets invoked whenever
		2852	C<prev> != C<attr>, or, more precisely, one or more of these members
		2853	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2854	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2855
		2856	=item ev_tstamp interval [read-only]
		2857
		2858	The specified interval.
		2859
		2860	=item const char *path [read-only]
		2861
		2862	The file system path that is being watched.
		2863
		2864	=back
		2865
		2866	=head3 Examples
		2867
		2868	Example: Watch C</etc/passwd> for attribute changes.
		2869
		2870	static void
		2871	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2872	{
		2873	/* /etc/passwd changed in some way */
		2874	if (w->attr.st_nlink)
		2875	{
		2876	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2877	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2878	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2879	}
		2880	else
		2881	/* you shalt not abuse printf for puts */
		2882	puts ("wow, /etc/passwd is not there, expect problems. "
		2883	"if this is windows, they already arrived\n");
		2884	}
		2885
		2886	...
		2887	ev_stat passwd;
		2888
		2889	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2890	ev_stat_start (loop, &passwd);
		2891
		2892	Example: Like above, but additionally use a one-second delay so we do not
		2893	miss updates (however, frequent updates will delay processing, too, so
		2894	one might do the work both on C<ev_stat> callback invocation I<and> on
		2895	C<ev_timer> callback invocation).
		2896
		2897	static ev_stat passwd;
		2898	static ev_timer timer;
		2899
		2900	static void
		2901	timer_cb (EV_P_ ev_timer *w, int revents)
		2902	{
		2903	ev_timer_stop (EV_A_ w);
		2904
		2905	/* now it's one second after the most recent passwd change */
		2906	}
		2907
		2908	static void
		2909	stat_cb (EV_P_ ev_stat *w, int revents)
		2910	{
		2911	/* reset the one-second timer */
		2912	ev_timer_again (EV_A_ &timer);
		2913	}
		2914
		2915	...
		2916	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2917	ev_stat_start (loop, &passwd);
		2918	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2919
		2920
672	=head2 C<ev_idle> - when you've got nothing better to do	2921	=head2 C<ev_idle> - when you've got nothing better to do...
673		2922
674	Idle watchers trigger events when there are no other events are pending	2923	Idle watchers trigger events when no other events of the same or higher
675	(prepare, check and other idle watchers do not count). That is, as long	2924	priority are pending (prepare, check and other idle watchers do not count
676	as your process is busy handling sockets or timeouts (or even signals,	2925	as receiving "events").
677	imagine) it will not be triggered. But when your process is idle all idle	2926
678	watchers are being called again and again, once per event loop iteration -	2927	That is, as long as your process is busy handling sockets or timeouts
		2928	(or even signals, imagine) of the same or higher priority it will not be
		2929	triggered. But when your process is idle (or only lower-priority watchers
		2930	are pending), the idle watchers are being called once per event loop
679	until stopped, that is, or your process receives more events and becomes	2931	iteration - until stopped, that is, or your process receives more events
680	busy.	2932	and becomes busy again with higher priority stuff.
681		2933
682	The most noteworthy effect is that as long as any idle watchers are	2934	The most noteworthy effect is that as long as any idle watchers are
683	active, the process will not block when waiting for new events.	2935	active, the process will not block when waiting for new events.
684		2936
685	Apart from keeping your process non-blocking (which is a useful	2937	Apart from keeping your process non-blocking (which is a useful
686	effect on its own sometimes), idle watchers are a good place to do	2938	effect on its own sometimes), idle watchers are a good place to do
687	"pseudo-background processing", or delay processing stuff to after the	2939	"pseudo-background processing", or delay processing stuff to after the
688	event loop has handled all outstanding events.	2940	event loop has handled all outstanding events.
689		2941
		2942	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2943
		2944	As long as there is at least one active idle watcher, libev will never
		2945	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2946	For this to work, the idle watcher doesn't need to be invoked at all - the
		2947	lowest priority will do.
		2948
		2949	This mode of operation can be useful together with an C<ev_check> watcher,
		2950	to do something on each event loop iteration - for example to balance load
		2951	between different connections.
		2952
		2953	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2954	example.
		2955
		2956	=head3 Watcher-Specific Functions and Data Members
		2957
690	=over 4	2958	=over 4
691		2959
692	=item ev_idle_init (ev_signal *, callback)	2960	=item ev_idle_init (ev_idle *, callback)
693		2961
694	Initialises and configures the idle watcher - it has no parameters of any	2962	Initialises and configures the idle watcher - it has no parameters of any
695	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2963	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
696	believe me.	2964	believe me.
697		2965
698	=back	2966	=back
699		2967
		2968	=head3 Examples
		2969
		2970	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2971	callback, free it. Also, use no error checking, as usual.
		2972
		2973	static void
		2974	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2975	{
		2976	// stop the watcher
		2977	ev_idle_stop (loop, w);
		2978
		2979	// now we can free it
		2980	free (w);
		2981
		2982	// now do something you wanted to do when the program has
		2983	// no longer anything immediate to do.
		2984	}
		2985
		2986	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2987	ev_idle_init (idle_watcher, idle_cb);
		2988	ev_idle_start (loop, idle_watcher);
		2989
		2990
700	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2991	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
701		2992
702	Prepare and check watchers are usually (but not always) used in tandem:	2993	Prepare and check watchers are often (but not always) used in pairs:
703	prepare watchers get invoked before the process blocks and check watchers	2994	prepare watchers get invoked before the process blocks and check watchers
704	afterwards.	2995	afterwards.
705		2996
		2997	You I<must not> call C<ev_run> (or similar functions that enter the
		2998	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
		2999	C<ev_check> watchers. Other loops than the current one are fine,
		3000	however. The rationale behind this is that you do not need to check
		3001	for recursion in those watchers, i.e. the sequence will always be
		3002	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
		3003	kind they will always be called in pairs bracketing the blocking call.
		3004
706	Their main purpose is to integrate other event mechanisms into libev. This	3005	Their main purpose is to integrate other event mechanisms into libev and
707	could be used, for example, to track variable changes, implement your own	3006	their use is somewhat advanced. They could be used, for example, to track
708	watchers, integrate net-snmp or a coroutine library and lots more.	3007	variable changes, implement your own watchers, integrate net-snmp or a
		3008	coroutine library and lots more. They are also occasionally useful if
		3009	you cache some data and want to flush it before blocking (for example,
		3010	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		3011	watcher).
709		3012
710	This is done by examining in each prepare call which file descriptors need	3013	This is done by examining in each prepare call which file descriptors
711	to be watched by the other library, registering C<ev_io> watchers for	3014	need to be watched by the other library, registering C<ev_io> watchers
712	them and starting an C<ev_timer> watcher for any timeouts (many libraries	3015	for them and starting an C<ev_timer> watcher for any timeouts (many
713	provide just this functionality). Then, in the check watcher you check for	3016	libraries provide exactly this functionality). Then, in the check watcher,
714	any events that occured (by checking the pending status of all watchers	3017	you check for any events that occurred (by checking the pending status
715	and stopping them) and call back into the library. The I/O and timer	3018	of all watchers and stopping them) and call back into the library. The
716	callbacks will never actually be called (but must be valid nevertheless,	3019	I/O and timer callbacks will never actually be called (but must be valid
717	because you never know, you know?).	3020	nevertheless, because you never know, you know?).
718		3021
719	As another example, the Perl Coro module uses these hooks to integrate	3022	As another example, the Perl Coro module uses these hooks to integrate
720	coroutines into libev programs, by yielding to other active coroutines	3023	coroutines into libev programs, by yielding to other active coroutines
721	during each prepare and only letting the process block if no coroutines	3024	during each prepare and only letting the process block if no coroutines
722	are ready to run (it's actually more complicated: it only runs coroutines	3025	are ready to run (it's actually more complicated: it only runs coroutines
723	with priority higher than or equal to the event loop and one coroutine	3026	with priority higher than or equal to the event loop and one coroutine
724	of lower priority, but only once, using idle watchers to keep the event	3027	of lower priority, but only once, using idle watchers to keep the event
725	loop from blocking if lower-priority coroutines are active, thus mapping	3028	loop from blocking if lower-priority coroutines are active, thus mapping
726	low-priority coroutines to idle/background tasks).	3029	low-priority coroutines to idle/background tasks).
727		3030
		3031	When used for this purpose, it is recommended to give C<ev_check> watchers
		3032	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
		3033	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		3034	watchers).
		3035
		3036	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		3037	activate ("feed") events into libev. While libev fully supports this, they
		3038	might get executed before other C<ev_check> watchers did their job. As
		3039	C<ev_check> watchers are often used to embed other (non-libev) event
		3040	loops those other event loops might be in an unusable state until their
		3041	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		3042	others).
		3043
		3044	=head3 Abusing an C<ev_check> watcher for its side-effect
		3045
		3046	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		3047	useful because they are called once per event loop iteration. For
		3048	example, if you want to handle a large number of connections fairly, you
		3049	normally only do a bit of work for each active connection, and if there
		3050	is more work to do, you wait for the next event loop iteration, so other
		3051	connections have a chance of making progress.
		3052
		3053	Using an C<ev_check> watcher is almost enough: it will be called on the
		3054	next event loop iteration. However, that isn't as soon as possible -
		3055	without external events, your C<ev_check> watcher will not be invoked.
		3056
		3057	This is where C<ev_idle> watchers come in handy - all you need is a
		3058	single global idle watcher that is active as long as you have one active
		3059	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		3060	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		3061	invoked. Neither watcher alone can do that.
		3062
		3063	=head3 Watcher-Specific Functions and Data Members
		3064
728	=over 4	3065	=over 4
729		3066
730	=item ev_prepare_init (ev_prepare *, callback)	3067	=item ev_prepare_init (ev_prepare *, callback)
731		3068
732	=item ev_check_init (ev_check *, callback)	3069	=item ev_check_init (ev_check *, callback)
733		3070
734	Initialises and configures the prepare or check watcher - they have no	3071	Initialises and configures the prepare or check watcher - they have no
735	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	3072	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
736	macros, but using them is utterly, utterly and completely pointless.	3073	macros, but using them is utterly, utterly, utterly and completely
		3074	pointless.
737		3075
738	=back	3076	=back
739		3077
		3078	=head3 Examples
		3079
		3080	There are a number of principal ways to embed other event loops or modules
		3081	into libev. Here are some ideas on how to include libadns into libev
		3082	(there is a Perl module named C<EV::ADNS> that does this, which you could
		3083	use as a working example. Another Perl module named C<EV::Glib> embeds a
		3084	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		3085	Glib event loop).
		3086
		3087	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		3088	and in a check watcher, destroy them and call into libadns. What follows
		3089	is pseudo-code only of course. This requires you to either use a low
		3090	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		3091	the callbacks for the IO/timeout watchers might not have been called yet.
		3092
		3093	static ev_io iow [nfd];
		3094	static ev_timer tw;
		3095
		3096	static void
		3097	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		3098	{
		3099	}
		3100
		3101	// create io watchers for each fd and a timer before blocking
		3102	static void
		3103	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		3104	{
		3105	int timeout = 3600000;
		3106	struct pollfd fds [nfd];
		3107	// actual code will need to loop here and realloc etc.
		3108	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		3109
		3110	/* the callback is illegal, but won't be called as we stop during check */
		3111	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		3112	ev_timer_start (loop, &tw);
		3113
		3114	// create one ev_io per pollfd
		3115	for (int i = 0; i < nfd; ++i)
		3116	{
		3117	ev_io_init (iow + i, io_cb, fds [i].fd,
		3118	((fds [i].events & POLLIN ? EV_READ : 0)
		3119	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		3120
		3121	fds [i].revents = 0;
		3122	ev_io_start (loop, iow + i);
		3123	}
		3124	}
		3125
		3126	// stop all watchers after blocking
		3127	static void
		3128	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		3129	{
		3130	ev_timer_stop (loop, &tw);
		3131
		3132	for (int i = 0; i < nfd; ++i)
		3133	{
		3134	// set the relevant poll flags
		3135	// could also call adns_processreadable etc. here
		3136	struct pollfd *fd = fds + i;
		3137	int revents = ev_clear_pending (iow + i);
		3138	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		3139	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		3140
		3141	// now stop the watcher
		3142	ev_io_stop (loop, iow + i);
		3143	}
		3144
		3145	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		3146	}
		3147
		3148	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		3149	in the prepare watcher and would dispose of the check watcher.
		3150
		3151	Method 3: If the module to be embedded supports explicit event
		3152	notification (libadns does), you can also make use of the actual watcher
		3153	callbacks, and only destroy/create the watchers in the prepare watcher.
