ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.8 by root, Mon Nov 12 08:20:02 2007 UTC vs.
Revision 1.449 by root, Sun Jun 23 02:02:30 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 AND OTHER GLOBAL FUNCTIONS	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
		168
		169	=head1 GLOBAL FUNCTIONS
		170
		171	These functions can be called anytime, even before initialising the
		172	library in any way.
51		173
52	=over 4	174	=over 4
53		175
54	=item ev_tstamp ev_time ()	176	=item ev_tstamp ev_time ()
55		177
56	Returns the current time as libev would use it.	178	Returns the current time as libev would use it. Please note that the
		179	C<ev_now> function is usually faster and also often returns the timestamp
		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 >>).
57		194
58	=item int ev_version_major ()	195	=item int ev_version_major ()
59		196
60	=item int ev_version_minor ()	197	=item int ev_version_minor ()
61		198
62	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
63	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
64	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
65	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
66	version of the library your program was compiled against.	203	version of the library your program was compiled against.
67		204
		205	These version numbers refer to the ABI version of the library, not the
		206	release version.
		207
68	Usually, its a good idea to terminate if the major versions mismatch,	208	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	209	as this indicates an incompatible change. Minor versions are usually
70	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
71	not a problem.	211	not a problem.
72		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
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	254	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
74		255
75	Sets the allocation function to use (the prototype is similar to the	256	Sets the allocation function to use (the prototype is similar - the
76	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
77	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
78	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
79	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.
80		265
81	You could override this function in high-availability programs to, say,	266	You could override this function in high-availability programs to, say,
82	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,
83	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.
84		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
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	310	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
86		311
87	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
88	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
89	indicating the system call or subsystem causing the problem. If this	314	indicating the system call or subsystem causing the problem. If this
90	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
91	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
92	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
93	(such as abort).	318	(such as abort).
94		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
95	=back	345	=back
96		346
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	347	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
98		348
99	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
100	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
101	events, and dynamically created loops which do not.	351	libev 3 had an C<ev_loop> function colliding with the struct name).
102		352
103	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
104	in your main thread (or in a separate thrad) and for each thread you	354	supports child process events, and dynamically created event loops which
105	create, you also create another event loop. Libev itself does no locking	355	do not.
106	whatsoever, so if you mix calls to the same event loop in different
107	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because its hideous and inefficient).
109		356
110	=over 4	357	=over 4
111		358
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	359	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		360
114	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
115	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
116	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
117	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".
118		371
119	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
120	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.
121		404
122	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
123	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>).
124		407
125	It supports the following flags:	408	The following flags are supported:
126		409
127	=over 4	410	=over 4
128		411
129	=item EVFLAG_AUTO	412	=item C<EVFLAG_AUTO>
130		413
131	The default flags value. Use this if you have no clue (its the right	414	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	415	thing, believe me).
133		416
134	=item EVFLAG_NOENV	417	=item C<EVFLAG_NOENV>
135		418
136	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
137	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
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	421	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	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
140	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
141	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).
142		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
143	=item EVMETHOD_SELECT portable select backend	483	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		484
145	=item EVMETHOD_POLL poll backend (everywhere except windows)	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.
146		490
147	=item EVMETHOD_EPOLL linux only	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.
148		497
149	=item EVMETHOD_KQUEUE some bsds only	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).
150		501
151	=item EVMETHOD_DEVPOLL solaris 8 only	502	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
152		503
153	=item EVMETHOD_PORT solaris 10 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.
154		510
155	If one or more of these are ored into the flags value, then only these	511	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
156	backends will be tried (in the reverse order as given here). If one are	512	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157	specified, any backend will do.	513
		514	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		515
		516	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		517	kernels).
		518
		519	For few fds, this backend is a bit little slower than poll and select, but
		520	it scales phenomenally better. While poll and select usually scale like
		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) >> but C<<
		580	io_submit(2) >>) event interface available in post-4.18 kernels.
		581
		582	If this backend works for you (as of this writing, it was very
		583	experimental), it is the best event interface available on linux and might
		584	be well worth enabling it - if it isn't available in your kernel this will
		585	be detected and this backend will be skipped.
		586
		587	This backend can batch oneshot requests and supports a user-space ring
		588	buffer to receive events. It also doesn't suffer from most of the design
		589	problems of epoll (such as not being able to remove event sources from
		590	the epoll set), and generally sounds too good to be true. Because, this
		591	being the linux kernel, of course it suffers from a whole new set of
		592	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 a system wide
		596	limit that can be configured in F</proc/sys/fs/aio-max-nr> - each loop
		597	currently requires C<61> of this number. If no aio requests are left, this
		598	backend will be skipped during initialisation.
		599
		600	Most problematic in practise, however, is that not all file descriptors
		601	work with it. For example, in linux 5.1, tcp sockets, pipes, event fds,
		602	files, F</dev/null> and a few others are supported, but ttys do not work
		603	(probably because of a bug), so this is not (yet?) a generic event polling
		604	interface.
		605
		606	To work around this latter problem, the current version of libev uses
		607	epoll as a fallback for file deescriptor types that do not work. Epoll
		608	is used in, kind of, slow mode that hopefully avoids most of its design
		609	problems and requires 1-3 extra syscalls per active fd every iteration.
		610
		611	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		612	C<EVBACKEND_POLL>.
		613
		614	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		615
		616	Kqueue deserves special mention, as at the time of this writing, it
		617	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		618	with anything but sockets and pipes, except on Darwin, where of course
		619	it's completely useless). Unlike epoll, however, whose brokenness
		620	is by design, these kqueue bugs can (and eventually will) be fixed
		621	without API changes to existing programs. For this reason it's not being
		622	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		623	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		624	system like NetBSD.
		625
		626	You still can embed kqueue into a normal poll or select backend and use it
		627	only for sockets (after having made sure that sockets work with kqueue on
		628	the target platform). See C<ev_embed> watchers for more info.
		629
		630	It scales in the same way as the epoll backend, but the interface to the
		631	kernel is more efficient (which says nothing about its actual speed, of
		632	course). While stopping, setting and starting an I/O watcher does never
		633	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		634	two event changes per incident. Support for C<fork ()> is very bad (you
		635	might have to leak fd's on fork, but it's more sane than epoll) and it
		636	drops fds silently in similarly hard-to-detect cases.
		637
		638	This backend usually performs well under most conditions.
		639
		640	While nominally embeddable in other event loops, this doesn't work
		641	everywhere, so you might need to test for this. And since it is broken
		642	almost everywhere, you should only use it when you have a lot of sockets
		643	(for which it usually works), by embedding it into another event loop
		644	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		645	also broken on OS X)) and, did I mention it, using it only for sockets.
		646
		647	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		648	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		649	C<NOTE_EOF>.
		650
		651	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		652
		653	This is not implemented yet (and might never be, unless you send me an
		654	implementation). According to reports, C</dev/poll> only supports sockets
		655	and is not embeddable, which would limit the usefulness of this backend
		656	immensely.
		657
		658	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		659
		660	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		661	it's really slow, but it still scales very well (O(active_fds)).
		662
		663	While this backend scales well, it requires one system call per active
		664	file descriptor per loop iteration. For small and medium numbers of file
		665	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		666	might perform better.
		667
		668	On the positive side, this backend actually performed fully to
		669	specification in all tests and is fully embeddable, which is a rare feat
		670	among the OS-specific backends (I vastly prefer correctness over speed
		671	hacks).
		672
		673	On the negative side, the interface is I<bizarre> - so bizarre that
		674	even sun itself gets it wrong in their code examples: The event polling
		675	function sometimes returns events to the caller even though an error
		676	occurred, but with no indication whether it has done so or not (yes, it's
		677	even documented that way) - deadly for edge-triggered interfaces where you
		678	absolutely have to know whether an event occurred or not because you have
		679	to re-arm the watcher.
		680
		681	Fortunately libev seems to be able to work around these idiocies.
		682
		683	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		684	C<EVBACKEND_POLL>.
		685
		686	=item C<EVBACKEND_ALL>
		687
		688	Try all backends (even potentially broken ones that wouldn't be tried
		689	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		690	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		691
		692	It is definitely not recommended to use this flag, use whatever
		693	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		694	at all.
		695
		696	=item C<EVBACKEND_MASK>
		697
		698	Not a backend at all, but a mask to select all backend bits from a
		699	C<flags> value, in case you want to mask out any backends from a flags
		700	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
158		701
159	=back	702	=back
160		703
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	704	If one or more of the backend flags are or'ed into the flags value,
		705	then only these backends will be tried (in the reverse order as listed
		706	here). If none are specified, all backends in C<ev_recommended_backends
		707	()> will be tried.
162		708
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	709	Example: Try to create a event loop that uses epoll and nothing else.
164	always distinct from the default loop. Unlike the default loop, it cannot
165	handle signal and child watchers, and attempts to do so will be greeted by
166	undefined behaviour (or a failed assertion if assertions are enabled).
167		710
168	=item ev_default_destroy ()	711	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		712	if (!epoller)
		713	fatal ("no epoll found here, maybe it hides under your chair");
169		714
170	Destroys the default loop again (frees all memory and kernel state	715	Example: Use whatever libev has to offer, but make sure that kqueue is
171	etc.). This stops all registered event watchers (by not touching them in	716	used if available.
172	any way whatsoever, although you cnanot rely on this :).	717
		718	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		719
		720	Example: Similarly, on linux, you mgiht want to take advantage of the
		721	linux aio backend if possible, but fall back to something else if that
		722	isn't available.
		723
		724	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
173		725
174	=item ev_loop_destroy (loop)	726	=item ev_loop_destroy (loop)
175		727
176	Like C<ev_default_destroy>, but destroys an event loop created by an	728	Destroys an event loop object (frees all memory and kernel state
177	earlier call to C<ev_loop_new>.	729	etc.). None of the active event watchers will be stopped in the normal
		730	sense, so e.g. C<ev_is_active> might still return true. It is your
		731	responsibility to either stop all watchers cleanly yourself I<before>
		732	calling this function, or cope with the fact afterwards (which is usually
		733	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		734	for example).
178		735
179	=item ev_default_fork ()	736	Note that certain global state, such as signal state (and installed signal
		737	handlers), will not be freed by this function, and related watchers (such
		738	as signal and child watchers) would need to be stopped manually.
180		739
181	This function reinitialises the kernel state for backends that have	740	This function is normally used on loop objects allocated by
182	one. Despite the name, you can call it anytime, but it makes most sense	741	C<ev_loop_new>, but it can also be used on the default loop returned by
183	after forking, in either the parent or child process (or both, but that	742	C<ev_default_loop>, in which case it is not thread-safe.
184	again makes little sense).
185		743
186	You I<must> call this function after forking if and only if you want to	744	Note that it is not advisable to call this function on the default loop
187	use the event library in both processes. If you just fork+exec, you don't	745	except in the rare occasion where you really need to free its resources.