		3154
		3155	static void
		3156	timer_cb (EV_P_ ev_timer *w, int revents)
		3157	{
		3158	adns_state ads = (adns_state)w->data;
		3159	update_now (EV_A);
		3160
		3161	adns_processtimeouts (ads, &tv_now);
		3162	}
		3163
		3164	static void
		3165	io_cb (EV_P_ ev_io *w, int revents)
		3166	{
		3167	adns_state ads = (adns_state)w->data;
		3168	update_now (EV_A);
		3169
		3170	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		3171	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		3172	}
		3173
		3174	// do not ever call adns_afterpoll
		3175
		3176	Method 4: Do not use a prepare or check watcher because the module you
		3177	want to embed is not flexible enough to support it. Instead, you can
		3178	override their poll function. The drawback with this solution is that the
		3179	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		3180	this approach, effectively embedding EV as a client into the horrible
		3181	libglib event loop.
		3182
		3183	static gint
		3184	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		3185	{
		3186	int got_events = 0;
		3187
		3188	for (n = 0; n < nfds; ++n)
		3189	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		3190
		3191	if (timeout >= 0)
		3192	// create/start timer
		3193
		3194	// poll
		3195	ev_run (EV_A_ 0);
		3196
		3197	// stop timer again
		3198	if (timeout >= 0)
		3199	ev_timer_stop (EV_A_ &to);
		3200
		3201	// stop io watchers again - their callbacks should have set
		3202	for (n = 0; n < nfds; ++n)
		3203	ev_io_stop (EV_A_ iow [n]);
		3204
		3205	return got_events;
		3206	}
		3207
		3208
		3209	=head2 C<ev_embed> - when one backend isn't enough...
		3210
		3211	This is a rather advanced watcher type that lets you embed one event loop
		3212	into another (currently only C<ev_io> events are supported in the embedded
		3213	loop, other types of watchers might be handled in a delayed or incorrect
		3214	fashion and must not be used).
		3215
		3216	There are primarily two reasons you would want that: work around bugs and
		3217	prioritise I/O.
		3218
		3219	As an example for a bug workaround, the kqueue backend might only support
		3220	sockets on some platform, so it is unusable as generic backend, but you
		3221	still want to make use of it because you have many sockets and it scales
		3222	so nicely. In this case, you would create a kqueue-based loop and embed
		3223	it into your default loop (which might use e.g. poll). Overall operation
		3224	will be a bit slower because first libev has to call C<poll> and then
		3225	C<kevent>, but at least you can use both mechanisms for what they are
		3226	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		3227
		3228	As for prioritising I/O: under rare circumstances you have the case where
		3229	some fds have to be watched and handled very quickly (with low latency),
		3230	and even priorities and idle watchers might have too much overhead. In
		3231	this case you would put all the high priority stuff in one loop and all
		3232	the rest in a second one, and embed the second one in the first.
		3233
		3234	As long as the watcher is active, the callback will be invoked every
		3235	time there might be events pending in the embedded loop. The callback
		3236	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		3237	sweep and invoke their callbacks (the callback doesn't need to invoke the
		3238	C<ev_embed_sweep> function directly, it could also start an idle watcher
		3239	to give the embedded loop strictly lower priority for example).
		3240
		3241	You can also set the callback to C<0>, in which case the embed watcher
		3242	will automatically execute the embedded loop sweep whenever necessary.
		3243
		3244	Fork detection will be handled transparently while the C<ev_embed> watcher
		3245	is active, i.e., the embedded loop will automatically be forked when the
		3246	embedding loop forks. In other cases, the user is responsible for calling
		3247	C<ev_loop_fork> on the embedded loop.
		3248
		3249	Unfortunately, not all backends are embeddable: only the ones returned by
		3250	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		3251	portable one.
		3252
		3253	So when you want to use this feature you will always have to be prepared
		3254	that you cannot get an embeddable loop. The recommended way to get around
		3255	this is to have a separate variables for your embeddable loop, try to
		3256	create it, and if that fails, use the normal loop for everything.
		3257
		3258	=head3 C<ev_embed> and fork
		3259
		3260	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3261	automatically be applied to the embedded loop as well, so no special
		3262	fork handling is required in that case. When the watcher is not running,
		3263	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3264	as applicable.
		3265
		3266	=head3 Watcher-Specific Functions and Data Members
		3267
		3268	=over 4
		3269
		3270	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		3271
		3272	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
		3273
		3274	Configures the watcher to embed the given loop, which must be
		3275	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		3276	invoked automatically, otherwise it is the responsibility of the callback
		3277	to invoke it (it will continue to be called until the sweep has been done,
		3278	if you do not want that, you need to temporarily stop the embed watcher).
		3279
		3280	=item ev_embed_sweep (loop, ev_embed *)
		3281
		3282	Make a single, non-blocking sweep over the embedded loop. This works
		3283	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3284	appropriate way for embedded loops.
		3285
		3286	=item struct ev_loop *other [read-only]
		3287
		3288	The embedded event loop.
		3289
		3290	=back
		3291
		3292	=head3 Examples
		3293
		3294	Example: Try to get an embeddable event loop and embed it into the default
		3295	event loop. If that is not possible, use the default loop. The default
		3296	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3297	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3298	used).
		3299
		3300	struct ev_loop *loop_hi = ev_default_init (0);
		3301	struct ev_loop *loop_lo = 0;
		3302	ev_embed embed;
		3303
		3304	// see if there is a chance of getting one that works
		3305	// (remember that a flags value of 0 means autodetection)
		3306	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3307	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3308	: 0;
		3309
		3310	// if we got one, then embed it, otherwise default to loop_hi
		3311	if (loop_lo)
		3312	{
		3313	ev_embed_init (&embed, 0, loop_lo);
		3314	ev_embed_start (loop_hi, &embed);
		3315	}
		3316	else
		3317	loop_lo = loop_hi;
		3318
		3319	Example: Check if kqueue is available but not recommended and create
		3320	a kqueue backend for use with sockets (which usually work with any
		3321	kqueue implementation). Store the kqueue/socket-only event loop in
		3322	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3323
		3324	struct ev_loop *loop = ev_default_init (0);
		3325	struct ev_loop *loop_socket = 0;
		3326	ev_embed embed;
		3327
		3328	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3329	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3330	{
		3331	ev_embed_init (&embed, 0, loop_socket);
		3332	ev_embed_start (loop, &embed);
		3333	}
		3334
		3335	if (!loop_socket)
		3336	loop_socket = loop;
		3337
		3338	// now use loop_socket for all sockets, and loop for everything else
		3339
		3340
		3341	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3342
		3343	Fork watchers are called when a C<fork ()> was detected (usually because
		3344	whoever is a good citizen cared to tell libev about it by calling
		3345	C<ev_loop_fork>). The invocation is done before the event loop blocks next
		3346	and before C<ev_check> watchers are being called, and only in the child
		3347	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
		3348	and calls it in the wrong process, the fork handlers will be invoked, too,
		3349	of course.
		3350
		3351	=head3 The special problem of life after fork - how is it possible?
		3352
		3353	Most uses of C<fork ()> consist of forking, then some simple calls to set
		3354	up/change the process environment, followed by a call to C<exec()>. This
		3355	sequence should be handled by libev without any problems.
		3356
		3357	This changes when the application actually wants to do event handling
		3358	in the child, or both parent in child, in effect "continuing" after the
		3359	fork.
		3360
		3361	The default mode of operation (for libev, with application help to detect
		3362	forks) is to duplicate all the state in the child, as would be expected
		3363	when I<either> the parent I<or> the child process continues.
		3364
		3365	When both processes want to continue using libev, then this is usually the
		3366	wrong result. In that case, usually one process (typically the parent) is
		3367	supposed to continue with all watchers in place as before, while the other
		3368	process typically wants to start fresh, i.e. without any active watchers.
		3369
		3370	The cleanest and most efficient way to achieve that with libev is to
		3371	simply create a new event loop, which of course will be "empty", and
		3372	use that for new watchers. This has the advantage of not touching more
		3373	memory than necessary, and thus avoiding the copy-on-write, and the
		3374	disadvantage of having to use multiple event loops (which do not support
		3375	signal watchers).
		3376
		3377	When this is not possible, or you want to use the default loop for
		3378	other reasons, then in the process that wants to start "fresh", call
		3379	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3380	Destroying the default loop will "orphan" (not stop) all registered
		3381	watchers, so you have to be careful not to execute code that modifies
		3382	those watchers. Note also that in that case, you have to re-register any
		3383	signal watchers.
		3384
		3385	=head3 Watcher-Specific Functions and Data Members
		3386
		3387	=over 4
		3388
		3389	=item ev_fork_init (ev_fork *, callback)
		3390
		3391	Initialises and configures the fork watcher - it has no parameters of any
		3392	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3393	really.
		3394
		3395	=back
		3396
		3397
		3398	=head2 C<ev_cleanup> - even the best things end
		3399
		3400	Cleanup watchers are called just before the event loop is being destroyed
		3401	by a call to C<ev_loop_destroy>.
		3402
		3403	While there is no guarantee that the event loop gets destroyed, cleanup
		3404	watchers provide a convenient method to install cleanup hooks for your
		3405	program, worker threads and so on - you just to make sure to destroy the
		3406	loop when you want them to be invoked.
		3407
		3408	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3409	all other watchers, they do not keep a reference to the event loop (which
		3410	makes a lot of sense if you think about it). Like all other watchers, you
		3411	can call libev functions in the callback, except C<ev_cleanup_start>.
		3412
		3413	=head3 Watcher-Specific Functions and Data Members
		3414
		3415	=over 4
		3416
		3417	=item ev_cleanup_init (ev_cleanup *, callback)
		3418
		3419	Initialises and configures the cleanup watcher - it has no parameters of
		3420	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3421	pointless, I assure you.
		3422
		3423	=back
		3424
		3425	Example: Register an atexit handler to destroy the default loop, so any
		3426	cleanup functions are called.
		3427
		3428	static void
		3429	program_exits (void)
		3430	{
		3431	ev_loop_destroy (EV_DEFAULT_UC);
		3432	}
		3433
		3434	...
		3435	atexit (program_exits);
		3436
		3437
		3438	=head2 C<ev_async> - how to wake up an event loop
		3439
		3440	In general, you cannot use an C<ev_loop> from multiple threads or other
		3441	asynchronous sources such as signal handlers (as opposed to multiple event
		3442	loops - those are of course safe to use in different threads).
		3443
		3444	Sometimes, however, you need to wake up an event loop you do not control,
		3445	for example because it belongs to another thread. This is what C<ev_async>
		3446	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3447	it by calling C<ev_async_send>, which is thread- and signal safe.
		3448
		3449	This functionality is very similar to C<ev_signal> watchers, as signals,
		3450	too, are asynchronous in nature, and signals, too, will be compressed
		3451	(i.e. the number of callback invocations may be less than the number of
		3452	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
		3453	of "global async watchers" by using a watcher on an otherwise unused
		3454	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3455	even without knowing which loop owns the signal.
		3456
		3457	=head3 Queueing
		3458
		3459	C<ev_async> does not support queueing of data in any way. The reason
		3460	is that the author does not know of a simple (or any) algorithm for a
		3461	multiple-writer-single-reader queue that works in all cases and doesn't
		3462	need elaborate support such as pthreads or unportable memory access
		3463	semantics.
		3464
		3465	That means that if you want to queue data, you have to provide your own
		3466	queue. But at least I can tell you how to implement locking around your
		3467	queue:
		3468
		3469	=over 4
		3470
		3471	=item queueing from a signal handler context
		3472
		3473	To implement race-free queueing, you simply add to the queue in the signal
		3474	handler but you block the signal handler in the watcher callback. Here is
		3475	an example that does that for some fictitious SIGUSR1 handler:
		3476
		3477	static ev_async mysig;
		3478
		3479	static void
		3480	sigusr1_handler (void)
		3481	{
		3482	sometype data;
		3483
		3484	// no locking etc.
		3485	queue_put (data);
		3486	ev_async_send (EV_DEFAULT_ &mysig);
		3487	}
		3488
		3489	static void
		3490	mysig_cb (EV_P_ ev_async *w, int revents)
		3491	{
		3492	sometype data;
		3493	sigset_t block, prev;
		3494
		3495	sigemptyset (&block);
		3496	sigaddset (&block, SIGUSR1);
		3497	sigprocmask (SIG_BLOCK, &block, &prev);
		3498
		3499	while (queue_get (&data))
		3500	process (data);
		3501
		3502	if (sigismember (&prev, SIGUSR1)
		3503	sigprocmask (SIG_UNBLOCK, &block, 0);
		3504	}
		3505
		3506	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3507	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3508	either...).