188	have to call it.	746	If you need dynamically allocated loops it is better to use C<ev_loop_new>
189		747	and C<ev_loop_destroy>.
190	The function itself is quite fast and its usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in
192	quite nicely into a call to C<pthread_atfork>:
193
194	pthread_atfork (0, 0, ev_default_fork);
195		748
196	=item ev_loop_fork (loop)	749	=item ev_loop_fork (loop)
197		750
198	Like C<ev_default_fork>, but acts on an event loop created by	751	This function sets a flag that causes subsequent C<ev_run> iterations
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	752	to reinitialise the kernel state for backends that have one. Despite
200	after fork, and how you do this is entirely your own problem.	753	the name, you can call it anytime you are allowed to start or stop
		754	watchers (except inside an C<ev_prepare> callback), but it makes most
		755	sense after forking, in the child process. You I<must> call it (or use
		756	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
201		757
		758	In addition, if you want to reuse a loop (via this function or
		759	C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
		760
		761	Again, you I<have> to call it on I<any> loop that you want to re-use after
		762	a fork, I<even if you do not plan to use the loop in the parent>. This is
		763	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		764	during fork.
		765
		766	On the other hand, you only need to call this function in the child
		767	process if and only if you want to use the event loop in the child. If
		768	you just fork+exec or create a new loop in the child, you don't have to
		769	call it at all (in fact, C<epoll> is so badly broken that it makes a
		770	difference, but libev will usually detect this case on its own and do a
		771	costly reset of the backend).
		772
		773	The function itself is quite fast and it's usually not a problem to call
		774	it just in case after a fork.
		775
		776	Example: Automate calling C<ev_loop_fork> on the default loop when
		777	using pthreads.
		778
		779	static void
		780	post_fork_child (void)
		781	{
		782	ev_loop_fork (EV_DEFAULT);
		783	}
		784
		785	...
		786	pthread_atfork (0, 0, post_fork_child);
		787
		788	=item int ev_is_default_loop (loop)
		789
		790	Returns true when the given loop is, in fact, the default loop, and false
		791	otherwise.
		792
		793	=item unsigned int ev_iteration (loop)
		794
		795	Returns the current iteration count for the event loop, which is identical
		796	to the number of times libev did poll for new events. It starts at C<0>
		797	and happily wraps around with enough iterations.
		798
		799	This value can sometimes be useful as a generation counter of sorts (it
		800	"ticks" the number of loop iterations), as it roughly corresponds with
		801	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		802	prepare and check phases.
		803
202	=item unsigned int ev_method (loop)	804	=item unsigned int ev_depth (loop)
203		805
		806	Returns the number of times C<ev_run> was entered minus the number of
		807	times C<ev_run> was exited normally, in other words, the recursion depth.
		808
		809	Outside C<ev_run>, this number is zero. In a callback, this number is
		810	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		811	in which case it is higher.
		812
		813	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		814	throwing an exception etc.), doesn't count as "exit" - consider this
		815	as a hint to avoid such ungentleman-like behaviour unless it's really
		816	convenient, in which case it is fully supported.
		817
		818	=item unsigned int ev_backend (loop)
		819
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	820	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	821	use.
206		822
207	=item ev_tstamp = ev_now (loop)	823	=item ev_tstamp ev_now (loop)
208		824
209	Returns the current "event loop time", which is the time the event loop	825	Returns the current "event loop time", which is the time the event loop
210	got events and started processing them. This timestamp does not change	826	received events and started processing them. This timestamp does not
211	as long as callbacks are being processed, and this is also the base time	827	change as long as callbacks are being processed, and this is also the base
212	used for relative timers. You can treat it as the timestamp of the event	828	time used for relative timers. You can treat it as the timestamp of the
213	occuring (or more correctly, the mainloop finding out about it).	829	event occurring (or more correctly, libev finding out about it).
214		830
		831	=item ev_now_update (loop)
		832
		833	Establishes the current time by querying the kernel, updating the time
		834	returned by C<ev_now ()> in the progress. This is a costly operation and
		835	is usually done automatically within C<ev_run ()>.
		836
		837	This function is rarely useful, but when some event callback runs for a
		838	very long time without entering the event loop, updating libev's idea of
		839	the current time is a good idea.
		840
		841	See also L</The special problem of time updates> in the C<ev_timer> section.
		842
		843	=item ev_suspend (loop)
		844
		845	=item ev_resume (loop)
		846
		847	These two functions suspend and resume an event loop, for use when the
		848	loop is not used for a while and timeouts should not be processed.
		849
		850	A typical use case would be an interactive program such as a game: When
		851	the user presses C<^Z> to suspend the game and resumes it an hour later it
		852	would be best to handle timeouts as if no time had actually passed while
		853	the program was suspended. This can be achieved by calling C<ev_suspend>
		854	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		855	C<ev_resume> directly afterwards to resume timer processing.
		856
		857	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		858	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		859	will be rescheduled (that is, they will lose any events that would have
		860	occurred while suspended).
		861
		862	After calling C<ev_suspend> you B<must not> call I<any> function on the
		863	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		864	without a previous call to C<ev_suspend>.
		865
		866	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		867	event loop time (see C<ev_now_update>).
		868
215	=item ev_loop (loop, int flags)	869	=item bool ev_run (loop, int flags)
216		870
217	Finally, this is it, the event handler. This function usually is called	871	Finally, this is it, the event handler. This function usually is called
218	after you initialised all your watchers and you want to start handling	872	after you have initialised all your watchers and you want to start
219	events.	873	handling events. It will ask the operating system for any new events, call
		874	the watcher callbacks, and then repeat the whole process indefinitely: This
		875	is why event loops are called I<loops>.
220		876
221	If the flags argument is specified as 0, it will not return until either	877	If the flags argument is specified as C<0>, it will keep handling events
222	no event watchers are active anymore or C<ev_unloop> was called.	878	until either no event watchers are active anymore or C<ev_break> was
		879	called.
223		880
		881	The return value is false if there are no more active watchers (which
		882	usually means "all jobs done" or "deadlock"), and true in all other cases
		883	(which usually means " you should call C<ev_run> again").
		884
		885	Please note that an explicit C<ev_break> is usually better than
		886	relying on all watchers to be stopped when deciding when a program has
		887	finished (especially in interactive programs), but having a program
		888	that automatically loops as long as it has to and no longer by virtue
		889	of relying on its watchers stopping correctly, that is truly a thing of
		890	beauty.
		891
		892	This function is I<mostly> exception-safe - you can break out of a
		893	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		894	exception and so on. This does not decrement the C<ev_depth> value, nor
		895	will it clear any outstanding C<EVBREAK_ONE> breaks.
		896
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	897	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
225	those events and any outstanding ones, but will not block your process in	898	those events and any already outstanding ones, but will not wait and
226	case there are no events.	899	block your process in case there are no events and will return after one
		900	iteration of the loop. This is sometimes useful to poll and handle new
		901	events while doing lengthy calculations, to keep the program responsive.
227		902
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	903	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
229	neccessary) and will handle those and any outstanding ones. It will block	904	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives.	905	will block your process until at least one new event arrives (which could
		906	be an event internal to libev itself, so there is no guarantee that a
		907	user-registered callback will be called), and will return after one
		908	iteration of the loop.
231		909
232	This flags value could be used to implement alternative looping	910	This is useful if you are waiting for some external event in conjunction
233	constructs, but the C<prepare> and C<check> watchers provide a better and	911	with something not expressible using other libev watchers (i.e. "roll your
234	more generic mechanism.	912	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		913	usually a better approach for this kind of thing.
235		914
		915	Here are the gory details of what C<ev_run> does (this is for your
		916	understanding, not a guarantee that things will work exactly like this in
		917	future versions):
		918
		919	- Increment loop depth.
		920	- Reset the ev_break status.
		921	- Before the first iteration, call any pending watchers.
		922	LOOP:
		923	- If EVFLAG_FORKCHECK was used, check for a fork.
		924	- If a fork was detected (by any means), queue and call all fork watchers.
		925	- Queue and call all prepare watchers.
		926	- If ev_break was called, goto FINISH.
		927	- If we have been forked, detach and recreate the kernel state
		928	as to not disturb the other process.
		929	- Update the kernel state with all outstanding changes.
		930	- Update the "event loop time" (ev_now ()).
		931	- Calculate for how long to sleep or block, if at all
		932	(active idle watchers, EVRUN_NOWAIT or not having
		933	any active watchers at all will result in not sleeping).
		934	- Sleep if the I/O and timer collect interval say so.
		935	- Increment loop iteration counter.
		936	- Block the process, waiting for any events.
		937	- Queue all outstanding I/O (fd) events.
		938	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		939	- Queue all expired timers.
		940	- Queue all expired periodics.
		941	- Queue all idle watchers with priority higher than that of pending events.
		942	- Queue all check watchers.
		943	- Call all queued watchers in reverse order (i.e. check watchers first).
		944	Signals and child watchers are implemented as I/O watchers, and will
		945	be handled here by queueing them when their watcher gets executed.
		946	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		947	were used, or there are no active watchers, goto FINISH, otherwise
		948	continue with step LOOP.
		949	FINISH:
		950	- Reset the ev_break status iff it was EVBREAK_ONE.
		951	- Decrement the loop depth.
		952	- Return.
		953
		954	Example: Queue some jobs and then loop until no events are outstanding
		955	anymore.
		956
		957	... queue jobs here, make sure they register event watchers as long
		958	... as they still have work to do (even an idle watcher will do..)
		959	ev_run (my_loop, 0);
		960	... jobs done or somebody called break. yeah!
		961
236	=item ev_unloop (loop, how)	962	=item ev_break (loop, how)
237		963
238	Can be used to make a call to C<ev_loop> return early. The C<how> argument	964	Can be used to make a call to C<ev_run> return early (but only after it
239	must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop>	965	has processed all outstanding events). The C<how> argument must be either
240	call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop>	966	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
241	calls return.	967	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		968
		969	This "break state" will be cleared on the next call to C<ev_run>.
		970
		971	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		972	which case it will have no effect.
242		973
243	=item ev_ref (loop)	974	=item ev_ref (loop)
244		975
245	=item ev_unref (loop)	976	=item ev_unref (loop)
246		977
247	Ref/unref can be used to add or remove a refcount on the event loop: Every	978	Ref/unref can be used to add or remove a reference count on the event
248	watcher keeps one reference. If you have a long-runing watcher you never	979	loop: Every watcher keeps one reference, and as long as the reference
249	unregister that should not keep ev_loop from running, ev_unref() after	980	count is nonzero, C<ev_run> will not return on its own.
250	starting, and ev_ref() before stopping it. Libev itself uses this for	981
251	example for its internal signal pipe: It is not visible to you as a user	982	This is useful when you have a watcher that you never intend to
252	and should not keep C<ev_loop> from exiting if the work is done. It is	983	unregister, but that nevertheless should not keep C<ev_run> from
		984	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		985	before stopping it.
		986
		987	As an example, libev itself uses this for its internal signal pipe: It
		988	is not visible to the libev user and should not keep C<ev_run> from
		989	exiting if no event watchers registered by it are active. It is also an
253	also an excellent way to do this for generic recurring timers or from	990	excellent way to do this for generic recurring timers or from within
254	within third-party libraries. Just remember to unref after start and ref	991	third-party libraries. Just remember to I<unref after start> and I<ref
255	before stop.	992	before stop> (but only if the watcher wasn't active before, or was active
		993	before, respectively. Note also that libev might stop watchers itself
		994	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		995	in the callback).
		996
		997	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		998	running when nothing else is active.
		999
		1000	ev_signal exitsig;
		1001	ev_signal_init (&exitsig, sig_cb, SIGINT);
		1002	ev_signal_start (loop, &exitsig);
		1003	ev_unref (loop);
		1004
		1005	Example: For some weird reason, unregister the above signal handler again.
		1006
		1007	ev_ref (loop);
		1008	ev_signal_stop (loop, &exitsig);
		1009
		1010	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		1011
		1012	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		1013
		1014	These advanced functions influence the time that libev will spend waiting
		1015	for events. Both time intervals are by default C<0>, meaning that libev
		1016	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		1017	latency.
		1018
		1019	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		1020	allows libev to delay invocation of I/O and timer/periodic callbacks
		1021	to increase efficiency of loop iterations (or to increase power-saving
		1022	opportunities).
		1023
		1024	The idea is that sometimes your program runs just fast enough to handle
		1025	one (or very few) event(s) per loop iteration. While this makes the
		1026	program responsive, it also wastes a lot of CPU time to poll for new
		1027	events, especially with backends like C<select ()> which have a high
		1028	overhead for the actual polling but can deliver many events at once.
		1029
		1030	By setting a higher I<io collect interval> you allow libev to spend more
		1031	time collecting I/O events, so you can handle more events per iteration,
		1032	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		1033	C<ev_timer>) will not be affected. Setting this to a non-null value will
		1034	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		1035	sleep time ensures that libev will not poll for I/O events more often then
		1036	once per this interval, on average (as long as the host time resolution is
		1037	good enough).
		1038
		1039	Likewise, by setting a higher I<timeout collect interval> you allow libev
		1040	to spend more time collecting timeouts, at the expense of increased
		1041	latency/jitter/inexactness (the watcher callback will be called
		1042	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		1043	value will not introduce any overhead in libev.
		1044
		1045	Many (busy) programs can usually benefit by setting the I/O collect
		1046	interval to a value near C<0.1> or so, which is often enough for
		1047	interactive servers (of course not for games), likewise for timeouts. It
		1048	usually doesn't make much sense to set it to a lower value than C<0.01>,
		1049	as this approaches the timing granularity of most systems. Note that if
		1050	you do transactions with the outside world and you can't increase the
		1051	parallelity, then this setting will limit your transaction rate (if you
		1052	need to poll once per transaction and the I/O collect interval is 0.01,
		1053	then you can't do more than 100 transactions per second).
		1054
		1055	Setting the I<timeout collect interval> can improve the opportunity for
		1056	saving power, as the program will "bundle" timer callback invocations that
		1057	are "near" in time together, by delaying some, thus reducing the number of
		1058	times the process sleeps and wakes up again. Another useful technique to
		1059	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		1060	they fire on, say, one-second boundaries only.
		1061
		1062	Example: we only need 0.1s timeout granularity, and we wish not to poll
		1063	more often than 100 times per second:
		1064
		1065	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		1066	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		1067
		1068	=item ev_invoke_pending (loop)
		1069
		1070	This call will simply invoke all pending watchers while resetting their
		1071	pending state. Normally, C<ev_run> does this automatically when required,
		1072	but when overriding the invoke callback this call comes handy. This
		1073	function can be invoked from a watcher - this can be useful for example
		1074	when you want to do some lengthy calculation and want to pass further
		1075	event handling to another thread (you still have to make sure only one
		1076	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		1077
		1078	=item int ev_pending_count (loop)
		1079
		1080	Returns the number of pending watchers - zero indicates that no watchers
		1081	are pending.
		1082
		1083	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1084
		1085	This overrides the invoke pending functionality of the loop: Instead of
		1086	invoking all pending watchers when there are any, C<ev_run> will call
		1087	this callback instead. This is useful, for example, when you want to
		1088	invoke the actual watchers inside another context (another thread etc.).
		1089
		1090	If you want to reset the callback, use C<ev_invoke_pending> as new
		1091	callback.
		1092
		1093	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
		1094
		1095	Sometimes you want to share the same loop between multiple threads. This
		1096	can be done relatively simply by putting mutex_lock/unlock calls around
		1097	each call to a libev function.
		1098
		1099	However, C<ev_run> can run an indefinite time, so it is not feasible
		1100	to wait for it to return. One way around this is to wake up the event
		1101	loop via C<ev_break> and C<ev_async_send>, another way is to set these
		1102	I<release> and I<acquire> callbacks on the loop.
		1103
		1104	When set, then C<release> will be called just before the thread is
		1105	suspended waiting for new events, and C<acquire> is called just
		1106	afterwards.
		1107
		1108	Ideally, C<release> will just call your mutex_unlock function, and
		1109	C<acquire> will just call the mutex_lock function again.
		1110
		1111	While event loop modifications are allowed between invocations of
		1112	C<release> and C<acquire> (that's their only purpose after all), no
		1113	modifications done will affect the event loop, i.e. adding watchers will
		1114	have no effect on the set of file descriptors being watched, or the time
		1115	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1116	to take note of any changes you made.
		1117
		1118	In theory, threads executing C<ev_run> will be async-cancel safe between
		1119	invocations of C<release> and C<acquire>.
		1120
		1121	See also the locking example in the C<THREADS> section later in this
		1122	document.
		1123
		1124	=item ev_set_userdata (loop, void *data)
		1125
		1126	=item void *ev_userdata (loop)
		1127
		1128	Set and retrieve a single C<void *> associated with a loop. When
		1129	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1130	C<0>.
		1131
		1132	These two functions can be used to associate arbitrary data with a loop,
		1133	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1134	C<acquire> callbacks described above, but of course can be (ab-)used for
		1135	any other purpose as well.
		1136
		1137	=item ev_verify (loop)
		1138
		1139	This function only does something when C<EV_VERIFY> support has been
		1140	compiled in, which is the default for non-minimal builds. It tries to go
		1141	through all internal structures and checks them for validity. If anything
		1142	is found to be inconsistent, it will print an error message to standard
		1143	error and call C<abort ()>.
		1144
		1145	This can be used to catch bugs inside libev itself: under normal
		1146	circumstances, this function will never abort as of course libev keeps its
		1147	data structures consistent.
256		1148
257	=back	1149	=back
258		1150
		1151
259	=head1 ANATOMY OF A WATCHER	1152	=head1 ANATOMY OF A WATCHER
260		1153
		1154	In the following description, uppercase C<TYPE> in names stands for the
		1155	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1156	watchers and C<ev_io_start> for I/O watchers.
		1157
261	A watcher is a structure that you create and register to record your	1158	A watcher is an opaque structure that you allocate and register to record
262	interest in some event. For instance, if you want to wait for STDIN to	1159	your interest in some event. To make a concrete example, imagine you want
263	become readable, you would create an ev_io watcher for that:	1160	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1161	for that:
264		1162
265	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1163	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
266	{	1164	{
267	ev_io_stop (w);	1165	ev_io_stop (w);
268	ev_unloop (loop, EVUNLOOP_ALL);	1166	ev_break (loop, EVBREAK_ALL);
269	}	1167	}
270		1168
271	struct ev_loop *loop = ev_default_loop (0);	1169	struct ev_loop *loop = ev_default_loop (0);
		1170
272	struct ev_io stdin_watcher;	1171	ev_io stdin_watcher;
		1172
273	ev_init (&stdin_watcher, my_cb);	1173	ev_init (&stdin_watcher, my_cb);
274	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1174	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
275	ev_io_start (loop, &stdin_watcher);	1175	ev_io_start (loop, &stdin_watcher);
		1176
276	ev_loop (loop, 0);	1177	ev_run (loop, 0);
277		1178
278	As you can see, you are responsible for allocating the memory for your	1179	As you can see, you are responsible for allocating the memory for your
279	watcher structures (and it is usually a bad idea to do this on the stack,	1180	watcher structures (and it is I<usually> a bad idea to do this on the
280	although this can sometimes be quite valid).	1181	stack).
281		1182
		1183	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1184	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1185
282	Each watcher structure must be initialised by a call to C<ev_init	1186	Each watcher structure must be initialised by a call to C<ev_init (watcher
283	(watcher *, callback)>, which expects a callback to be provided. This	1187	*, callback)>, which expects a callback to be provided. This callback is
284	callback gets invoked each time the event occurs (or, in the case of io	1188	invoked each time the event occurs (or, in the case of I/O watchers, each
285	watchers, each time the event loop detects that the file descriptor given	1189	time the event loop detects that the file descriptor given is readable
286	is readable and/or writable).	1190	and/or writable).
287		1191
288	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1192	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
289	with arguments specific to this watcher type. There is also a macro	1193	macro to configure it, with arguments specific to the watcher type. There
290	to combine initialisation and setting in one call: C<< ev_<type>_init	1194	is also a macro to combine initialisation and setting in one call: C<<
291	(watcher *, callback, ...) >>.	1195	ev_TYPE_init (watcher *, callback, ...) >>.
292		1196
293	To make the watcher actually watch out for events, you have to start it	1197	To make the watcher actually watch out for events, you have to start it
294	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1198	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
295	*) >>), and you can stop watching for events at any time by calling the	1199	*) >>), and you can stop watching for events at any time by calling the
296	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1200	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
297		1201
298	As long as your watcher is active (has been started but not stopped) you	1202	As long as your watcher is active (has been started but not stopped) you
299	must not touch the values stored in it. Most specifically you must never	1203	must not touch the values stored in it. Most specifically you must never
300	reinitialise it or call its set method.	1204	reinitialise it or call its C<ev_TYPE_set> macro.
301
302	You cna check whether an event is active by calling the C<ev_is_active
303	(watcher *)> macro. To see whether an event is outstanding (but the
304	callback for it has not been called yet) you cna use the C<ev_is_pending
305	(watcher *)> macro.
306		1205
307	Each and every callback receives the event loop pointer as first, the	1206	Each and every callback receives the event loop pointer as first, the
308	registered watcher structure as second, and a bitset of received events as	1207	registered watcher structure as second, and a bitset of received events as
309	third argument.	1208	third argument.