		3509
		3510	=item queueing from a thread context
		3511
		3512	The strategy for threads is different, as you cannot (easily) block
		3513	threads but you can easily preempt them, so to queue safely you need to
		3514	employ a traditional mutex lock, such as in this pthread example:
		3515
		3516	static ev_async mysig;
		3517	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3518
		3519	static void
		3520	otherthread (void)
		3521	{
		3522	// only need to lock the actual queueing operation
		3523	pthread_mutex_lock (&mymutex);
		3524	queue_put (data);
		3525	pthread_mutex_unlock (&mymutex);
		3526
		3527	ev_async_send (EV_DEFAULT_ &mysig);
		3528	}
		3529
		3530	static void
		3531	mysig_cb (EV_P_ ev_async *w, int revents)
		3532	{
		3533	pthread_mutex_lock (&mymutex);
		3534
		3535	while (queue_get (&data))
		3536	process (data);
		3537
		3538	pthread_mutex_unlock (&mymutex);
		3539	}
		3540
		3541	=back
		3542
		3543
		3544	=head3 Watcher-Specific Functions and Data Members
		3545
		3546	=over 4
		3547
		3548	=item ev_async_init (ev_async *, callback)
		3549
		3550	Initialises and configures the async watcher - it has no parameters of any
		3551	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3552	trust me.
		3553
		3554	=item ev_async_send (loop, ev_async *)
		3555
		3556	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3557	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3558	returns.
		3559
		3560	Unlike C<ev_feed_event>, this call is safe to do from other threads,
		3561	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
		3562	embedding section below on what exactly this means).
		3563
		3564	Note that, as with other watchers in libev, multiple events might get
		3565	compressed into a single callback invocation (another way to look at
		3566	this is that C<ev_async> watchers are level-triggered: they are set on
		3567	C<ev_async_send>, reset when the event loop detects that).
		3568
		3569	This call incurs the overhead of at most one extra system call per event
		3570	loop iteration, if the event loop is blocked, and no syscall at all if
		3571	the event loop (or your program) is processing events. That means that
		3572	repeated calls are basically free (there is no need to avoid calls for
		3573	performance reasons) and that the overhead becomes smaller (typically
		3574	zero) under load.
		3575
		3576	=item bool = ev_async_pending (ev_async *)
		3577
		3578	Returns a non-zero value when C<ev_async_send> has been called on the
		3579	watcher but the event has not yet been processed (or even noted) by the
		3580	event loop.
		3581
		3582	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3583	the loop iterates next and checks for the watcher to have become active,
		3584	it will reset the flag again. C<ev_async_pending> can be used to very
		3585	quickly check whether invoking the loop might be a good idea.
		3586
		3587	Not that this does I<not> check whether the watcher itself is pending,
		3588	only whether it has been requested to make this watcher pending: there
		3589	is a time window between the event loop checking and resetting the async
		3590	notification, and the callback being invoked.
		3591
		3592	=back
		3593
		3594
740	=head1 OTHER FUNCTIONS	3595	=head1 OTHER FUNCTIONS
741		3596
742	There are some other functions of possible interest. Described. Here. Now.	3597	There are some other functions of possible interest. Described. Here. Now.
743		3598
744	=over 4	3599	=over 4
745		3600
746	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3601	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
747		3602
748	This function combines a simple timer and an I/O watcher, calls your	3603	This function combines a simple timer and an I/O watcher, calls your
749	callback on whichever event happens first and automatically stop both	3604	callback on whichever event happens first and automatically stops both
750	watchers. This is useful if you want to wait for a single event on an fd	3605	watchers. This is useful if you want to wait for a single event on an fd
751	or timeout without having to allocate/configure/start/stop/free one or	3606	or timeout without having to allocate/configure/start/stop/free one or
752	more watchers yourself.	3607	more watchers yourself.
753		3608
754	If C<fd> is less than 0, then no I/O watcher will be started and events	3609	If C<fd> is less than 0, then no I/O watcher will be started and the
755	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3610	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
756	C<events> set will be craeted and started.	3611	the given C<fd> and C<events> set will be created and started.
757		3612
758	If C<timeout> is less than 0, then no timeout watcher will be	3613	If C<timeout> is less than 0, then no timeout watcher will be
759	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3614	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
760	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3615	repeat = 0) will be started. C<0> is a valid timeout.
761	dubious value.
762		3616
763	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3617	The callback has the type C<void (*cb)(int revents, void *arg)> and is
764	passed an C<revents> set like normal event callbacks (a combination of	3618	passed an C<revents> set like normal event callbacks (a combination of
765	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3619	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
766	value passed to C<ev_once>:	3620	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3621	a timeout and an io event at the same time - you probably should give io
		3622	events precedence.
767		3623
		3624	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3625
768	static void stdin_ready (int revents, void *arg)	3626	static void stdin_ready (int revents, void *arg)
		3627	{
		3628	if (revents & EV_READ)
		3629	/* stdin might have data for us, joy! */;
		3630	else if (revents & EV_TIMER)
		3631	/* doh, nothing entered */;
		3632	}
		3633
		3634	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
		3635
		3636	=item ev_feed_fd_event (loop, int fd, int revents)
		3637
		3638	Feed an event on the given fd, as if a file descriptor backend detected
		3639	the given events.
		3640
		3641	=item ev_feed_signal_event (loop, int signum)
		3642
		3643	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3644	which is async-safe.
		3645
		3646	=back
		3647
		3648
		3649	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3650
		3651	This section explains some common idioms that are not immediately
		3652	obvious. Note that examples are sprinkled over the whole manual, and this
		3653	section only contains stuff that wouldn't fit anywhere else.
		3654
		3655	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3656
		3657	Each watcher has, by default, a C<void *data> member that you can read
		3658	or modify at any time: libev will completely ignore it. This can be used
		3659	to associate arbitrary data with your watcher. If you need more data and
		3660	don't want to allocate memory separately and store a pointer to it in that
		3661	data member, you can also "subclass" the watcher type and provide your own
		3662	data:
		3663
		3664	struct my_io
		3665	{
		3666	ev_io io;
		3667	int otherfd;
		3668	void *somedata;
		3669	struct whatever *mostinteresting;
		3670	};
		3671
		3672	...
		3673	struct my_io w;
		3674	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3675
		3676	And since your callback will be called with a pointer to the watcher, you
		3677	can cast it back to your own type:
		3678
		3679	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3680	{
		3681	struct my_io *w = (struct my_io *)w_;
		3682	...
		3683	}
		3684
		3685	More interesting and less C-conformant ways of casting your callback
		3686	function type instead have been omitted.
		3687
		3688	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3689
		3690	Another common scenario is to use some data structure with multiple
		3691	embedded watchers, in effect creating your own watcher that combines
		3692	multiple libev event sources into one "super-watcher":
		3693
		3694	struct my_biggy
		3695	{
		3696	int some_data;
		3697	ev_timer t1;
		3698	ev_timer t2;
		3699	}
		3700
		3701	In this case getting the pointer to C<my_biggy> is a bit more
		3702	complicated: Either you store the address of your C<my_biggy> struct in
		3703	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3704	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3705	real programmers):
		3706
		3707	#include <stddef.h>
		3708
		3709	static void
		3710	t1_cb (EV_P_ ev_timer *w, int revents)
		3711	{
		3712	struct my_biggy big = (struct my_biggy *)
		3713	(((char *)w) - offsetof (struct my_biggy, t1));
		3714	}
		3715
		3716	static void
		3717	t2_cb (EV_P_ ev_timer *w, int revents)
		3718	{
		3719	struct my_biggy big = (struct my_biggy *)
		3720	(((char *)w) - offsetof (struct my_biggy, t2));
		3721	}
		3722
		3723	=head2 AVOIDING FINISHING BEFORE RETURNING
		3724
		3725	Often you have structures like this in event-based programs:
		3726
		3727	callback ()
769	{	3728	{
770	if (revents & EV_TIMEOUT)	3729	free (request);
771	/* doh, nothing entered */;
772	else if (revents & EV_READ)
773	/* stdin might have data for us, joy! */;
774	}	3730	}
775		3731
776	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3732	request = start_new_request (..., callback);
777		3733
778	=item ev_feed_event (loop, watcher, int events)	3734	The intent is to start some "lengthy" operation. The C<request> could be
		3735	used to cancel the operation, or do other things with it.
779		3736
780	Feeds the given event set into the event loop, as if the specified event	3737	It's not uncommon to have code paths in C<start_new_request> that
781	had happened for the specified watcher (which must be a pointer to an	3738	immediately invoke the callback, for example, to report errors. Or you add
782	initialised but not necessarily started event watcher).	3739	some caching layer that finds that it can skip the lengthy aspects of the
		3740	operation and simply invoke the callback with the result.
783		3741
784	=item ev_feed_fd_event (loop, int fd, int revents)	3742	The problem here is that this will happen I<before> C<start_new_request>
		3743	has returned, so C<request> is not set.
785		3744
786	Feed an event on the given fd, as if a file descriptor backend detected	3745	Even if you pass the request by some safer means to the callback, you
787	the given events it.	3746	might want to do something to the request after starting it, such as
		3747	canceling it, which probably isn't working so well when the callback has
		3748	already been invoked.
788		3749
789	=item ev_feed_signal_event (loop, int signum)	3750	A common way around all these issues is to make sure that
		3751	C<start_new_request> I<always> returns before the callback is invoked. If
		3752	C<start_new_request> immediately knows the result, it can artificially
		3753	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3754	example, or more sneakily, by reusing an existing (stopped) watcher and
		3755	pushing it into the pending queue:
790		3756
791	Feed an event as if the given signal occured (loop must be the default loop!).	3757	ev_set_cb (watcher, callback);
		3758	ev_feed_event (EV_A_ watcher, 0);
792		3759
793	=back	3760	This way, C<start_new_request> can safely return before the callback is
		3761	invoked, while not delaying callback invocation too much.
		3762
		3763	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3764
		3765	Often (especially in GUI toolkits) there are places where you have
		3766	I<modal> interaction, which is most easily implemented by recursively
		3767	invoking C<ev_run>.
		3768
		3769	This brings the problem of exiting - a callback might want to finish the
		3770	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3771	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3772	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3773	other combination: In these cases, a simple C<ev_break> will not work.
		3774
		3775	The solution is to maintain "break this loop" variable for each C<ev_run>
		3776	invocation, and use a loop around C<ev_run> until the condition is
		3777	triggered, using C<EVRUN_ONCE>:
		3778
		3779	// main loop
		3780	int exit_main_loop = 0;
		3781
		3782	while (!exit_main_loop)
		3783	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3784
		3785	// in a modal watcher
		3786	int exit_nested_loop = 0;
		3787
		3788	while (!exit_nested_loop)
		3789	ev_run (EV_A_ EVRUN_ONCE);
		3790
		3791	To exit from any of these loops, just set the corresponding exit variable:
		3792
		3793	// exit modal loop
		3794	exit_nested_loop = 1;
		3795
		3796	// exit main program, after modal loop is finished
		3797	exit_main_loop = 1;
		3798
		3799	// exit both
		3800	exit_main_loop = exit_nested_loop = 1;
		3801
		3802	=head2 THREAD LOCKING EXAMPLE
		3803
		3804	Here is a fictitious example of how to run an event loop in a different
		3805	thread from where callbacks are being invoked and watchers are
		3806	created/added/removed.
		3807
		3808	For a real-world example, see the C<EV::Loop::Async> perl module,
		3809	which uses exactly this technique (which is suited for many high-level
		3810	languages).
		3811
		3812	The example uses a pthread mutex to protect the loop data, a condition
		3813	variable to wait for callback invocations, an async watcher to notify the
		3814	event loop thread and an unspecified mechanism to wake up the main thread.
		3815
		3816	First, you need to associate some data with the event loop:
		3817
		3818	typedef struct {
		3819	mutex_t lock; /* global loop lock */
		3820	ev_async async_w;
		3821	thread_t tid;
		3822	cond_t invoke_cv;
		3823	} userdata;
		3824
		3825	void prepare_loop (EV_P)
		3826	{
		3827	// for simplicity, we use a static userdata struct.
		3828	static userdata u;
		3829
		3830	ev_async_init (&u->async_w, async_cb);
		3831	ev_async_start (EV_A_ &u->async_w);
		3832
		3833	pthread_mutex_init (&u->lock, 0);
		3834	pthread_cond_init (&u->invoke_cv, 0);
		3835
		3836	// now associate this with the loop
		3837	ev_set_userdata (EV_A_ u);
		3838	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3839	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3840
		3841	// then create the thread running ev_run
		3842	pthread_create (&u->tid, 0, l_run, EV_A);
		3843	}
		3844
		3845	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3846	solely to wake up the event loop so it takes notice of any new watchers
		3847	that might have been added:
		3848
		3849	static void
		3850	async_cb (EV_P_ ev_async *w, int revents)
		3851	{
		3852	// just used for the side effects
		3853	}
		3854
		3855	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3856	protecting the loop data, respectively.