310		1209
311	The rceeived events usually include a single bit per event type received	1210	The received events usually include a single bit per event type received
312	(you can receive multiple events at the same time). The possible bit masks	1211	(you can receive multiple events at the same time). The possible bit masks
313	are:	1212	are:
314		1213
315	=over 4	1214	=over 4
316		1215
317	=item EV_READ	1216	=item C<EV_READ>
318		1217
319	=item EV_WRITE	1218	=item C<EV_WRITE>
320		1219
321	The file descriptor in the ev_io watcher has become readable and/or	1220	The file descriptor in the C<ev_io> watcher has become readable and/or
322	writable.	1221	writable.
323		1222
324	=item EV_TIMEOUT	1223	=item C<EV_TIMER>
325		1224
326	The ev_timer watcher has timed out.	1225	The C<ev_timer> watcher has timed out.
327		1226
328	=item EV_PERIODIC	1227	=item C<EV_PERIODIC>
329		1228
330	The ev_periodic watcher has timed out.	1229	The C<ev_periodic> watcher has timed out.
331		1230
332	=item EV_SIGNAL	1231	=item C<EV_SIGNAL>
333		1232
334	The signal specified in the ev_signal watcher has been received by a thread.	1233	The signal specified in the C<ev_signal> watcher has been received by a thread.
335		1234
336	=item EV_CHILD	1235	=item C<EV_CHILD>
337		1236
338	The pid specified in the ev_child watcher has received a status change.	1237	The pid specified in the C<ev_child> watcher has received a status change.
339		1238
		1239	=item C<EV_STAT>
		1240
		1241	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1242
340	=item EV_IDLE	1243	=item C<EV_IDLE>
341		1244
342	The ev_idle watcher has determined that you have nothing better to do.	1245	The C<ev_idle> watcher has determined that you have nothing better to do.
343		1246
344	=item EV_PREPARE	1247	=item C<EV_PREPARE>
345		1248
346	=item EV_CHECK	1249	=item C<EV_CHECK>
347		1250
348	All ev_prepare watchers are invoked just I<before> C<ev_loop> starts	1251	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
349	to gather new events, and all ev_check watchers are invoked just after	1252	gather new events, and all C<ev_check> watchers are queued (not invoked)
350	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1253	just after C<ev_run> has gathered them, but before it queues any callbacks
		1254	for any received events. That means C<ev_prepare> watchers are the last
		1255	watchers invoked before the event loop sleeps or polls for new events, and
		1256	C<ev_check> watchers will be invoked before any other watchers of the same
		1257	or lower priority within an event loop iteration.
		1258
351	received events. Callbacks of both watcher types can start and stop as	1259	Callbacks of both watcher types can start and stop as many watchers as
352	many watchers as they want, and all of them will be taken into account	1260	they want, and all of them will be taken into account (for example, a
353	(for example, a ev_prepare watcher might start an idle watcher to keep	1261	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
354	C<ev_loop> from blocking).	1262	blocking).
355		1263
		1264	=item C<EV_EMBED>
		1265
		1266	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1267
		1268	=item C<EV_FORK>
		1269
		1270	The event loop has been resumed in the child process after fork (see
		1271	C<ev_fork>).
		1272
		1273	=item C<EV_CLEANUP>
		1274
		1275	The event loop is about to be destroyed (see C<ev_cleanup>).
		1276
		1277	=item C<EV_ASYNC>
		1278
		1279	The given async watcher has been asynchronously notified (see C<ev_async>).
		1280
		1281	=item C<EV_CUSTOM>
		1282
		1283	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1284	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1285
356	=item EV_ERROR	1286	=item C<EV_ERROR>
357		1287
358	An unspecified error has occured, the watcher has been stopped. This might	1288	An unspecified error has occurred, the watcher has been stopped. This might
359	happen because the watcher could not be properly started because libev	1289	happen because the watcher could not be properly started because libev
360	ran out of memory, a file descriptor was found to be closed or any other	1290	ran out of memory, a file descriptor was found to be closed or any other
		1291	problem. Libev considers these application bugs.
		1292
361	problem. You best act on it by reporting the problem and somehow coping	1293	You best act on it by reporting the problem and somehow coping with the
362	with the watcher being stopped.	1294	watcher being stopped. Note that well-written programs should not receive
		1295	an error ever, so when your watcher receives it, this usually indicates a
		1296	bug in your program.
363		1297
364	Libev will usually signal a few "dummy" events together with an error,	1298	Libev will usually signal a few "dummy" events together with an error, for
365	for example it might indicate that a fd is readable or writable, and if	1299	example it might indicate that a fd is readable or writable, and if your
366	your callbacks is well-written it can just attempt the operation and cope	1300	callbacks is well-written it can just attempt the operation and cope with
367	with the error from read() or write(). This will not work in multithreaded	1301	the error from read() or write(). This will not work in multi-threaded
368	programs, though, so beware.	1302	programs, though, as the fd could already be closed and reused for another
		1303	thing, so beware.
369		1304
370	=back	1305	=back
371		1306
372	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1307	=head2 GENERIC WATCHER FUNCTIONS
373		1308
374	Each watcher has, by default, a member C<void *data> that you can change	1309	=over 4
375	and read at any time, libev will completely ignore it. This cna be used
376	to associate arbitrary data with your watcher. If you need more data and
377	don't want to allocate memory and store a pointer to it in that data
378	member, you can also "subclass" the watcher type and provide your own
379	data:
380		1310
381	struct my_io	1311	=item C<ev_init> (ev_TYPE *watcher, callback)
		1312
		1313	This macro initialises the generic portion of a watcher. The contents
		1314	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1315	the generic parts of the watcher are initialised, you I<need> to call
		1316	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1317	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1318	which rolls both calls into one.
		1319
		1320	You can reinitialise a watcher at any time as long as it has been stopped
		1321	(or never started) and there are no pending events outstanding.
		1322
		1323	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1324	int revents)>.
		1325
		1326	Example: Initialise an C<ev_io> watcher in two steps.
		1327
		1328	ev_io w;
		1329	ev_init (&w, my_cb);
		1330	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1331
		1332	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1333
		1334	This macro initialises the type-specific parts of a watcher. You need to
		1335	call C<ev_init> at least once before you call this macro, but you can
		1336	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1337	macro on a watcher that is active (it can be pending, however, which is a
		1338	difference to the C<ev_init> macro).
		1339
		1340	Although some watcher types do not have type-specific arguments
		1341	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1342
		1343	See C<ev_init>, above, for an example.
		1344
		1345	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1346
		1347	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1348	calls into a single call. This is the most convenient method to initialise
		1349	a watcher. The same limitations apply, of course.
		1350
		1351	Example: Initialise and set an C<ev_io> watcher in one step.
		1352
		1353	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1354
		1355	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1356
		1357	Starts (activates) the given watcher. Only active watchers will receive
		1358	events. If the watcher is already active nothing will happen.
		1359
		1360	Example: Start the C<ev_io> watcher that is being abused as example in this
		1361	whole section.
		1362
		1363	ev_io_start (EV_DEFAULT_UC, &w);
		1364
		1365	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1366
		1367	Stops the given watcher if active, and clears the pending status (whether
		1368	the watcher was active or not).
		1369
		1370	It is possible that stopped watchers are pending - for example,
		1371	non-repeating timers are being stopped when they become pending - but
		1372	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1373	pending. If you want to free or reuse the memory used by the watcher it is
		1374	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1375
		1376	=item bool ev_is_active (ev_TYPE *watcher)
		1377
		1378	Returns a true value iff the watcher is active (i.e. it has been started
		1379	and not yet been stopped). As long as a watcher is active you must not modify
		1380	it.
		1381
		1382	=item bool ev_is_pending (ev_TYPE *watcher)
		1383
		1384	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1385	events but its callback has not yet been invoked). As long as a watcher
		1386	is pending (but not active) you must not call an init function on it (but
		1387	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1388	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1389	it).
		1390
		1391	=item callback ev_cb (ev_TYPE *watcher)
		1392
		1393	Returns the callback currently set on the watcher.
		1394
		1395	=item ev_set_cb (ev_TYPE *watcher, callback)
		1396
		1397	Change the callback. You can change the callback at virtually any time
		1398	(modulo threads).
		1399
		1400	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1401
		1402	=item int ev_priority (ev_TYPE *watcher)
		1403
		1404	Set and query the priority of the watcher. The priority is a small
		1405	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1406	(default: C<-2>). Pending watchers with higher priority will be invoked
		1407	before watchers with lower priority, but priority will not keep watchers
		1408	from being executed (except for C<ev_idle> watchers).
		1409
		1410	If you need to suppress invocation when higher priority events are pending
		1411	you need to look at C<ev_idle> watchers, which provide this functionality.
		1412
		1413	You I<must not> change the priority of a watcher as long as it is active or
		1414	pending.
		1415
		1416	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1417	fine, as long as you do not mind that the priority value you query might
		1418	or might not have been clamped to the valid range.
		1419
		1420	The default priority used by watchers when no priority has been set is
		1421	always C<0>, which is supposed to not be too high and not be too low :).
		1422
		1423	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1424	priorities.
		1425
		1426	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1427
		1428	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1429	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1430	can deal with that fact, as both are simply passed through to the
		1431	callback.
		1432
		1433	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1434
		1435	If the watcher is pending, this function clears its pending status and
		1436	returns its C<revents> bitset (as if its callback was invoked). If the
		1437	watcher isn't pending it does nothing and returns C<0>.
		1438
		1439	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1440	callback to be invoked, which can be accomplished with this function.
		1441
		1442	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1443
		1444	Feeds the given event set into the event loop, as if the specified event
		1445	had happened for the specified watcher (which must be a pointer to an
		1446	initialised but not necessarily started event watcher). Obviously you must
		1447	not free the watcher as long as it has pending events.
		1448
		1449	Stopping the watcher, letting libev invoke it, or calling
		1450	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1451	not started in the first place.
		1452
		1453	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1454	functions that do not need a watcher.
		1455
		1456	=back
		1457
		1458	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
		1459	OWN COMPOSITE WATCHERS> idioms.
		1460
		1461	=head2 WATCHER STATES
		1462
		1463	There are various watcher states mentioned throughout this manual -
		1464	active, pending and so on. In this section these states and the rules to
		1465	transition between them will be described in more detail - and while these
		1466	rules might look complicated, they usually do "the right thing".
		1467
		1468	=over 4
		1469
		1470	=item initialised
		1471
		1472	Before a watcher can be registered with the event loop it has to be
		1473	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1474	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1475
		1476	In this state it is simply some block of memory that is suitable for
		1477	use in an event loop. It can be moved around, freed, reused etc. at
		1478	will - as long as you either keep the memory contents intact, or call
		1479	C<ev_TYPE_init> again.
		1480
		1481	=item started/running/active
		1482
		1483	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1484	property of the event loop, and is actively waiting for events. While in
		1485	this state it cannot be accessed (except in a few documented ways), moved,
		1486	freed or anything else - the only legal thing is to keep a pointer to it,
		1487	and call libev functions on it that are documented to work on active watchers.
		1488
		1489	=item pending
		1490
		1491	If a watcher is active and libev determines that an event it is interested
		1492	in has occurred (such as a timer expiring), it will become pending. It will
		1493	stay in this pending state until either it is stopped or its callback is
		1494	about to be invoked, so it is not normally pending inside the watcher
		1495	callback.
		1496
		1497	The watcher might or might not be active while it is pending (for example,
		1498	an expired non-repeating timer can be pending but no longer active). If it
		1499	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1500	but it is still property of the event loop at this time, so cannot be
		1501	moved, freed or reused. And if it is active the rules described in the
		1502	previous item still apply.
		1503
		1504	It is also possible to feed an event on a watcher that is not active (e.g.
		1505	via C<ev_feed_event>), in which case it becomes pending without being
		1506	active.
		1507
		1508	=item stopped
		1509
		1510	A watcher can be stopped implicitly by libev (in which case it might still
		1511	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1512	latter will clear any pending state the watcher might be in, regardless
		1513	of whether it was active or not, so stopping a watcher explicitly before
		1514	freeing it is often a good idea.
		1515
		1516	While stopped (and not pending) the watcher is essentially in the
		1517	initialised state, that is, it can be reused, moved, modified in any way
		1518	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1519	it again).
		1520
		1521	=back
		1522
		1523	=head2 WATCHER PRIORITY MODELS
		1524
		1525	Many event loops support I<watcher priorities>, which are usually small
		1526	integers that influence the ordering of event callback invocation
		1527	between watchers in some way, all else being equal.
		1528
		1529	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1530	description for the more technical details such as the actual priority
		1531	range.
		1532
		1533	There are two common ways how these these priorities are being interpreted
		1534	by event loops:
		1535
		1536	In the more common lock-out model, higher priorities "lock out" invocation
		1537	of lower priority watchers, which means as long as higher priority
		1538	watchers receive events, lower priority watchers are not being invoked.
		1539
		1540	The less common only-for-ordering model uses priorities solely to order
		1541	callback invocation within a single event loop iteration: Higher priority
		1542	watchers are invoked before lower priority ones, but they all get invoked
		1543	before polling for new events.
		1544
		1545	Libev uses the second (only-for-ordering) model for all its watchers
		1546	except for idle watchers (which use the lock-out model).
		1547
		1548	The rationale behind this is that implementing the lock-out model for
		1549	watchers is not well supported by most kernel interfaces, and most event
		1550	libraries will just poll for the same events again and again as long as
		1551	their callbacks have not been executed, which is very inefficient in the
		1552	common case of one high-priority watcher locking out a mass of lower
		1553	priority ones.
		1554
		1555	Static (ordering) priorities are most useful when you have two or more
		1556	watchers handling the same resource: a typical usage example is having an
		1557	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1558	timeouts. Under load, data might be received while the program handles
		1559	other jobs, but since timers normally get invoked first, the timeout
		1560	handler will be executed before checking for data. In that case, giving
		1561	the timer a lower priority than the I/O watcher ensures that I/O will be
		1562	handled first even under adverse conditions (which is usually, but not
		1563	always, what you want).
		1564
		1565	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1566	will only be executed when no same or higher priority watchers have
		1567	received events, they can be used to implement the "lock-out" model when
		1568	required.
		1569
		1570	For example, to emulate how many other event libraries handle priorities,
		1571	you can associate an C<ev_idle> watcher to each such watcher, and in
		1572	the normal watcher callback, you just start the idle watcher. The real
		1573	processing is done in the idle watcher callback. This causes libev to
		1574	continuously poll and process kernel event data for the watcher, but when
		1575	the lock-out case is known to be rare (which in turn is rare :), this is
		1576	workable.
		1577
		1578	Usually, however, the lock-out model implemented that way will perform
		1579	miserably under the type of load it was designed to handle. In that case,
		1580	it might be preferable to stop the real watcher before starting the
		1581	idle watcher, so the kernel will not have to process the event in case
		1582	the actual processing will be delayed for considerable time.
		1583
		1584	Here is an example of an I/O watcher that should run at a strictly lower
		1585	priority than the default, and which should only process data when no
		1586	other events are pending:
		1587
		1588	ev_idle idle; // actual processing watcher
		1589	ev_io io; // actual event watcher
		1590
		1591	static void
		1592	io_cb (EV_P_ ev_io *w, int revents)
382	{	1593	{
383	struct ev_io io;	1594	// stop the I/O watcher, we received the event, but
384	int otherfd;	1595	// are not yet ready to handle it.
385	void *somedata;	1596	ev_io_stop (EV_A_ w);
386	struct whatever *mostinteresting;	1597
		1598	// start the idle watcher to handle the actual event.
		1599	// it will not be executed as long as other watchers
		1600	// with the default priority are receiving events.
		1601	ev_idle_start (EV_A_ &idle);
387	}	1602	}
388		1603
389	And since your callback will be called with a pointer to the watcher, you	1604	static void
390	can cast it back to your own type:	1605	idle_cb (EV_P_ ev_idle *w, int revents)
391
392	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
393	{	1606	{
394	struct my_io *w = (struct my_io *)w_;	1607	// actual processing
395	...	1608	read (STDIN_FILENO, ...);
		1609
		1610	// have to start the I/O watcher again, as
		1611	// we have handled the event
		1612	ev_io_start (EV_P_ &io);
396	}	1613	}
397		1614
398	More interesting and less C-conformant ways of catsing your callback type	1615	// initialisation
399	have been omitted....	1616	ev_idle_init (&idle, idle_cb);
		1617	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1618	ev_io_start (EV_DEFAULT_ &io);
		1619
		1620	In the "real" world, it might also be beneficial to start a timer, so that
		1621	low-priority connections can not be locked out forever under load. This
		1622	enables your program to keep a lower latency for important connections
		1623	during short periods of high load, while not completely locking out less
		1624	important ones.
400		1625
401		1626
402	=head1 WATCHER TYPES	1627	=head1 WATCHER TYPES
403		1628
404	This section describes each watcher in detail, but will not repeat	1629	This section describes each watcher in detail, but will not repeat
405	information given in the last section.	1630	information given in the last section. Any initialisation/set macros,
		1631	functions and members specific to the watcher type are explained.
406		1632
		1633	Members are additionally marked with either I<[read-only]>, meaning that,
		1634	while the watcher is active, you can look at the member and expect some
		1635	sensible content, but you must not modify it (you can modify it while the
		1636	watcher is stopped to your hearts content), or I<[read-write]>, which
		1637	means you can expect it to have some sensible content while the watcher
		1638	is active, but you can also modify it. Modifying it may not do something
		1639	sensible or take immediate effect (or do anything at all), but libev will
		1640	not crash or malfunction in any way.
		1641
		1642
407	=head2 struct ev_io - is my file descriptor readable or writable	1643	=head2 C<ev_io> - is this file descriptor readable or writable?
408		1644
409	I/O watchers check whether a file descriptor is readable or writable	1645	I/O watchers check whether a file descriptor is readable or writable
410	in each iteration of the event loop (This behaviour is called	1646	in each iteration of the event loop, or, more precisely, when reading
411	level-triggering because you keep receiving events as long as the	1647	would not block the process and writing would at least be able to write
412	condition persists. Remember you cna stop the watcher if you don't want to	1648	some data. This behaviour is called level-triggering because you keep
413	act on the event and neither want to receive future events).	1649	receiving events as long as the condition persists. Remember you can stop
		1650	the watcher if you don't want to act on the event and neither want to
		1651	receive future events.
414		1652
415	In general you can register as many read and/or write event watchers oer	1653	In general you can register as many read and/or write event watchers per
416	fd as you want (as long as you don't confuse yourself). Setting all file	1654	fd as you want (as long as you don't confuse yourself). Setting all file
417	descriptors to non-blocking mode is also usually a good idea (but not	1655	descriptors to non-blocking mode is also usually a good idea (but not
418	required if you know what you are doing).	1656	required if you know what you are doing).
419		1657
420	You have to be careful with dup'ed file descriptors, though. Some backends	1658	Another thing you have to watch out for is that it is quite easy to
421	(the linux epoll backend is a notable example) cannot handle dup'ed file	1659	receive "spurious" readiness notifications, that is, your callback might
422	descriptors correctly if you register interest in two or more fds pointing	1660	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
423	to the same file/socket etc. description.	1661	because there is no data. It is very easy to get into this situation even
		1662	with a relatively standard program structure. Thus it is best to always
		1663	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
		1664	preferable to a program hanging until some data arrives.
424		1665
425	If you must do this, then force the use of a known-to-be-good backend	1666	If you cannot run the fd in non-blocking mode (for example you should
426	(at the time of this writing, this includes only EVMETHOD_SELECT and	1667	not play around with an Xlib connection), then you have to separately
427	EVMETHOD_POLL).	1668	re-test whether a file descriptor is really ready with a known-to-be good
		1669	interface such as poll (fortunately in the case of Xlib, it already does
		1670	this on its own, so its quite safe to use). Some people additionally
		1671	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1672	indefinitely.
		1673
		1674	But really, best use non-blocking mode.
		1675
		1676	=head3 The special problem of disappearing file descriptors
		1677
		1678	Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
		1679	a file descriptor (either due to calling C<close> explicitly or any other
		1680	means, such as C<dup2>). The reason is that you register interest in some
		1681	file descriptor, but when it goes away, the operating system will silently
		1682	drop this interest. If another file descriptor with the same number then
		1683	is registered with libev, there is no efficient way to see that this is,
		1684	in fact, a different file descriptor.
		1685
		1686	To avoid having to explicitly tell libev about such cases, libev follows
		1687	the following policy: Each time C<ev_io_set> is being called, libev
		1688	will assume that this is potentially a new file descriptor, otherwise
		1689	it is assumed that the file descriptor stays the same. That means that
		1690	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1691	descriptor even if the file descriptor number itself did not change.
		1692
		1693	This is how one would do it normally anyway, the important point is that
		1694	the libev application should not optimise around libev but should leave
		1695	optimisations to libev.
		1696
		1697	=head3 The special problem of dup'ed file descriptors
		1698
		1699	Some backends (e.g. epoll), cannot register events for file descriptors,
		1700	but only events for the underlying file descriptions. That means when you
		1701	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1702	events for them, only one file descriptor might actually receive events.
		1703
		1704	There is no workaround possible except not registering events
		1705	for potentially C<dup ()>'ed file descriptors, or to resort to
		1706	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1707
		1708	=head3 The special problem of files
		1709
		1710	Many people try to use C<select> (or libev) on file descriptors
		1711	representing files, and expect it to become ready when their program
		1712	doesn't block on disk accesses (which can take a long time on their own).