		3857
		3858	static void
		3859	l_release (EV_P)
		3860	{
		3861	userdata *u = ev_userdata (EV_A);
		3862	pthread_mutex_unlock (&u->lock);
		3863	}
		3864
		3865	static void
		3866	l_acquire (EV_P)
		3867	{
		3868	userdata *u = ev_userdata (EV_A);
		3869	pthread_mutex_lock (&u->lock);
		3870	}
		3871
		3872	The event loop thread first acquires the mutex, and then jumps straight
		3873	into C<ev_run>:
		3874
		3875	void *
		3876	l_run (void *thr_arg)
		3877	{
		3878	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3879
		3880	l_acquire (EV_A);
		3881	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3882	ev_run (EV_A_ 0);
		3883	l_release (EV_A);
		3884
		3885	return 0;
		3886	}
		3887
		3888	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3889	signal the main thread via some unspecified mechanism (signals? pipe
		3890	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3891	have been called (in a while loop because a) spurious wakeups are possible
		3892	and b) skipping inter-thread-communication when there are no pending
		3893	watchers is very beneficial):
		3894
		3895	static void
		3896	l_invoke (EV_P)
		3897	{
		3898	userdata *u = ev_userdata (EV_A);
		3899
		3900	while (ev_pending_count (EV_A))
		3901	{
		3902	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3903	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3904	}
		3905	}
		3906
		3907	Now, whenever the main thread gets told to invoke pending watchers, it
		3908	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3909	thread to continue:
		3910
		3911	static void
		3912	real_invoke_pending (EV_P)
		3913	{
		3914	userdata *u = ev_userdata (EV_A);
		3915
		3916	pthread_mutex_lock (&u->lock);
		3917	ev_invoke_pending (EV_A);
		3918	pthread_cond_signal (&u->invoke_cv);
		3919	pthread_mutex_unlock (&u->lock);
		3920	}
		3921
		3922	Whenever you want to start/stop a watcher or do other modifications to an
		3923	event loop, you will now have to lock:
		3924
		3925	ev_timer timeout_watcher;
		3926	userdata *u = ev_userdata (EV_A);
		3927
		3928	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3929
		3930	pthread_mutex_lock (&u->lock);
		3931	ev_timer_start (EV_A_ &timeout_watcher);
		3932	ev_async_send (EV_A_ &u->async_w);
		3933	pthread_mutex_unlock (&u->lock);
		3934
		3935	Note that sending the C<ev_async> watcher is required because otherwise
		3936	an event loop currently blocking in the kernel will have no knowledge
		3937	about the newly added timer. By waking up the loop it will pick up any new
		3938	watchers in the next event loop iteration.
		3939
		3940	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3941
		3942	While the overhead of a callback that e.g. schedules a thread is small, it
		3943	is still an overhead. If you embed libev, and your main usage is with some
		3944	kind of threads or coroutines, you might want to customise libev so that
		3945	doesn't need callbacks anymore.
		3946
		3947	Imagine you have coroutines that you can switch to using a function
		3948	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3949	and that due to some magic, the currently active coroutine is stored in a
		3950	global called C<current_coro>. Then you can build your own "wait for libev
		3951	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3952	the differing C<;> conventions):
		3953
		3954	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3955	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3956
		3957	That means instead of having a C callback function, you store the
		3958	coroutine to switch to in each watcher, and instead of having libev call
		3959	your callback, you instead have it switch to that coroutine.
		3960
		3961	A coroutine might now wait for an event with a function called
		3962	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3963	matter when, or whether the watcher is active or not when this function is
		3964	called):
		3965
		3966	void
		3967	wait_for_event (ev_watcher *w)
		3968	{
		3969	ev_set_cb (w, current_coro);
		3970	switch_to (libev_coro);
		3971	}
		3972
		3973	That basically suspends the coroutine inside C<wait_for_event> and
		3974	continues the libev coroutine, which, when appropriate, switches back to
		3975	this or any other coroutine.
		3976
		3977	You can do similar tricks if you have, say, threads with an event queue -
		3978	instead of storing a coroutine, you store the queue object and instead of
		3979	switching to a coroutine, you push the watcher onto the queue and notify
		3980	any waiters.
		3981
		3982	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		3983	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3984
		3985	// my_ev.h
		3986	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3987	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3988	#include "../libev/ev.h"
		3989
		3990	// my_ev.c
		3991	#define EV_H "my_ev.h"
		3992	#include "../libev/ev.c"
		3993
		3994	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3995	F<my_ev.c> into your project. When properly specifying include paths, you
		3996	can even use F<ev.h> as header file name directly.
		3997
794		3998
795	=head1 LIBEVENT EMULATION	3999	=head1 LIBEVENT EMULATION
796		4000
797	Libev offers a compatibility emulation layer for libevent. It cannot	4001	Libev offers a compatibility emulation layer for libevent. It cannot
798	emulate the internals of libevent, so here are some usage hints:	4002	emulate the internals of libevent, so here are some usage hints:
799		4003
800	=over 4	4004	=over 4
		4005
		4006	=item * Only the libevent-1.4.1-beta API is being emulated.
		4007
		4008	This was the newest libevent version available when libev was implemented,
		4009	and is still mostly unchanged in 2010.
801		4010
802	=item * Use it by including <event.h>, as usual.	4011	=item * Use it by including <event.h>, as usual.
803		4012
804	=item * The following members are fully supported: ev_base, ev_callback,	4013	=item * The following members are fully supported: ev_base, ev_callback,
805	ev_arg, ev_fd, ev_res, ev_events.	4014	ev_arg, ev_fd, ev_res, ev_events.
…		…
810		4019
811	=item * Priorities are not currently supported. Initialising priorities	4020	=item * Priorities are not currently supported. Initialising priorities
812	will fail and all watchers will have the same priority, even though there	4021	will fail and all watchers will have the same priority, even though there
813	is an ev_pri field.	4022	is an ev_pri field.
814		4023
		4024	=item * In libevent, the last base created gets the signals, in libev, the
		4025	base that registered the signal gets the signals.
		4026
815	=item * Other members are not supported.	4027	=item * Other members are not supported.
816		4028
817	=item * The libev emulation is I<not> ABI compatible to libevent, you need	4029	=item * The libev emulation is I<not> ABI compatible to libevent, you need
818	to use the libev header file and library.	4030	to use the libev header file and library.
819		4031
820	=back	4032	=back
821		4033
822	=head1 C++ SUPPORT	4034	=head1 C++ SUPPORT
823		4035
824	TBD.	4036	=head2 C API
		4037
		4038	The normal C API should work fine when used from C++: both ev.h and the
		4039	libev sources can be compiled as C++. Therefore, code that uses the C API
		4040	will work fine.
		4041
		4042	Proper exception specifications might have to be added to callbacks passed
		4043	to libev: exceptions may be thrown only from watcher callbacks, all other
		4044	callbacks (allocator, syserr, loop acquire/release and periodic reschedule
		4045	callbacks) must not throw exceptions, and might need a C<noexcept>
		4046	specification. If you have code that needs to be compiled as both C and
		4047	C++ you can use the C<EV_NOEXCEPT> macro for this:
		4048
		4049	static void
		4050	fatal_error (const char *msg) EV_NOEXCEPT
		4051	{
		4052	perror (msg);
		4053	abort ();
		4054	}
		4055
		4056	...
		4057	ev_set_syserr_cb (fatal_error);
		4058
		4059	The only API functions that can currently throw exceptions are C<ev_run>,
		4060	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		4061	because it runs cleanup watchers).
		4062
		4063	Throwing exceptions in watcher callbacks is only supported if libev itself
		4064	is compiled with a C++ compiler or your C and C++ environments allow
		4065	throwing exceptions through C libraries (most do).
		4066
		4067	=head2 C++ API
		4068
		4069	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		4070	you to use some convenience methods to start/stop watchers and also change
		4071	the callback model to a model using method callbacks on objects.
		4072
		4073	To use it,
		4074
		4075	#include <ev++.h>
		4076
		4077	This automatically includes F<ev.h> and puts all of its definitions (many
		4078	of them macros) into the global namespace. All C++ specific things are
		4079	put into the C<ev> namespace. It should support all the same embedding
		4080	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		4081
		4082	Care has been taken to keep the overhead low. The only data member the C++
		4083	classes add (compared to plain C-style watchers) is the event loop pointer
		4084	that the watcher is associated with (or no additional members at all if
		4085	you disable C<EV_MULTIPLICITY> when embedding libev).
		4086
		4087	Currently, functions, static and non-static member functions and classes
		4088	with C<operator ()> can be used as callbacks. Other types should be easy
		4089	to add as long as they only need one additional pointer for context. If
		4090	you need support for other types of functors please contact the author
		4091	(preferably after implementing it).
		4092
		4093	For all this to work, your C++ compiler either has to use the same calling
		4094	conventions as your C compiler (for static member functions), or you have
		4095	to embed libev and compile libev itself as C++.
		4096
		4097	Here is a list of things available in the C<ev> namespace:
		4098
		4099	=over 4
		4100
		4101	=item C<ev::READ>, C<ev::WRITE> etc.
		4102
		4103	These are just enum values with the same values as the C<EV_READ> etc.
		4104	macros from F<ev.h>.
		4105
		4106	=item C<ev::tstamp>, C<ev::now>
		4107
		4108	Aliases to the same types/functions as with the C<ev_> prefix.
		4109
		4110	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		4111
		4112	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		4113	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		4114	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		4115	defined by many implementations.
		4116
		4117	All of those classes have these methods:
		4118
		4119	=over 4
		4120
		4121	=item ev::TYPE::TYPE ()
		4122
		4123	=item ev::TYPE::TYPE (loop)
		4124
		4125	=item ev::TYPE::~TYPE
		4126
		4127	The constructor (optionally) takes an event loop to associate the watcher
		4128	with. If it is omitted, it will use C<EV_DEFAULT>.
		4129
		4130	The constructor calls C<ev_init> for you, which means you have to call the
		4131	C<set> method before starting it.
		4132
		4133	It will not set a callback, however: You have to call the templated C<set>
		4134	method to set a callback before you can start the watcher.
		4135
		4136	(The reason why you have to use a method is a limitation in C++ which does
		4137	not allow explicit template arguments for constructors).
		4138
		4139	The destructor automatically stops the watcher if it is active.
		4140
		4141	=item w->set<class, &class::method> (object *)
		4142
		4143	This method sets the callback method to call. The method has to have a
		4144	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		4145	first argument and the C<revents> as second. The object must be given as
		4146	parameter and is stored in the C<data> member of the watcher.
		4147
		4148	This method synthesizes efficient thunking code to call your method from
		4149	the C callback that libev requires. If your compiler can inline your
		4150	callback (i.e. it is visible to it at the place of the C<set> call and
		4151	your compiler is good :), then the method will be fully inlined into the
		4152	thunking function, making it as fast as a direct C callback.
		4153
		4154	Example: simple class declaration and watcher initialisation
		4155
		4156	struct myclass
		4157	{
		4158	void io_cb (ev::io &w, int revents) { }
		4159	}
		4160
		4161	myclass obj;
		4162	ev::io iow;
		4163	iow.set <myclass, &myclass::io_cb> (&obj);
		4164
		4165	=item w->set (object *)
		4166
		4167	This is a variation of a method callback - leaving out the method to call
		4168	will default the method to C<operator ()>, which makes it possible to use
		4169	functor objects without having to manually specify the C<operator ()> all
		4170	the time. Incidentally, you can then also leave out the template argument
		4171	list.
		4172
		4173	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		4174	int revents)>.
		4175
		4176	See the method-C<set> above for more details.
		4177
		4178	Example: use a functor object as callback.
		4179
		4180	struct myfunctor
		4181	{
		4182	void operator() (ev::io &w, int revents)
		4183	{
		4184	...
		4185	}
		4186	}
		4187
		4188	myfunctor f;
		4189
		4190	ev::io w;
		4191	w.set (&f);
		4192
		4193	=item w->set<function> (void *data = 0)
		4194
		4195	Also sets a callback, but uses a static method or plain function as
		4196	callback. The optional C<data> argument will be stored in the watcher's
		4197	C<data> member and is free for you to use.
		4198
		4199	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		4200
		4201	See the method-C<set> above for more details.
		4202
		4203	Example: Use a plain function as callback.
		4204
		4205	static void io_cb (ev::io &w, int revents) { }
		4206	iow.set <io_cb> ();
		4207
		4208	=item w->set (loop)
		4209
		4210	Associates a different C<struct ev_loop> with this watcher. You can only
		4211	do this when the watcher is inactive (and not pending either).
		4212
		4213	=item w->set ([arguments])
		4214
		4215	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
		4216	with the same arguments. Either this method or a suitable start method
		4217	must be called at least once. Unlike the C counterpart, an active watcher
		4218	gets automatically stopped and restarted when reconfiguring it with this
		4219	method.
		4220
		4221	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4222	clashing with the C<set (loop)> method.
		4223
		4224	=item w->start ()
		4225
		4226	Starts the watcher. Note that there is no C<loop> argument, as the
		4227	constructor already stores the event loop.