		1713
		1714	However, this cannot ever work in the "expected" way - you get a readiness
		1715	notification as soon as the kernel knows whether and how much data is
		1716	there, and in the case of open files, that's always the case, so you
		1717	always get a readiness notification instantly, and your read (or possibly
		1718	write) will still block on the disk I/O.
		1719
		1720	Another way to view it is that in the case of sockets, pipes, character
		1721	devices and so on, there is another party (the sender) that delivers data
		1722	on its own, but in the case of files, there is no such thing: the disk
		1723	will not send data on its own, simply because it doesn't know what you
		1724	wish to read - you would first have to request some data.
		1725
		1726	Since files are typically not-so-well supported by advanced notification
		1727	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1728	to files, even though you should not use it. The reason for this is
		1729	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1730	usually a tty, often a pipe, but also sometimes files or special devices
		1731	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1732	F</dev/urandom>), and even though the file might better be served with
		1733	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1734	it "just works" instead of freezing.
		1735
		1736	So avoid file descriptors pointing to files when you know it (e.g. use
		1737	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1738	when you rarely read from a file instead of from a socket, and want to
		1739	reuse the same code path.
		1740
		1741	=head3 The special problem of fork
		1742
		1743	Some backends (epoll, kqueue, probably linuxaio) do not support C<fork ()>
		1744	at all or exhibit useless behaviour. Libev fully supports fork, but needs
		1745	to be told about it in the child if you want to continue to use it in the
		1746	child.
		1747
		1748	To support fork in your child processes, you have to call C<ev_loop_fork
		1749	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
		1750	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1751
		1752	=head3 The special problem of SIGPIPE
		1753
		1754	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1755	when writing to a pipe whose other end has been closed, your program gets
		1756	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1757	this is sensible behaviour, for daemons, this is usually undesirable.
		1758
		1759	So when you encounter spurious, unexplained daemon exits, make sure you
		1760	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1761	somewhere, as that would have given you a big clue).
		1762
		1763	=head3 The special problem of accept()ing when you can't
		1764
		1765	Many implementations of the POSIX C<accept> function (for example,
		1766	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1767	connection from the pending queue in all error cases.
		1768
		1769	For example, larger servers often run out of file descriptors (because
		1770	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1771	rejecting the connection, leading to libev signalling readiness on
		1772	the next iteration again (the connection still exists after all), and
		1773	typically causing the program to loop at 100% CPU usage.
		1774
		1775	Unfortunately, the set of errors that cause this issue differs between
		1776	operating systems, there is usually little the app can do to remedy the
		1777	situation, and no known thread-safe method of removing the connection to
		1778	cope with overload is known (to me).
		1779
		1780	One of the easiest ways to handle this situation is to just ignore it
		1781	- when the program encounters an overload, it will just loop until the
		1782	situation is over. While this is a form of busy waiting, no OS offers an
		1783	event-based way to handle this situation, so it's the best one can do.
		1784
		1785	A better way to handle the situation is to log any errors other than
		1786	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1787	messages, and continue as usual, which at least gives the user an idea of
		1788	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1789	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1790	usage.
		1791
		1792	If your program is single-threaded, then you could also keep a dummy file
		1793	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1794	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1795	close that fd, and create a new dummy fd. This will gracefully refuse
		1796	clients under typical overload conditions.
		1797
		1798	The last way to handle it is to simply log the error and C<exit>, as
		1799	is often done with C<malloc> failures, but this results in an easy
		1800	opportunity for a DoS attack.
		1801
		1802	=head3 Watcher-Specific Functions
428		1803
429	=over 4	1804	=over 4
430		1805
431	=item ev_io_init (ev_io *, callback, int fd, int events)	1806	=item ev_io_init (ev_io *, callback, int fd, int events)
432		1807
433	=item ev_io_set (ev_io *, int fd, int events)	1808	=item ev_io_set (ev_io *, int fd, int events)
434		1809
435	Configures an ev_io watcher. The fd is the file descriptor to rceeive	1810	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
436	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1811	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
437	EV_WRITE> to receive the given events.	1812	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1813
		1814	=item int fd [read-only]
		1815
		1816	The file descriptor being watched.
		1817
		1818	=item int events [read-only]
		1819
		1820	The events being watched.
438		1821
439	=back	1822	=back
440		1823
		1824	=head3 Examples
		1825
		1826	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1827	readable, but only once. Since it is likely line-buffered, you could
		1828	attempt to read a whole line in the callback.
		1829
		1830	static void
		1831	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1832	{
		1833	ev_io_stop (loop, w);
		1834	.. read from stdin here (or from w->fd) and handle any I/O errors
		1835	}
		1836
		1837	...
		1838	struct ev_loop *loop = ev_default_init (0);
		1839	ev_io stdin_readable;
		1840	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1841	ev_io_start (loop, &stdin_readable);
		1842	ev_run (loop, 0);
		1843
		1844
441	=head2 struct ev_timer - relative and optionally recurring timeouts	1845	=head2 C<ev_timer> - relative and optionally repeating timeouts
442		1846
443	Timer watchers are simple relative timers that generate an event after a	1847	Timer watchers are simple relative timers that generate an event after a
444	given time, and optionally repeating in regular intervals after that.	1848	given time, and optionally repeating in regular intervals after that.
445		1849
446	The timers are based on real time, that is, if you register an event that	1850	The timers are based on real time, that is, if you register an event that
447	times out after an hour and youreset your system clock to last years	1851	times out after an hour and you reset your system clock to January last
448	time, it will still time out after (roughly) and hour. "Roughly" because	1852	year, it will still time out after (roughly) one hour. "Roughly" because
449	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1853	detecting time jumps is hard, and some inaccuracies are unavoidable (the
450	monotonic clock option helps a lot here).	1854	monotonic clock option helps a lot here).
451		1855
		1856	The callback is guaranteed to be invoked only I<after> its timeout has
		1857	passed (not I<at>, so on systems with very low-resolution clocks this
		1858	might introduce a small delay, see "the special problem of being too
		1859	early", below). If multiple timers become ready during the same loop
		1860	iteration then the ones with earlier time-out values are invoked before
		1861	ones of the same priority with later time-out values (but this is no
		1862	longer true when a callback calls C<ev_run> recursively).
		1863
		1864	=head3 Be smart about timeouts
		1865
		1866	Many real-world problems involve some kind of timeout, usually for error
		1867	recovery. A typical example is an HTTP request - if the other side hangs,
		1868	you want to raise some error after a while.
		1869
		1870	What follows are some ways to handle this problem, from obvious and
		1871	inefficient to smart and efficient.
		1872
		1873	In the following, a 60 second activity timeout is assumed - a timeout that
		1874	gets reset to 60 seconds each time there is activity (e.g. each time some
		1875	data or other life sign was received).
		1876
452	=over 4	1877	=over 4
453		1878
		1879	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1880
		1881	This is the most obvious, but not the most simple way: In the beginning,
		1882	start the watcher:
		1883
		1884	ev_timer_init (timer, callback, 60., 0.);
		1885	ev_timer_start (loop, timer);
		1886
		1887	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1888	and start it again:
		1889
		1890	ev_timer_stop (loop, timer);
		1891	ev_timer_set (timer, 60., 0.);
		1892	ev_timer_start (loop, timer);
		1893
		1894	This is relatively simple to implement, but means that each time there is
		1895	some activity, libev will first have to remove the timer from its internal
		1896	data structure and then add it again. Libev tries to be fast, but it's
		1897	still not a constant-time operation.
		1898
		1899	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1900
		1901	This is the easiest way, and involves using C<ev_timer_again> instead of
		1902	C<ev_timer_start>.
		1903
		1904	To implement this, configure an C<ev_timer> with a C<repeat> value
		1905	of C<60> and then call C<ev_timer_again> at start and each time you
		1906	successfully read or write some data. If you go into an idle state where
		1907	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1908	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1909
		1910	That means you can ignore both the C<ev_timer_start> function and the
		1911	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1912	member and C<ev_timer_again>.
		1913
		1914	At start:
		1915
		1916	ev_init (timer, callback);
		1917	timer->repeat = 60.;
		1918	ev_timer_again (loop, timer);
		1919
		1920	Each time there is some activity:
		1921
		1922	ev_timer_again (loop, timer);
		1923
		1924	It is even possible to change the time-out on the fly, regardless of
		1925	whether the watcher is active or not:
		1926
		1927	timer->repeat = 30.;
		1928	ev_timer_again (loop, timer);
		1929
		1930	This is slightly more efficient then stopping/starting the timer each time
		1931	you want to modify its timeout value, as libev does not have to completely
		1932	remove and re-insert the timer from/into its internal data structure.
		1933
		1934	It is, however, even simpler than the "obvious" way to do it.
		1935
		1936	=item 3. Let the timer time out, but then re-arm it as required.
		1937
		1938	This method is more tricky, but usually most efficient: Most timeouts are
		1939	relatively long compared to the intervals between other activity - in
		1940	our example, within 60 seconds, there are usually many I/O events with
		1941	associated activity resets.
		1942
		1943	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1944	but remember the time of last activity, and check for a real timeout only
		1945	within the callback:
		1946
		1947	ev_tstamp timeout = 60.;
		1948	ev_tstamp last_activity; // time of last activity
		1949	ev_timer timer;
		1950
		1951	static void
		1952	callback (EV_P_ ev_timer *w, int revents)
		1953	{
		1954	// calculate when the timeout would happen
		1955	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
		1956
		1957	// if negative, it means we the timeout already occurred
		1958	if (after < 0.)
		1959	{
		1960	// timeout occurred, take action
		1961	}
		1962	else
		1963	{
		1964	// callback was invoked, but there was some recent
		1965	// activity. simply restart the timer to time out
		1966	// after "after" seconds, which is the earliest time
		1967	// the timeout can occur.
		1968	ev_timer_set (w, after, 0.);
		1969	ev_timer_start (EV_A_ w);
		1970	}
		1971	}
		1972
		1973	To summarise the callback: first calculate in how many seconds the
		1974	timeout will occur (by calculating the absolute time when it would occur,
		1975	C<last_activity + timeout>, and subtracting the current time, C<ev_now
		1976	(EV_A)> from that).
		1977
		1978	If this value is negative, then we are already past the timeout, i.e. we
		1979	timed out, and need to do whatever is needed in this case.
		1980
		1981	Otherwise, we now the earliest time at which the timeout would trigger,
		1982	and simply start the timer with this timeout value.
		1983
		1984	In other words, each time the callback is invoked it will check whether
		1985	the timeout occurred. If not, it will simply reschedule itself to check
		1986	again at the earliest time it could time out. Rinse. Repeat.
		1987
		1988	This scheme causes more callback invocations (about one every 60 seconds
		1989	minus half the average time between activity), but virtually no calls to
		1990	libev to change the timeout.