		4228
		4229	=item w->start ([arguments])
		4230
		4231	Instead of calling C<set> and C<start> methods separately, it is often
		4232	convenient to wrap them in one call. Uses the same type of arguments as
		4233	the configure C<set> method of the watcher.
		4234
		4235	=item w->stop ()
		4236
		4237	Stops the watcher if it is active. Again, no C<loop> argument.
		4238
		4239	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		4240
		4241	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		4242	C<ev_TYPE_again> function.
		4243
		4244	=item w->sweep () (C<ev::embed> only)
		4245
		4246	Invokes C<ev_embed_sweep>.
		4247
		4248	=item w->update () (C<ev::stat> only)
		4249
		4250	Invokes C<ev_stat_stat>.
		4251
		4252	=back
		4253
		4254	=back
		4255
		4256	Example: Define a class with two I/O and idle watchers, start the I/O
		4257	watchers in the constructor.
		4258
		4259	class myclass
		4260	{
		4261	ev::io io ; void io_cb (ev::io &w, int revents);
		4262	ev::io io2 ; void io2_cb (ev::io &w, int revents);
		4263	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		4264
		4265	myclass (int fd)
		4266	{
		4267	io .set <myclass, &myclass::io_cb > (this);
		4268	io2 .set <myclass, &myclass::io2_cb > (this);
		4269	idle.set <myclass, &myclass::idle_cb> (this);
		4270
		4271	io.set (fd, ev::WRITE); // configure the watcher
		4272	io.start (); // start it whenever convenient
		4273
		4274	io2.start (fd, ev::READ); // set + start in one call
		4275	}
		4276	};
		4277
		4278
		4279	=head1 OTHER LANGUAGE BINDINGS
		4280
		4281	Libev does not offer other language bindings itself, but bindings for a
		4282	number of languages exist in the form of third-party packages. If you know
		4283	any interesting language binding in addition to the ones listed here, drop
		4284	me a note.
		4285
		4286	=over 4
		4287
		4288	=item Perl
		4289
		4290	The EV module implements the full libev API and is actually used to test
		4291	libev. EV is developed together with libev. Apart from the EV core module,
		4292	there are additional modules that implement libev-compatible interfaces
		4293	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		4294	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		4295	and C<EV::Glib>).
		4296
		4297	It can be found and installed via CPAN, its homepage is at
		4298	L<http://software.schmorp.de/pkg/EV>.
		4299
		4300	=item Python
		4301
		4302	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		4303	seems to be quite complete and well-documented.
		4304
		4305	=item Ruby
		4306
		4307	Tony Arcieri has written a ruby extension that offers access to a subset
		4308	of the libev API and adds file handle abstractions, asynchronous DNS and
		4309	more on top of it. It can be found via gem servers. Its homepage is at
		4310	L<http://rev.rubyforge.org/>.
		4311
		4312	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		4313	makes rev work even on mingw.
		4314
		4315	=item Haskell
		4316
		4317	A haskell binding to libev is available at
		4318	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		4319
		4320	=item D
		4321
		4322	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		4323	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
		4324
		4325	=item Ocaml
		4326
		4327	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4328	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4329
		4330	=item Lua
		4331
		4332	Brian Maher has written a partial interface to libev for lua (at the
		4333	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4334	L<http://github.com/brimworks/lua-ev>.
		4335
		4336	=item Javascript
		4337
		4338	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4339
		4340	=item Others
		4341
		4342	There are others, and I stopped counting.
		4343
		4344	=back
		4345
		4346
		4347	=head1 MACRO MAGIC
		4348
		4349	Libev can be compiled with a variety of options, the most fundamental
		4350	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		4351	functions and callbacks have an initial C<struct ev_loop *> argument.
		4352
		4353	To make it easier to write programs that cope with either variant, the
		4354	following macros are defined:
		4355
		4356	=over 4
		4357
		4358	=item C<EV_A>, C<EV_A_>
		4359
		4360	This provides the loop I<argument> for functions, if one is required ("ev
		4361	loop argument"). The C<EV_A> form is used when this is the sole argument,
		4362	C<EV_A_> is used when other arguments are following. Example:
		4363
		4364	ev_unref (EV_A);
		4365	ev_timer_add (EV_A_ watcher);
		4366	ev_run (EV_A_ 0);
		4367
		4368	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		4369	which is often provided by the following macro.
		4370
		4371	=item C<EV_P>, C<EV_P_>
		4372
		4373	This provides the loop I<parameter> for functions, if one is required ("ev
		4374	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		4375	C<EV_P_> is used when other parameters are following. Example:
		4376
		4377	// this is how ev_unref is being declared
		4378	static void ev_unref (EV_P);
		4379
		4380	// this is how you can declare your typical callback
		4381	static void cb (EV_P_ ev_timer *w, int revents)
		4382
		4383	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		4384	suitable for use with C<EV_A>.
		4385
		4386	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		4387
		4388	Similar to the other two macros, this gives you the value of the default
		4389	loop, if multiple loops are supported ("ev loop default"). The default loop
		4390	will be initialised if it isn't already initialised.
		4391
		4392	For non-multiplicity builds, these macros do nothing, so you always have
		4393	to initialise the loop somewhere.
		4394
		4395	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		4396
		4397	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		4398	default loop has been initialised (C<UC> == unchecked). Their behaviour
		4399	is undefined when the default loop has not been initialised by a previous
		4400	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		4401
		4402	It is often prudent to use C<EV_DEFAULT> when initialising the first
		4403	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		4404
		4405	=back
		4406
		4407	Example: Declare and initialise a check watcher, utilising the above
		4408	macros so it will work regardless of whether multiple loops are supported
		4409	or not.
		4410
		4411	static void
		4412	check_cb (EV_P_ ev_timer *w, int revents)
		4413	{
		4414	ev_check_stop (EV_A_ w);
		4415	}
		4416
		4417	ev_check check;
		4418	ev_check_init (&check, check_cb);
		4419	ev_check_start (EV_DEFAULT_ &check);
		4420	ev_run (EV_DEFAULT_ 0);
		4421
		4422	=head1 EMBEDDING
		4423
		4424	Libev can (and often is) directly embedded into host
		4425	applications. Examples of applications that embed it include the Deliantra
		4426	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		4427	and rxvt-unicode.
		4428
		4429	The goal is to enable you to just copy the necessary files into your
		4430	source directory without having to change even a single line in them, so
		4431	you can easily upgrade by simply copying (or having a checked-out copy of
		4432	libev somewhere in your source tree).
		4433
		4434	=head2 FILESETS
		4435
		4436	Depending on what features you need you need to include one or more sets of files
		4437	in your application.
		4438
		4439	=head3 CORE EVENT LOOP
		4440
		4441	To include only the libev core (all the C<ev_*> functions), with manual
		4442	configuration (no autoconf):
		4443
		4444	#define EV_STANDALONE 1
		4445	#include "ev.c"
		4446
		4447	This will automatically include F<ev.h>, too, and should be done in a
		4448	single C source file only to provide the function implementations. To use
		4449	it, do the same for F<ev.h> in all files wishing to use this API (best
		4450	done by writing a wrapper around F<ev.h> that you can include instead and
		4451	where you can put other configuration options):
		4452
		4453	#define EV_STANDALONE 1
		4454	#include "ev.h"
		4455
		4456	Both header files and implementation files can be compiled with a C++
		4457	compiler (at least, that's a stated goal, and breakage will be treated
		4458	as a bug).
		4459
		4460	You need the following files in your source tree, or in a directory
		4461	in your include path (e.g. in libev/ when using -Ilibev):
		4462
		4463	ev.h
		4464	ev.c
		4465	ev_vars.h
		4466	ev_wrap.h
		4467
		4468	ev_win32.c required on win32 platforms only
		4469
		4470	ev_select.c only when select backend is enabled
		4471	ev_poll.c only when poll backend is enabled
		4472	ev_epoll.c only when the epoll backend is enabled
		4473	ev_linuxaio.c only when the linux aio backend is enabled
		4474	ev_kqueue.c only when the kqueue backend is enabled
		4475	ev_port.c only when the solaris port backend is enabled
		4476
		4477	F<ev.c> includes the backend files directly when enabled, so you only need
		4478	to compile this single file.
		4479
		4480	=head3 LIBEVENT COMPATIBILITY API
		4481
		4482	To include the libevent compatibility API, also include:
		4483
		4484	#include "event.c"
		4485
		4486	in the file including F<ev.c>, and:
		4487
		4488	#include "event.h"
		4489
		4490	in the files that want to use the libevent API. This also includes F<ev.h>.
		4491
		4492	You need the following additional files for this:
		4493
		4494	event.h
		4495	event.c
		4496
		4497	=head3 AUTOCONF SUPPORT
		4498
		4499	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		4500	whatever way you want, you can also C<m4_include([libev.m4])> in your
		4501	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		4502	include F<config.h> and configure itself accordingly.
		4503
		4504	For this of course you need the m4 file:
		4505
		4506	libev.m4
		4507
		4508	=head2 PREPROCESSOR SYMBOLS/MACROS
		4509
		4510	Libev can be configured via a variety of preprocessor symbols you have to
		4511	define before including (or compiling) any of its files. The default in
		4512	the absence of autoconf is documented for every option.
		4513
		4514	Symbols marked with "(h)" do not change the ABI, and can have different
		4515	values when compiling libev vs. including F<ev.h>, so it is permissible
		4516	to redefine them before including F<ev.h> without breaking compatibility
		4517	to a compiled library. All other symbols change the ABI, which means all
		4518	users of libev and the libev code itself must be compiled with compatible
		4519	settings.
		4520
		4521	=over 4
		4522
		4523	=item EV_COMPAT3 (h)
		4524
		4525	Backwards compatibility is a major concern for libev. This is why this
		4526	release of libev comes with wrappers for the functions and symbols that
		4527	have been renamed between libev version 3 and 4.
		4528
		4529	You can disable these wrappers (to test compatibility with future
		4530	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4531	sources. This has the additional advantage that you can drop the C<struct>
		4532	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4533	typedef in that case.
		4534
		4535	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4536	and in some even more future version the compatibility code will be
		4537	removed completely.
		4538
		4539	=item EV_STANDALONE (h)
		4540
		4541	Must always be C<1> if you do not use autoconf configuration, which
		4542	keeps libev from including F<config.h>, and it also defines dummy
		4543	implementations for some libevent functions (such as logging, which is not
		4544	supported). It will also not define any of the structs usually found in
		4545	F<event.h> that are not directly supported by the libev core alone.
		4546
		4547	In standalone mode, libev will still try to automatically deduce the
		4548	configuration, but has to be more conservative.
		4549
		4550	=item EV_USE_FLOOR
		4551
		4552	If defined to be C<1>, libev will use the C<floor ()> function for its
		4553	periodic reschedule calculations, otherwise libev will fall back on a
		4554	portable (slower) implementation. If you enable this, you usually have to
		4555	link against libm or something equivalent. Enabling this when the C<floor>
		4556	function is not available will fail, so the safe default is to not enable
		4557	this.
		4558
		4559	=item EV_USE_MONOTONIC
		4560
		4561	If defined to be C<1>, libev will try to detect the availability of the
		4562	monotonic clock option at both compile time and runtime. Otherwise no
		4563	use of the monotonic clock option will be attempted. If you enable this,
		4564	you usually have to link against librt or something similar. Enabling it
		4565	when the functionality isn't available is safe, though, although you have
		4566	to make sure you link against any libraries where the C<clock_gettime>
		4567	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		4568
		4569	=item EV_USE_REALTIME
		4570
		4571	If defined to be C<1>, libev will try to detect the availability of the
		4572	real-time clock option at compile time (and assume its availability
		4573	at runtime if successful). Otherwise no use of the real-time clock
		4574	option will be attempted. This effectively replaces C<gettimeofday>
		4575	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		4576	correctness. See the note about libraries in the description of
		4577	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4578	C<EV_USE_CLOCK_SYSCALL>.
		4579
		4580	=item EV_USE_CLOCK_SYSCALL
		4581
		4582	If defined to be C<1>, libev will try to use a direct syscall instead
		4583	of calling the system-provided C<clock_gettime> function. This option
		4584	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4585	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4586	programs needlessly. Using a direct syscall is slightly slower (in
		4587	theory), because no optimised vdso implementation can be used, but avoids
		4588	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4589	higher, as it simplifies linking (no need for C<-lrt>).
		4590
		4591	=item EV_USE_NANOSLEEP
		4592
		4593	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4594	and will use it for delays. Otherwise it will use C<select ()>.
		4595
		4596	=item EV_USE_EVENTFD
		4597
		4598	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4599	available and will probe for kernel support at runtime. This will improve
		4600	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4601	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4602	2.7 or newer, otherwise disabled.