		1991
		1992	To start the machinery, simply initialise the watcher and set
		1993	C<last_activity> to the current time (meaning there was some activity just
		1994	now), then call the callback, which will "do the right thing" and start
		1995	the timer:
		1996
		1997	last_activity = ev_now (EV_A);
		1998	ev_init (&timer, callback);
		1999	callback (EV_A_ &timer, 0);
		2000
		2001	When there is some activity, simply store the current time in
		2002	C<last_activity>, no libev calls at all:
		2003
		2004	if (activity detected)
		2005	last_activity = ev_now (EV_A);
		2006
		2007	When your timeout value changes, then the timeout can be changed by simply
		2008	providing a new value, stopping the timer and calling the callback, which
		2009	will again do the right thing (for example, time out immediately :).
		2010
		2011	timeout = new_value;
		2012	ev_timer_stop (EV_A_ &timer);
		2013	callback (EV_A_ &timer, 0);
		2014
		2015	This technique is slightly more complex, but in most cases where the
		2016	time-out is unlikely to be triggered, much more efficient.
		2017
		2018	=item 4. Wee, just use a double-linked list for your timeouts.
		2019
		2020	If there is not one request, but many thousands (millions...), all
		2021	employing some kind of timeout with the same timeout value, then one can
		2022	do even better:
		2023
		2024	When starting the timeout, calculate the timeout value and put the timeout
		2025	at the I<end> of the list.
		2026
		2027	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		2028	the list is expected to fire (for example, using the technique #3).
		2029
		2030	When there is some activity, remove the timer from the list, recalculate
		2031	the timeout, append it to the end of the list again, and make sure to
		2032	update the C<ev_timer> if it was taken from the beginning of the list.
		2033
		2034	This way, one can manage an unlimited number of timeouts in O(1) time for
		2035	starting, stopping and updating the timers, at the expense of a major
		2036	complication, and having to use a constant timeout. The constant timeout
		2037	ensures that the list stays sorted.
		2038
		2039	=back
		2040
		2041	So which method the best?
		2042
		2043	Method #2 is a simple no-brain-required solution that is adequate in most
		2044	situations. Method #3 requires a bit more thinking, but handles many cases
		2045	better, and isn't very complicated either. In most case, choosing either
		2046	one is fine, with #3 being better in typical situations.
		2047
		2048	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		2049	rather complicated, but extremely efficient, something that really pays
		2050	off after the first million or so of active timers, i.e. it's usually
		2051	overkill :)
		2052
		2053	=head3 The special problem of being too early
		2054
		2055	If you ask a timer to call your callback after three seconds, then
		2056	you expect it to be invoked after three seconds - but of course, this
		2057	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		2058	guaranteed to any precision by libev - imagine somebody suspending the
		2059	process with a STOP signal for a few hours for example.
		2060
		2061	So, libev tries to invoke your callback as soon as possible I<after> the
		2062	delay has occurred, but cannot guarantee this.
		2063
		2064	A less obvious failure mode is calling your callback too early: many event
		2065	loops compare timestamps with a "elapsed delay >= requested delay", but
		2066	this can cause your callback to be invoked much earlier than you would
		2067	expect.
		2068
		2069	To see why, imagine a system with a clock that only offers full second
		2070	resolution (think windows if you can't come up with a broken enough OS
		2071	yourself). If you schedule a one-second timer at the time 500.9, then the
		2072	event loop will schedule your timeout to elapse at a system time of 500
		2073	(500.9 truncated to the resolution) + 1, or 501.
		2074
		2075	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2076	501" and invoke the callback 0.1s after it was started, even though a
		2077	one-second delay was requested - this is being "too early", despite best
		2078	intentions.
		2079
		2080	This is the reason why libev will never invoke the callback if the elapsed
		2081	delay equals the requested delay, but only when the elapsed delay is
		2082	larger than the requested delay. In the example above, libev would only invoke
		2083	the callback at system time 502, or 1.1s after the timer was started.
		2084
		2085	So, while libev cannot guarantee that your callback will be invoked
		2086	exactly when requested, it I<can> and I<does> guarantee that the requested
		2087	delay has actually elapsed, or in other words, it always errs on the "too
		2088	late" side of things.
		2089
		2090	=head3 The special problem of time updates
		2091
		2092	Establishing the current time is a costly operation (it usually takes
		2093	at least one system call): EV therefore updates its idea of the current
		2094	time only before and after C<ev_run> collects new events, which causes a
		2095	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		2096	lots of events in one iteration.
		2097
		2098	The relative timeouts are calculated relative to the C<ev_now ()>
		2099	time. This is usually the right thing as this timestamp refers to the time
		2100	of the event triggering whatever timeout you are modifying/starting. If
		2101	you suspect event processing to be delayed and you I<need> to base the
		2102	timeout on the current time, use something like the following to adjust
		2103	for it:
		2104
		2105	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
		2106
		2107	If the event loop is suspended for a long time, you can also force an
		2108	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		2109	()>, although that will push the event time of all outstanding events
		2110	further into the future.
		2111
		2112	=head3 The special problem of unsynchronised clocks
		2113
		2114	Modern systems have a variety of clocks - libev itself uses the normal
		2115	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2116	jumps).
		2117
		2118	Neither of these clocks is synchronised with each other or any other clock
		2119	on the system, so C<ev_time ()> might return a considerably different time
		2120	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2121	a call to C<gettimeofday> might return a second count that is one higher
		2122	than a directly following call to C<time>.
		2123
		2124	The moral of this is to only compare libev-related timestamps with
		2125	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2126	a second or so.
		2127
		2128	One more problem arises due to this lack of synchronisation: if libev uses
		2129	the system monotonic clock and you compare timestamps from C<ev_time>
		2130	or C<ev_now> from when you started your timer and when your callback is
		2131	invoked, you will find that sometimes the callback is a bit "early".
		2132
		2133	This is because C<ev_timer>s work in real time, not wall clock time, so
		2134	libev makes sure your callback is not invoked before the delay happened,
		2135	I<measured according to the real time>, not the system clock.
		2136
		2137	If your timeouts are based on a physical timescale (e.g. "time out this
		2138	connection after 100 seconds") then this shouldn't bother you as it is
		2139	exactly the right behaviour.
		2140
		2141	If you want to compare wall clock/system timestamps to your timers, then
		2142	you need to use C<ev_periodic>s, as these are based on the wall clock
		2143	time, where your comparisons will always generate correct results.
		2144
		2145	=head3 The special problems of suspended animation
		2146
		2147	When you leave the server world it is quite customary to hit machines that
		2148	can suspend/hibernate - what happens to the clocks during such a suspend?
		2149
		2150	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		2151	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2152	to run until the system is suspended, but they will not advance while the
		2153	system is suspended. That means, on resume, it will be as if the program
		2154	was frozen for a few seconds, but the suspend time will not be counted
		2155	towards C<ev_timer> when a monotonic clock source is used. The real time
		2156	clock advanced as expected, but if it is used as sole clocksource, then a
		2157	long suspend would be detected as a time jump by libev, and timers would
		2158	be adjusted accordingly.
		2159
		2160	I would not be surprised to see different behaviour in different between
		2161	operating systems, OS versions or even different hardware.
		2162
		2163	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2164	time jump in the monotonic clocks and the realtime clock. If the program
		2165	is suspended for a very long time, and monotonic clock sources are in use,
		2166	then you can expect C<ev_timer>s to expire as the full suspension time
		2167	will be counted towards the timers. When no monotonic clock source is in
		2168	use, then libev will again assume a timejump and adjust accordingly.
		2169
		2170	It might be beneficial for this latter case to call C<ev_suspend>
		2171	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2172	deterministic behaviour in this case (you can do nothing against
		2173	C<SIGSTOP>).
		2174
		2175	=head3 Watcher-Specific Functions and Data Members
		2176
		2177	=over 4
		2178
454	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2179	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
455		2180
456	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2181	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
457		2182
458	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2183	Configure the timer to trigger after C<after> seconds (fractional and
459	C<0.>, then it will automatically be stopped. If it is positive, then the	2184	negative values are supported). If C<repeat> is C<0.>, then it will
		2185	automatically be stopped once the timeout is reached. If it is positive,
460	timer will automatically be configured to trigger again C<repeat> seconds	2186	then the timer will automatically be configured to trigger again C<repeat>
461	later, again, and again, until stopped manually.	2187	seconds later, again, and again, until stopped manually.
462		2188
463	The timer itself will do a best-effort at avoiding drift, that is, if you	2189	The timer itself will do a best-effort at avoiding drift, that is, if
464	configure a timer to trigger every 10 seconds, then it will trigger at	2190	you configure a timer to trigger every 10 seconds, then it will normally
465	exactly 10 second intervals. If, however, your program cannot keep up with	2191	trigger at exactly 10 second intervals. If, however, your program cannot
466	the timer (ecause it takes longer than those 10 seconds to do stuff) the	2192	keep up with the timer (because it takes longer than those 10 seconds to
467	timer will not fire more than once per event loop iteration.	2193	do stuff) the timer will not fire more than once per event loop iteration.
468		2194
469	=item ev_timer_again (loop)	2195	=item ev_timer_again (loop, ev_timer *)
470		2196
471	This will act as if the timer timed out and restart it again if it is	2197	This will act as if the timer timed out, and restarts it again if it is
472	repeating. The exact semantics are:	2198	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2199	timeout to the C<repeat> value and calling C<ev_timer_start>.
473		2200
		2201	The exact semantics are as in the following rules, all of which will be
		2202	applied to the watcher:
		2203
		2204	=over 4
		2205
		2206	=item If the timer is pending, the pending status is always cleared.
		2207
474	If the timer is started but nonrepeating, stop it.	2208	=item If the timer is started but non-repeating, stop it (as if it timed
		2209	out, without invoking it).
475		2210
476	If the timer is repeating, either start it if necessary (with the repeat	2211	=item If the timer is repeating, make the C<repeat> value the new timeout
477	value), or reset the running timer to the repeat value.	2212	and start the timer, if necessary.
478
479	This sounds a bit complicated, but here is a useful and typical
480	example: Imagine you have a tcp connection and you want a so-called idle
481	timeout, that is, you want to be called when there have been, say, 60
482	seconds of inactivity on the socket. The easiest way to do this is to
483	configure an ev_timer with after=repeat=60 and calling ev_timer_again each
484	time you successfully read or write some data. If you go into an idle
485	state where you do not expect data to travel on the socket, you can stop
486	the timer, and again will automatically restart it if need be.
487		2213
488	=back	2214	=back
489		2215
		2216	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
		2217	usage example.
		2218
		2219	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
		2220
		2221	Returns the remaining time until a timer fires. If the timer is active,
		2222	then this time is relative to the current event loop time, otherwise it's
		2223	the timeout value currently configured.
		2224
		2225	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2226	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2227	will return C<4>. When the timer expires and is restarted, it will return
		2228	roughly C<7> (likely slightly less as callback invocation takes some time,
		2229	too), and so on.