		4603
		4604	=item EV_USE_SELECT
		4605
		4606	If undefined or defined to be C<1>, libev will compile in support for the
		4607	C<select>(2) backend. No attempt at auto-detection will be done: if no
		4608	other method takes over, select will be it. Otherwise the select backend
		4609	will not be compiled in.
		4610
		4611	=item EV_SELECT_USE_FD_SET
		4612
		4613	If defined to C<1>, then the select backend will use the system C<fd_set>
		4614	structure. This is useful if libev doesn't compile due to a missing
		4615	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		4616	on exotic systems. This usually limits the range of file descriptors to
		4617	some low limit such as 1024 or might have other limitations (winsocket
		4618	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		4619	configures the maximum size of the C<fd_set>.
		4620
		4621	=item EV_SELECT_IS_WINSOCKET
		4622
		4623	When defined to C<1>, the select backend will assume that
		4624	select/socket/connect etc. don't understand file descriptors but
		4625	wants osf handles on win32 (this is the case when the select to
		4626	be used is the winsock select). This means that it will call
		4627	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		4628	it is assumed that all these functions actually work on fds, even
		4629	on win32. Should not be defined on non-win32 platforms.
		4630
		4631	=item EV_FD_TO_WIN32_HANDLE(fd)
		4632
		4633	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4634	file descriptors to socket handles. When not defining this symbol (the
		4635	default), then libev will call C<_get_osfhandle>, which is usually
		4636	correct. In some cases, programs use their own file descriptor management,
		4637	in which case they can provide this function to map fds to socket handles.
		4638
		4639	=item EV_WIN32_HANDLE_TO_FD(handle)
		4640
		4641	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4642	using the standard C<_open_osfhandle> function. For programs implementing
		4643	their own fd to handle mapping, overwriting this function makes it easier
		4644	to do so. This can be done by defining this macro to an appropriate value.
		4645
		4646	=item EV_WIN32_CLOSE_FD(fd)
		4647
		4648	If programs implement their own fd to handle mapping on win32, then this
		4649	macro can be used to override the C<close> function, useful to unregister
		4650	file descriptors again. Note that the replacement function has to close
		4651	the underlying OS handle.
		4652
		4653	=item EV_USE_WSASOCKET
		4654
		4655	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4656	communication socket, which works better in some environments. Otherwise,
		4657	the normal C<socket> function will be used, which works better in other
		4658	environments.
		4659
		4660	=item EV_USE_POLL
		4661
		4662	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		4663	backend. Otherwise it will be enabled on non-win32 platforms. It
		4664	takes precedence over select.
		4665
		4666	=item EV_USE_EPOLL
		4667
		4668	If defined to be C<1>, libev will compile in support for the Linux
		4669	C<epoll>(7) backend. Its availability will be detected at runtime,
		4670	otherwise another method will be used as fallback. This is the preferred
		4671	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4672	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4673
		4674	=item EV_USE_LINUXAIO
		4675
		4676	If defined to be C<1>, libev will compile in support for the Linux
		4677	aio backend. Due to it's currenbt limitations it has to be requested
		4678	explicitly. If undefined, it will be enabled on linux, otherwise
		4679	disabled.
		4680
		4681	=item EV_USE_KQUEUE
		4682
		4683	If defined to be C<1>, libev will compile in support for the BSD style
		4684	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		4685	otherwise another method will be used as fallback. This is the preferred
		4686	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		4687	supports some types of fds correctly (the only platform we found that
		4688	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		4689	not be used unless explicitly requested. The best way to use it is to find
		4690	out whether kqueue supports your type of fd properly and use an embedded
		4691	kqueue loop.
		4692
		4693	=item EV_USE_PORT
		4694
		4695	If defined to be C<1>, libev will compile in support for the Solaris
		4696	10 port style backend. Its availability will be detected at runtime,
		4697	otherwise another method will be used as fallback. This is the preferred
		4698	backend for Solaris 10 systems.
		4699
		4700	=item EV_USE_DEVPOLL
		4701
		4702	Reserved for future expansion, works like the USE symbols above.
		4703
		4704	=item EV_USE_INOTIFY
		4705
		4706	If defined to be C<1>, libev will compile in support for the Linux inotify
		4707	interface to speed up C<ev_stat> watchers. Its actual availability will
		4708	be detected at runtime. If undefined, it will be enabled if the headers
		4709	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4710
		4711	=item EV_NO_SMP
		4712
		4713	If defined to be C<1>, libev will assume that memory is always coherent
		4714	between threads, that is, threads can be used, but threads never run on
		4715	different cpus (or different cpu cores). This reduces dependencies
		4716	and makes libev faster.
		4717
		4718	=item EV_NO_THREADS
		4719
		4720	If defined to be C<1>, libev will assume that it will never be called from
		4721	different threads (that includes signal handlers), which is a stronger
		4722	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4723	libev faster.
		4724
		4725	=item EV_ATOMIC_T
		4726
		4727	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4728	access is atomic with respect to other threads or signal contexts. No
		4729	such type is easily found in the C language, so you can provide your own
		4730	type that you know is safe for your purposes. It is used both for signal
		4731	handler "locking" as well as for signal and thread safety in C<ev_async>
		4732	watchers.
		4733
		4734	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4735	(from F<signal.h>), which is usually good enough on most platforms.
		4736
		4737	=item EV_H (h)
		4738
		4739	The name of the F<ev.h> header file used to include it. The default if
		4740	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4741	used to virtually rename the F<ev.h> header file in case of conflicts.
		4742
		4743	=item EV_CONFIG_H (h)
		4744
		4745	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4746	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4747	C<EV_H>, above.
		4748
		4749	=item EV_EVENT_H (h)
		4750
		4751	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4752	of how the F<event.h> header can be found, the default is C<"event.h">.
		4753
		4754	=item EV_PROTOTYPES (h)
		4755
		4756	If defined to be C<0>, then F<ev.h> will not define any function
		4757	prototypes, but still define all the structs and other symbols. This is
		4758	occasionally useful if you want to provide your own wrapper functions
		4759	around libev functions.
		4760
		4761	=item EV_MULTIPLICITY
		4762
		4763	If undefined or defined to C<1>, then all event-loop-specific functions
		4764	will have the C<struct ev_loop *> as first argument, and you can create
		4765	additional independent event loops. Otherwise there will be no support
		4766	for multiple event loops and there is no first event loop pointer
		4767	argument. Instead, all functions act on the single default loop.
		4768
		4769	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4770	default loop when multiplicity is switched off - you always have to
		4771	initialise the loop manually in this case.
		4772
		4773	=item EV_MINPRI
		4774
		4775	=item EV_MAXPRI
		4776
		4777	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4778	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4779	provide for more priorities by overriding those symbols (usually defined
		4780	to be C<-2> and C<2>, respectively).
		4781
		4782	When doing priority-based operations, libev usually has to linearly search
		4783	all the priorities, so having many of them (hundreds) uses a lot of space
		4784	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4785	fine.
		4786
		4787	If your embedding application does not need any priorities, defining these
		4788	both to C<0> will save some memory and CPU.
		4789
		4790	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4791	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4792	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4793
		4794	If undefined or defined to be C<1> (and the platform supports it), then
		4795	the respective watcher type is supported. If defined to be C<0>, then it
		4796	is not. Disabling watcher types mainly saves code size.
		4797
		4798	=item EV_FEATURES
		4799
		4800	If you need to shave off some kilobytes of code at the expense of some
		4801	speed (but with the full API), you can define this symbol to request
		4802	certain subsets of functionality. The default is to enable all features
		4803	that can be enabled on the platform.
		4804
		4805	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4806	with some broad features you want) and then selectively re-enable
		4807	additional parts you want, for example if you want everything minimal,
		4808	but multiple event loop support, async and child watchers and the poll
		4809	backend, use this:
		4810
		4811	#define EV_FEATURES 0
		4812	#define EV_MULTIPLICITY 1
		4813	#define EV_USE_POLL 1
		4814	#define EV_CHILD_ENABLE 1
		4815	#define EV_ASYNC_ENABLE 1
		4816
		4817	The actual value is a bitset, it can be a combination of the following
		4818	values (by default, all of these are enabled):
		4819
		4820	=over 4
		4821
		4822	=item C<1> - faster/larger code
		4823
		4824	Use larger code to speed up some operations.
		4825
		4826	Currently this is used to override some inlining decisions (enlarging the
		4827	code size by roughly 30% on amd64).
		4828
		4829	When optimising for size, use of compiler flags such as C<-Os> with
		4830	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4831	assertions.
		4832
		4833	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4834	(e.g. gcc with C<-Os>).
		4835
		4836	=item C<2> - faster/larger data structures
		4837
		4838	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4839	hash table sizes and so on. This will usually further increase code size
		4840	and can additionally have an effect on the size of data structures at
		4841	runtime.
		4842
		4843	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4844	(e.g. gcc with C<-Os>).
		4845
		4846	=item C<4> - full API configuration
		4847
		4848	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4849	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4850
		4851	=item C<8> - full API
		4852
		4853	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4854	details on which parts of the API are still available without this
		4855	feature, and do not complain if this subset changes over time.
		4856
		4857	=item C<16> - enable all optional watcher types
		4858
		4859	Enables all optional watcher types. If you want to selectively enable
		4860	only some watcher types other than I/O and timers (e.g. prepare,
		4861	embed, async, child...) you can enable them manually by defining
		4862	C<EV_watchertype_ENABLE> to C<1> instead.
		4863
		4864	=item C<32> - enable all backends
		4865
		4866	This enables all backends - without this feature, you need to enable at
		4867	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4868
		4869	=item C<64> - enable OS-specific "helper" APIs
		4870
		4871	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4872	default.
		4873
		4874	=back
		4875
		4876	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4877	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4878	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4879	watchers, timers and monotonic clock support.
		4880
		4881	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4882	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4883	your program might be left out as well - a binary starting a timer and an
		4884	I/O watcher then might come out at only 5Kb.
		4885
		4886	=item EV_API_STATIC
		4887
		4888	If this symbol is defined (by default it is not), then all identifiers
		4889	will have static linkage. This means that libev will not export any
		4890	identifiers, and you cannot link against libev anymore. This can be useful
		4891	when you embed libev, only want to use libev functions in a single file,
		4892	and do not want its identifiers to be visible.
		4893
		4894	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4895	wants to use libev.
		4896
		4897	This option only works when libev is compiled with a C compiler, as C++
		4898	doesn't support the required declaration syntax.
		4899
		4900	=item EV_AVOID_STDIO
		4901
		4902	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4903	functions (printf, scanf, perror etc.). This will increase the code size
		4904	somewhat, but if your program doesn't otherwise depend on stdio and your
		4905	libc allows it, this avoids linking in the stdio library which is quite
		4906	big.
		4907
		4908	Note that error messages might become less precise when this option is
		4909	enabled.
		4910
		4911	=item EV_NSIG
		4912
		4913	The highest supported signal number, +1 (or, the number of
		4914	signals): Normally, libev tries to deduce the maximum number of signals
		4915	automatically, but sometimes this fails, in which case it can be
		4916	specified. Also, using a lower number than detected (C<32> should be
		4917	good for about any system in existence) can save some memory, as libev
		4918	statically allocates some 12-24 bytes per signal number.
		4919
		4920	=item EV_PID_HASHSIZE
		4921
		4922	C<ev_child> watchers use a small hash table to distribute workload by
		4923	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4924	usually more than enough. If you need to manage thousands of children you
		4925	might want to increase this value (I<must> be a power of two).
		4926
		4927	=item EV_INOTIFY_HASHSIZE
		4928
		4929	C<ev_stat> watchers use a small hash table to distribute workload by
		4930	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4931	disabled), usually more than enough. If you need to manage thousands of
		4932	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4933	power of two).
		4934
		4935	=item EV_USE_4HEAP
		4936
		4937	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4938	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4939	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4940	faster performance with many (thousands) of watchers.
		4941
		4942	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4943	will be C<0>.
		4944
		4945	=item EV_HEAP_CACHE_AT
		4946
		4947	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4948	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4949	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4950	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4951	but avoids random read accesses on heap changes. This improves performance
		4952	noticeably with many (hundreds) of watchers.
		4953
		4954	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4955	will be C<0>.
		4956
		4957	=item EV_VERIFY
		4958
		4959	Controls how much internal verification (see C<ev_verify ()>) will
		4960	be done: If set to C<0>, no internal verification code will be compiled
		4961	in. If set to C<1>, then verification code will be compiled in, but not
		4962	called. If set to C<2>, then the internal verification code will be
		4963	called once per loop, which can slow down libev. If set to C<3>, then the
		4964	verification code will be called very frequently, which will slow down
		4965	libev considerably.
		4966
		4967	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4968	will be C<0>.
		4969
		4970	=item EV_COMMON
		4971
		4972	By default, all watchers have a C<void *data> member. By redefining
		4973	this macro to something else you can include more and other types of
		4974	members. You have to define it each time you include one of the files,
		4975	though, and it must be identical each time.
		4976
		4977	For example, the perl EV module uses something like this:
		4978
		4979	#define EV_COMMON \
		4980	SV *self; /* contains this struct */ \
		4981	SV *cb_sv, *fh /* note no trailing ";" */
		4982
		4983	=item EV_CB_DECLARE (type)
		4984
		4985	=item EV_CB_INVOKE (watcher, revents)
		4986
		4987	=item ev_set_cb (ev, cb)
		4988
		4989	Can be used to change the callback member declaration in each watcher,
		4990	and the way callbacks are invoked and set. Must expand to a struct member
		4991	definition and a statement, respectively. See the F<ev.h> header file for
		4992	their default definitions. One possible use for overriding these is to
		4993	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4994	method calls instead of plain function calls in C++.
		4995
		4996	=back
		4997
		4998	=head2 EXPORTED API SYMBOLS
		4999
		5000	If you need to re-export the API (e.g. via a DLL) and you need a list of
		5001	exported symbols, you can use the provided F<Symbol.*> files which list
		5002	all public symbols, one per line:
		5003
		5004	Symbols.ev for libev proper
		5005	Symbols.event for the libevent emulation
		5006
		5007	This can also be used to rename all public symbols to avoid clashes with
		5008	multiple versions of libev linked together (which is obviously bad in
		5009	itself, but sometimes it is inconvenient to avoid this).
		5010
		5011	A sed command like this will create wrapper C<#define>'s that you need to
		5012	include before including F<ev.h>:
		5013
		5014	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		5015
		5016	This would create a file F<wrap.h> which essentially looks like this:
		5017
		5018	#define ev_backend myprefix_ev_backend
		5019	#define ev_check_start myprefix_ev_check_start
		5020	#define ev_check_stop myprefix_ev_check_stop
		5021	...
		5022
		5023	=head2 EXAMPLES
		5024
		5025	For a real-world example of a program the includes libev
		5026	verbatim, you can have a look at the EV perl module
		5027	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		5028	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		5029	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		5030	will be compiled. It is pretty complex because it provides its own header
		5031	file.
		5032
		5033	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		5034	that everybody includes and which overrides some configure choices:
		5035
		5036	#define EV_FEATURES 8
		5037	#define EV_USE_SELECT 1
		5038	#define EV_PREPARE_ENABLE 1
		5039	#define EV_IDLE_ENABLE 1
		5040	#define EV_SIGNAL_ENABLE 1
		5041	#define EV_CHILD_ENABLE 1
		5042	#define EV_USE_STDEXCEPT 0
		5043	#define EV_CONFIG_H <config.h>
		5044
		5045	#include "ev++.h"
		5046
		5047	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		5048
		5049	#include "ev_cpp.h"
		5050	#include "ev.c"
		5051
		5052	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
		5053
		5054	=head2 THREADS AND COROUTINES
		5055
		5056	=head3 THREADS
		5057
		5058	All libev functions are reentrant and thread-safe unless explicitly
		5059	documented otherwise, but libev implements no locking itself. This means
		5060	that you can use as many loops as you want in parallel, as long as there
		5061	are no concurrent calls into any libev function with the same loop
		5062	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		5063	of course): libev guarantees that different event loops share no data
		5064	structures that need any locking.
		5065
		5066	Or to put it differently: calls with different loop parameters can be done
		5067	concurrently from multiple threads, calls with the same loop parameter
		5068	must be done serially (but can be done from different threads, as long as
		5069	only one thread ever is inside a call at any point in time, e.g. by using
		5070	a mutex per loop).
		5071
		5072	Specifically to support threads (and signal handlers), libev implements
		5073	so-called C<ev_async> watchers, which allow some limited form of
		5074	concurrency on the same event loop, namely waking it up "from the
		5075	outside".
		5076
		5077	If you want to know which design (one loop, locking, or multiple loops
		5078	without or something else still) is best for your problem, then I cannot
		5079	help you, but here is some generic advice:
		5080
		5081	=over 4
		5082
		5083	=item * most applications have a main thread: use the default libev loop
		5084	in that thread, or create a separate thread running only the default loop.
		5085
		5086	This helps integrating other libraries or software modules that use libev
		5087	themselves and don't care/know about threading.
		5088
		5089	=item * one loop per thread is usually a good model.
		5090
		5091	Doing this is almost never wrong, sometimes a better-performance model
		5092	exists, but it is always a good start.
		5093
		5094	=item * other models exist, such as the leader/follower pattern, where one
		5095	loop is handed through multiple threads in a kind of round-robin fashion.
		5096
		5097	Choosing a model is hard - look around, learn, know that usually you can do
		5098	better than you currently do :-)
		5099
		5100	=item * often you need to talk to some other thread which blocks in the
		5101	event loop.
		5102
		5103	C<ev_async> watchers can be used to wake them up from other threads safely
		5104	(or from signal contexts...).
		5105
		5106	An example use would be to communicate signals or other events that only
		5107	work in the default loop by registering the signal watcher with the
		5108	default loop and triggering an C<ev_async> watcher from the default loop
		5109	watcher callback into the event loop interested in the signal.
		5110
		5111	=back
		5112
		5113	See also L</THREAD LOCKING EXAMPLE>.
		5114
		5115	=head3 COROUTINES
		5116
		5117	Libev is very accommodating to coroutines ("cooperative threads"):
		5118	libev fully supports nesting calls to its functions from different
		5119	coroutines (e.g. you can call C<ev_run> on the same loop from two
		5120	different coroutines, and switch freely between both coroutines running
		5121	the loop, as long as you don't confuse yourself). The only exception is
		5122	that you must not do this from C<ev_periodic> reschedule callbacks.
		5123
		5124	Care has been taken to ensure that libev does not keep local state inside
		5125	C<ev_run>, and other calls do not usually allow for coroutine switches as
		5126	they do not call any callbacks.
		5127
		5128	=head2 COMPILER WARNINGS
		5129
		5130	Depending on your compiler and compiler settings, you might get no or a
		5131	lot of warnings when compiling libev code. Some people are apparently
		5132	scared by this.
		5133
		5134	However, these are unavoidable for many reasons. For one, each compiler
		5135	has different warnings, and each user has different tastes regarding
		5136	warning options. "Warn-free" code therefore cannot be a goal except when
		5137	targeting a specific compiler and compiler-version.
		5138
		5139	Another reason is that some compiler warnings require elaborate
		5140	workarounds, or other changes to the code that make it less clear and less
		5141	maintainable.
		5142
		5143	And of course, some compiler warnings are just plain stupid, or simply
		5144	wrong (because they don't actually warn about the condition their message
		5145	seems to warn about). For example, certain older gcc versions had some
		5146	warnings that resulted in an extreme number of false positives. These have
		5147	been fixed, but some people still insist on making code warn-free with
		5148	such buggy versions.
		5149
		5150	While libev is written to generate as few warnings as possible,
		5151	"warn-free" code is not a goal, and it is recommended not to build libev
		5152	with any compiler warnings enabled unless you are prepared to cope with
		5153	them (e.g. by ignoring them). Remember that warnings are just that:
		5154	warnings, not errors, or proof of bugs.
		5155
		5156
		5157	=head2 VALGRIND
		5158
		5159	Valgrind has a special section here because it is a popular tool that is
		5160	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		5161
		5162	If you think you found a bug (memory leak, uninitialised data access etc.)
		5163	in libev, then check twice: If valgrind reports something like:
		5164
		5165	==2274== definitely lost: 0 bytes in 0 blocks.
		5166	==2274== possibly lost: 0 bytes in 0 blocks.
		5167	==2274== still reachable: 256 bytes in 1 blocks.
		5168
		5169	Then there is no memory leak, just as memory accounted to global variables
		5170	is not a memleak - the memory is still being referenced, and didn't leak.
		5171
		5172	Similarly, under some circumstances, valgrind might report kernel bugs
		5173	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		5174	although an acceptable workaround has been found here), or it might be
		5175	confused.
		5176
		5177	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		5178	make it into some kind of religion.
		5179
		5180	If you are unsure about something, feel free to contact the mailing list
		5181	with the full valgrind report and an explanation on why you think this
		5182	is a bug in libev (best check the archives, too :). However, don't be
		5183	annoyed when you get a brisk "this is no bug" answer and take the chance
		5184	of learning how to interpret valgrind properly.
		5185
		5186	If you need, for some reason, empty reports from valgrind for your project
		5187	I suggest using suppression lists.
		5188
		5189
		5190	=head1 PORTABILITY NOTES
		5191
		5192	=head2 GNU/LINUX 32 BIT LIMITATIONS
		5193
		5194	GNU/Linux is the only common platform that supports 64 bit file/large file
		5195	interfaces but I<disables> them by default.
		5196
		5197	That means that libev compiled in the default environment doesn't support
		5198	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		5199
		5200	Unfortunately, many programs try to work around this GNU/Linux issue
		5201	by enabling the large file API, which makes them incompatible with the
		5202	standard libev compiled for their system.
		5203
		5204	Likewise, libev cannot enable the large file API itself as this would
		5205	suddenly make it incompatible to the default compile time environment,
		5206	i.e. all programs not using special compile switches.
		5207
		5208	=head2 OS/X AND DARWIN BUGS
		5209
		5210	The whole thing is a bug if you ask me - basically any system interface
		5211	you touch is broken, whether it is locales, poll, kqueue or even the
		5212	OpenGL drivers.
		5213
		5214	=head3 C<kqueue> is buggy
		5215
		5216	The kqueue syscall is broken in all known versions - most versions support
		5217	only sockets, many support pipes.
		5218
		5219	Libev tries to work around this by not using C<kqueue> by default on this
		5220	rotten platform, but of course you can still ask for it when creating a
		5221	loop - embedding a socket-only kqueue loop into a select-based one is
		5222	probably going to work well.
		5223
		5224	=head3 C<poll> is buggy
		5225
		5226	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		5227	implementation by something calling C<kqueue> internally around the 10.5.6
		5228	release, so now C<kqueue> I<and> C<poll> are broken.
		5229
		5230	Libev tries to work around this by not using C<poll> by default on
		5231	this rotten platform, but of course you can still ask for it when creating
		5232	a loop.
		5233
		5234	=head3 C<select> is buggy
		5235
		5236	All that's left is C<select>, and of course Apple found a way to fuck this
		5237	one up as well: On OS/X, C<select> actively limits the number of file
		5238	descriptors you can pass in to 1024 - your program suddenly crashes when
		5239	you use more.
		5240
		5241	There is an undocumented "workaround" for this - defining
		5242	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		5243	work on OS/X.
		5244
		5245	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		5246
		5247	=head3 C<errno> reentrancy
		5248
		5249	The default compile environment on Solaris is unfortunately so
		5250	thread-unsafe that you can't even use components/libraries compiled
		5251	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		5252	defined by default. A valid, if stupid, implementation choice.
		5253
		5254	If you want to use libev in threaded environments you have to make sure
		5255	it's compiled with C<_REENTRANT> defined.
		5256
		5257	=head3 Event port backend
		5258
		5259	The scalable event interface for Solaris is called "event
		5260	ports". Unfortunately, this mechanism is very buggy in all major
		5261	releases. If you run into high CPU usage, your program freezes or you get
		5262	a large number of spurious wakeups, make sure you have all the relevant
		5263	and latest kernel patches applied. No, I don't know which ones, but there
		5264	are multiple ones to apply, and afterwards, event ports actually work
		5265	great.
		5266
		5267	If you can't get it to work, you can try running the program by setting
		5268	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		5269	C<select> backends.
		5270
		5271	=head2 AIX POLL BUG
		5272
		5273	AIX unfortunately has a broken C<poll.h> header. Libev works around
		5274	this by trying to avoid the poll backend altogether (i.e. it's not even
		5275	compiled in), which normally isn't a big problem as C<select> works fine
		5276	with large bitsets on AIX, and AIX is dead anyway.
		5277
		5278	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		5279
		5280	=head3 General issues
		5281
		5282	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		5283	requires, and its I/O model is fundamentally incompatible with the POSIX
		5284	model. Libev still offers limited functionality on this platform in
		5285	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		5286	descriptors. This only applies when using Win32 natively, not when using
		5287	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		5288	as every compiler comes with a slightly differently broken/incompatible
		5289	environment.
		5290
		5291	Lifting these limitations would basically require the full
		5292	re-implementation of the I/O system. If you are into this kind of thing,
		5293	then note that glib does exactly that for you in a very portable way (note
		5294	also that glib is the slowest event library known to man).
		5295
		5296	There is no supported compilation method available on windows except
		5297	embedding it into other applications.
		5298
		5299	Sensible signal handling is officially unsupported by Microsoft - libev
		5300	tries its best, but under most conditions, signals will simply not work.
		5301
		5302	Not a libev limitation but worth mentioning: windows apparently doesn't
		5303	accept large writes: instead of resulting in a partial write, windows will
		5304	either accept everything or return C<ENOBUFS> if the buffer is too large,
		5305	so make sure you only write small amounts into your sockets (less than a
		5306	megabyte seems safe, but this apparently depends on the amount of memory
		5307	available).
		5308
		5309	Due to the many, low, and arbitrary limits on the win32 platform and
		5310	the abysmal performance of winsockets, using a large number of sockets
		5311	is not recommended (and not reasonable). If your program needs to use
		5312	more than a hundred or so sockets, then likely it needs to use a totally
		5313	different implementation for windows, as libev offers the POSIX readiness
		5314	notification model, which cannot be implemented efficiently on windows
		5315	(due to Microsoft monopoly games).
		5316
		5317	A typical way to use libev under windows is to embed it (see the embedding
		5318	section for details) and use the following F<evwrap.h> header file instead
		5319	of F<ev.h>:
		5320
		5321	#define EV_STANDALONE /* keeps ev from requiring config.h */
		5322	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		5323
		5324	#include "ev.h"
		5325
		5326	And compile the following F<evwrap.c> file into your project (make sure
		5327	you do I<not> compile the F<ev.c> or any other embedded source files!):
		5328
		5329	#include "evwrap.h"
		5330	#include "ev.c"
		5331
		5332	=head3 The winsocket C<select> function
		5333
		5334	The winsocket C<select> function doesn't follow POSIX in that it
		5335	requires socket I<handles> and not socket I<file descriptors> (it is
		5336	also extremely buggy). This makes select very inefficient, and also
		5337	requires a mapping from file descriptors to socket handles (the Microsoft
		5338	C runtime provides the function C<_open_osfhandle> for this). See the
		5339	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		5340	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		5341
		5342	The configuration for a "naked" win32 using the Microsoft runtime
		5343	libraries and raw winsocket select is:
		5344
		5345	#define EV_USE_SELECT 1
		5346	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		5347
		5348	Note that winsockets handling of fd sets is O(n), so you can easily get a
		5349	complexity in the O(n²) range when using win32.
		5350
		5351	=head3 Limited number of file descriptors
		5352
		5353	Windows has numerous arbitrary (and low) limits on things.
		5354
		5355	Early versions of winsocket's select only supported waiting for a maximum
		5356	of C<64> handles (probably owning to the fact that all windows kernels
		5357	can only wait for C<64> things at the same time internally; Microsoft
		5358	recommends spawning a chain of threads and wait for 63 handles and the
		5359	previous thread in each. Sounds great!).
		5360
		5361	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		5362	to some high number (e.g. C<2048>) before compiling the winsocket select
		5363	call (which might be in libev or elsewhere, for example, perl and many
		5364	other interpreters do their own select emulation on windows).
		5365
		5366	Another limit is the number of file descriptors in the Microsoft runtime
		5367	libraries, which by default is C<64> (there must be a hidden I<64>
		5368	fetish or something like this inside Microsoft). You can increase this
		5369	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		5370	(another arbitrary limit), but is broken in many versions of the Microsoft
		5371	runtime libraries. This might get you to about C<512> or C<2048> sockets
		5372	(depending on windows version and/or the phase of the moon). To get more,
		5373	you need to wrap all I/O functions and provide your own fd management, but
		5374	the cost of calling select (O(n²)) will likely make this unworkable.
		5375
		5376	=head2 PORTABILITY REQUIREMENTS
		5377
		5378	In addition to a working ISO-C implementation and of course the
		5379	backend-specific APIs, libev relies on a few additional extensions:
		5380
		5381	=over 4
		5382
		5383	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		5384	calling conventions regardless of C<ev_watcher_type *>.
		5385
		5386	Libev assumes not only that all watcher pointers have the same internal
		5387	structure (guaranteed by POSIX but not by ISO C for example), but it also
		5388	assumes that the same (machine) code can be used to call any watcher
		5389	callback: The watcher callbacks have different type signatures, but libev
		5390	calls them using an C<ev_watcher *> internally.
		5391
		5392	=item null pointers and integer zero are represented by 0 bytes
		5393
		5394	Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
		5395	relies on this setting pointers and integers to null.
		5396
		5397	=item pointer accesses must be thread-atomic
		5398
		5399	Accessing a pointer value must be atomic, it must both be readable and
		5400	writable in one piece - this is the case on all current architectures.
		5401
		5402	=item C<sig_atomic_t volatile> must be thread-atomic as well
		5403
		5404	The type C<sig_atomic_t volatile> (or whatever is defined as
		5405	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		5406	threads. This is not part of the specification for C<sig_atomic_t>, but is
		5407	believed to be sufficiently portable.
		5408
		5409	=item C<sigprocmask> must work in a threaded environment
		5410
		5411	Libev uses C<sigprocmask> to temporarily block signals. This is not
		5412	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		5413	pthread implementations will either allow C<sigprocmask> in the "main
		5414	thread" or will block signals process-wide, both behaviours would
		5415	be compatible with libev. Interaction between C<sigprocmask> and
		5416	C<pthread_sigmask> could complicate things, however.
		5417
		5418	The most portable way to handle signals is to block signals in all threads
		5419	except the initial one, and run the signal handling loop in the initial
		5420	thread as well.
		5421
		5422	=item C<long> must be large enough for common memory allocation sizes
		5423
		5424	To improve portability and simplify its API, libev uses C<long> internally
		5425	instead of C<size_t> when allocating its data structures. On non-POSIX
		5426	systems (Microsoft...) this might be unexpectedly low, but is still at
		5427	least 31 bits everywhere, which is enough for hundreds of millions of
		5428	watchers.
		5429
		5430	=item C<double> must hold a time value in seconds with enough accuracy
		5431
		5432	The type C<double> is used to represent timestamps. It is required to
		5433	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		5434	good enough for at least into the year 4000 with millisecond accuracy
		5435	(the design goal for libev). This requirement is overfulfilled by
		5436	implementations using IEEE 754, which is basically all existing ones.
		5437
		5438	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5439	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5440	is either obsolete or somebody patched it to use C<long double> or
		5441	something like that, just kidding).
		5442
		5443	=back
		5444
		5445	If you know of other additional requirements drop me a note.
		5446
		5447
		5448	=head1 ALGORITHMIC COMPLEXITIES
		5449
		5450	In this section the complexities of (many of) the algorithms used inside
		5451	libev will be documented. For complexity discussions about backends see
		5452	the documentation for C<ev_default_init>.
		5453
		5454	All of the following are about amortised time: If an array needs to be
		5455	extended, libev needs to realloc and move the whole array, but this
		5456	happens asymptotically rarer with higher number of elements, so O(1) might
		5457	mean that libev does a lengthy realloc operation in rare cases, but on
		5458	average it is much faster and asymptotically approaches constant time.
		5459
		5460	=over 4
		5461
		5462	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		5463
		5464	This means that, when you have a watcher that triggers in one hour and
		5465	there are 100 watchers that would trigger before that, then inserting will
		5466	have to skip roughly seven (C<ld 100>) of these watchers.
		5467
		5468	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		5469
		5470	That means that changing a timer costs less than removing/adding them,
		5471	as only the relative motion in the event queue has to be paid for.
		5472
		5473	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		5474
		5475	These just add the watcher into an array or at the head of a list.
		5476
		5477	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		5478
		5479	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		5480
		5481	These watchers are stored in lists, so they need to be walked to find the
		5482	correct watcher to remove. The lists are usually short (you don't usually
		5483	have many watchers waiting for the same fd or signal: one is typical, two
		5484	is rare).
		5485
		5486	=item Finding the next timer in each loop iteration: O(1)
		5487
		5488	By virtue of using a binary or 4-heap, the next timer is always found at a
		5489	fixed position in the storage array.
		5490
		5491	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5492
		5493	A change means an I/O watcher gets started or stopped, which requires
		5494	libev to recalculate its status (and possibly tell the kernel, depending
		5495	on backend and whether C<ev_io_set> was used).
		5496
		5497	=item Activating one watcher (putting it into the pending state): O(1)
		5498
		5499	=item Priority handling: O(number_of_priorities)
		5500
		5501	Priorities are implemented by allocating some space for each
		5502	priority. When doing priority-based operations, libev usually has to
		5503	linearly search all the priorities, but starting/stopping and activating
		5504	watchers becomes O(1) with respect to priority handling.
		5505
		5506	=item Sending an ev_async: O(1)
		5507
		5508	=item Processing ev_async_send: O(number_of_async_watchers)
		5509
		5510	=item Processing signals: O(max_signal_number)
		5511
		5512	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5513	calls in the current loop iteration and the loop is currently
		5514	blocked. Checking for async and signal events involves iterating over all
		5515	running async watchers or all signal numbers.
		5516
		5517	=back
		5518
		5519
		5520	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5521
		5522	The major version 4 introduced some incompatible changes to the API.
		5523
		5524	At the moment, the C<ev.h> header file provides compatibility definitions
		5525	for all changes, so most programs should still compile. The compatibility
		5526	layer might be removed in later versions of libev, so better update to the
		5527	new API early than late.
		5528
		5529	=over 4
		5530
		5531	=item C<EV_COMPAT3> backwards compatibility mechanism
		5532
		5533	The backward compatibility mechanism can be controlled by
		5534	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
		5535	section.
		5536
		5537	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5538
		5539	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5540
		5541	ev_loop_destroy (EV_DEFAULT_UC);
		5542	ev_loop_fork (EV_DEFAULT);
		5543
		5544	=item function/symbol renames
		5545
		5546	A number of functions and symbols have been renamed:
		5547
		5548	ev_loop => ev_run
		5549	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5550	EVLOOP_ONESHOT => EVRUN_ONCE
		5551
		5552	ev_unloop => ev_break
		5553	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5554	EVUNLOOP_ONE => EVBREAK_ONE
		5555	EVUNLOOP_ALL => EVBREAK_ALL
		5556
		5557	EV_TIMEOUT => EV_TIMER
		5558
		5559	ev_loop_count => ev_iteration
		5560	ev_loop_depth => ev_depth
		5561	ev_loop_verify => ev_verify
		5562
		5563	Most functions working on C<struct ev_loop> objects don't have an
		5564	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5565	associated constants have been renamed to not collide with the C<struct
		5566	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5567	as all other watcher types. Note that C<ev_loop_fork> is still called
		5568	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5569	typedef.
		5570
		5571	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5572
		5573	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5574	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5575	and work, but the library code will of course be larger.
		5576
		5577	=back
		5578
		5579
		5580	=head1 GLOSSARY
		5581
		5582	=over 4
		5583
		5584	=item active
		5585
		5586	A watcher is active as long as it has been started and not yet stopped.
		5587	See L</WATCHER STATES> for details.
		5588
		5589	=item application
		5590
		5591	In this document, an application is whatever is using libev.
		5592
		5593	=item backend
		5594
		5595	The part of the code dealing with the operating system interfaces.
		5596
		5597	=item callback
		5598
		5599	The address of a function that is called when some event has been
		5600	detected. Callbacks are being passed the event loop, the watcher that
		5601	received the event, and the actual event bitset.
		5602
		5603	=item callback/watcher invocation
		5604
		5605	The act of calling the callback associated with a watcher.
		5606
		5607	=item event
		5608
		5609	A change of state of some external event, such as data now being available
		5610	for reading on a file descriptor, time having passed or simply not having
		5611	any other events happening anymore.
		5612
		5613	In libev, events are represented as single bits (such as C<EV_READ> or
		5614	C<EV_TIMER>).
		5615
		5616	=item event library
		5617
		5618	A software package implementing an event model and loop.
		5619
		5620	=item event loop
		5621
		5622	An entity that handles and processes external events and converts them
		5623	into callback invocations.
		5624
		5625	=item event model
		5626
		5627	The model used to describe how an event loop handles and processes
		5628	watchers and events.
		5629
		5630	=item pending
		5631
		5632	A watcher is pending as soon as the corresponding event has been
		5633	detected. See L</WATCHER STATES> for details.
		5634
		5635	=item real time
		5636
		5637	The physical time that is observed. It is apparently strictly monotonic :)
		5638
		5639	=item wall-clock time
		5640
		5641	The time and date as shown on clocks. Unlike real time, it can actually
		5642	be wrong and jump forwards and backwards, e.g. when you adjust your
		5643	clock.
		5644
		5645	=item watcher
		5646
		5647	A data structure that describes interest in certain events. Watchers need
		5648	to be started (attached to an event loop) before they can receive events.
		5649
		5650	=back
825		5651
826	=head1 AUTHOR	5652	=head1 AUTHOR
827		5653
828	Marc Lehmann <libev@schmorp.de>.	5654	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5655	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
829		5656

Diff Legend

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