		2230
		2231	=item ev_tstamp repeat [read-write]
		2232
		2233	The current C<repeat> value. Will be used each time the watcher times out
		2234	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2235	which is also when any modifications are taken into account.
		2236
		2237	=back
		2238
		2239	=head3 Examples
		2240
		2241	Example: Create a timer that fires after 60 seconds.
		2242
		2243	static void
		2244	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2245	{
		2246	.. one minute over, w is actually stopped right here
		2247	}
		2248
		2249	ev_timer mytimer;
		2250	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2251	ev_timer_start (loop, &mytimer);
		2252
		2253	Example: Create a timeout timer that times out after 10 seconds of
		2254	inactivity.
		2255
		2256	static void
		2257	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2258	{
		2259	.. ten seconds without any activity
		2260	}
		2261
		2262	ev_timer mytimer;
		2263	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2264	ev_timer_again (&mytimer); /* start timer */
		2265	ev_run (loop, 0);
		2266
		2267	// and in some piece of code that gets executed on any "activity":
		2268	// reset the timeout to start ticking again at 10 seconds
		2269	ev_timer_again (&mytimer);
		2270
		2271
490	=head2 ev_periodic - to cron or not to cron it	2272	=head2 C<ev_periodic> - to cron or not to cron?
491		2273
492	Periodic watchers are also timers of a kind, but they are very versatile	2274	Periodic watchers are also timers of a kind, but they are very versatile
493	(and unfortunately a bit complex).	2275	(and unfortunately a bit complex).
494		2276
495	Unlike ev_timer's, they are not based on real time (or relative time)	2277	Unlike C<ev_timer>, periodic watchers are not based on real time (or
496	but on wallclock time (absolute time). You can tell a periodic watcher	2278	relative time, the physical time that passes) but on wall clock time
497	to trigger "at" some specific point in time. For example, if you tell a	2279	(absolute time, the thing you can read on your calendar or clock). The
498	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2280	difference is that wall clock time can run faster or slower than real
499	+ 10.>) and then reset your system clock to the last year, then it will	2281	time, and time jumps are not uncommon (e.g. when you adjust your
500	take a year to trigger the event (unlike an ev_timer, which would trigger	2282	wrist-watch).
501	roughly 10 seconds later and of course not if you reset your system time
502	again).
503		2283
504	They can also be used to implement vastly more complex timers, such as	2284	You can tell a periodic watcher to trigger after some specific point
		2285	in time: for example, if you tell a periodic watcher to trigger "in 10
		2286	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2287	not a delay) and then reset your system clock to January of the previous
		2288	year, then it will take a year or more to trigger the event (unlike an
		2289	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2290	it, as it uses a relative timeout).
		2291
		2292	C<ev_periodic> watchers can also be used to implement vastly more complex
505	triggering an event on eahc midnight, local time.	2293	timers, such as triggering an event on each "midnight, local time", or
		2294	other complicated rules. This cannot easily be done with C<ev_timer>
		2295	watchers, as those cannot react to time jumps.
		2296
		2297	As with timers, the callback is guaranteed to be invoked only when the
		2298	point in time where it is supposed to trigger has passed. If multiple
		2299	timers become ready during the same loop iteration then the ones with
		2300	earlier time-out values are invoked before ones with later time-out values
		2301	(but this is no longer true when a callback calls C<ev_run> recursively).
		2302
		2303	=head3 Watcher-Specific Functions and Data Members
506		2304
507	=over 4	2305	=over 4
508		2306
509	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2307	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
510		2308
511	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2309	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
512		2310
513	Lots of arguments, lets sort it out... There are basically three modes of	2311	Lots of arguments, let's sort it out... There are basically three modes of
514	operation, and we will explain them from simplest to complex:	2312	operation, and we will explain them from simplest to most complex:
515
516		2313
517	=over 4	2314	=over 4
518		2315
519	=item * absolute timer (interval = reschedule_cb = 0)	2316	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
520		2317
521	In this configuration the watcher triggers an event at the wallclock time	2318	In this configuration the watcher triggers an event after the wall clock
522	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2319	time C<offset> has passed. It will not repeat and will not adjust when a
523	that is, if it is to be run at January 1st 2011 then it will run when the	2320	time jump occurs, that is, if it is to be run at January 1st 2011 then it
524	system time reaches or surpasses this time.	2321	will be stopped and invoked when the system clock reaches or surpasses
		2322	this point in time.
525		2323
526	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2324	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
527		2325
528	In this mode the watcher will always be scheduled to time out at the next	2326	In this mode the watcher will always be scheduled to time out at the next
529	C<at + N * interval> time (for some integer N) and then repeat, regardless	2327	C<offset + N * interval> time (for some integer N, which can also be
530	of any time jumps.	2328	negative) and then repeat, regardless of any time jumps. The C<offset>
		2329	argument is merely an offset into the C<interval> periods.
531		2330
532	This can be used to create timers that do not drift with respect to system	2331	This can be used to create timers that do not drift with respect to the
533	time:	2332	system clock, for example, here is an C<ev_periodic> that triggers each
		2333	hour, on the hour (with respect to UTC):
534		2334
535	ev_periodic_set (&periodic, 0., 3600., 0);	2335	ev_periodic_set (&periodic, 0., 3600., 0);
536		2336
537	This doesn't mean there will always be 3600 seconds in between triggers,	2337	This doesn't mean there will always be 3600 seconds in between triggers,
538	but only that the the callback will be called when the system time shows a	2338	but only that the callback will be called when the system time shows a
539	full hour (UTC), or more correct, when the system time is evenly divisible	2339	full hour (UTC), or more correctly, when the system time is evenly divisible
540	by 3600.	2340	by 3600.
541		2341
542	Another way to think about it (for the mathematically inclined) is that	2342	Another way to think about it (for the mathematically inclined) is that
543	ev_periodic will try to run the callback in this mode at the next possible	2343	C<ev_periodic> will try to run the callback in this mode at the next possible
544	time where C<time = at (mod interval)>, regardless of any time jumps.	2344	time where C<time = offset (mod interval)>, regardless of any time jumps.
545		2345
		2346	The C<interval> I<MUST> be positive, and for numerical stability, the
		2347	interval value should be higher than C<1/8192> (which is around 100
		2348	microseconds) and C<offset> should be higher than C<0> and should have
		2349	at most a similar magnitude as the current time (say, within a factor of
		2350	ten). Typical values for offset are, in fact, C<0> or something between
		2351	C<0> and C<interval>, which is also the recommended range.
		2352
		2353	Note also that there is an upper limit to how often a timer can fire (CPU
		2354	speed for example), so if C<interval> is very small then timing stability
		2355	will of course deteriorate. Libev itself tries to be exact to be about one
		2356	millisecond (if the OS supports it and the machine is fast enough).
		2357
546	=item * manual reschedule mode (reschedule_cb = callback)	2358	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
547		2359
548	In this mode the values for C<interval> and C<at> are both being	2360	In this mode the values for C<interval> and C<offset> are both being
549	ignored. Instead, each time the periodic watcher gets scheduled, the	2361	ignored. Instead, each time the periodic watcher gets scheduled, the
550	reschedule callback will be called with the watcher as first, and the	2362	reschedule callback will be called with the watcher as first, and the
551	current time as second argument.	2363	current time as second argument.
552		2364
553	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	2365	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
554	periodic watcher, ever, or make any event loop modificstions>. If you need	2366	or make ANY other event loop modifications whatsoever, unless explicitly
555	to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards.	2367	allowed by documentation here>.
556		2368
		2369	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2370	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2371	only event loop modification you are allowed to do).
		2372
557	Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2373	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
558	ev_tstamp now)>, e.g.:	2374	*w, ev_tstamp now)>, e.g.:
559		2375
		2376	static ev_tstamp
560	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2377	my_rescheduler (ev_periodic *w, ev_tstamp now)
561	{	2378	{
562	return now + 60.;	2379	return now + 60.;
563	}	2380	}
564		2381
565	It must return the next time to trigger, based on the passed time value	2382	It must return the next time to trigger, based on the passed time value
566	(that is, the lowest time value larger than to the second argument). It	2383	(that is, the lowest time value larger than to the second argument). It
567	will usually be called just before the callback will be triggered, but	2384	will usually be called just before the callback will be triggered, but
568	might be called at other times, too.	2385	might be called at other times, too.
569		2386
		2387	NOTE: I<< This callback must always return a time that is higher than or
		2388	equal to the passed C<now> value >>.
		2389
570	This can be used to create very complex timers, such as a timer that	2390	This can be used to create very complex timers, such as a timer that
571	triggers on each midnight, local time. To do this, you would calculate the	2391	triggers on "next midnight, local time". To do this, you would calculate
572	next midnight after C<now> and return the timestamp value for this. How you do this	2392	the next midnight after C<now> and return the timestamp value for
573	is, again, up to you (but it is not trivial).	2393	this. Here is a (completely untested, no error checking) example on how to
		2394	do this:
		2395
		2396	#include <time.h>
		2397
		2398	static ev_tstamp
		2399	my_rescheduler (ev_periodic *w, ev_tstamp now)
		2400	{
		2401	time_t tnow = (time_t)now;
		2402	struct tm tm;
		2403	localtime_r (&tnow, &tm);
		2404
		2405	tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
		2406	++tm.tm_mday; // midnight next day
		2407
		2408	return mktime (&tm);
		2409	}
		2410
		2411	Note: this code might run into trouble on days that have more then two
		2412	midnights (beginning and end).
574		2413
575	=back	2414	=back
576		2415
577	=item ev_periodic_again (loop, ev_periodic *)	2416	=item ev_periodic_again (loop, ev_periodic *)
578		2417
579	Simply stops and restarts the periodic watcher again. This is only useful	2418	Simply stops and restarts the periodic watcher again. This is only useful
580	when you changed some parameters or the reschedule callback would return	2419	when you changed some parameters or the reschedule callback would return
581	a different time than the last time it was called (e.g. in a crond like	2420	a different time than the last time it was called (e.g. in a crond like
582	program when the crontabs have changed).	2421	program when the crontabs have changed).
583		2422
		2423	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2424
		2425	When active, returns the absolute time that the watcher is supposed
		2426	to trigger next. This is not the same as the C<offset> argument to
		2427	C<ev_periodic_set>, but indeed works even in interval and manual
		2428	rescheduling modes.
		2429
		2430	=item ev_tstamp offset [read-write]
		2431
		2432	When repeating, this contains the offset value, otherwise this is the
		2433	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2434	although libev might modify this value for better numerical stability).
		2435
		2436	Can be modified any time, but changes only take effect when the periodic
		2437	timer fires or C<ev_periodic_again> is being called.
		2438
		2439	=item ev_tstamp interval [read-write]
		2440
		2441	The current interval value. Can be modified any time, but changes only
		2442	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2443	called.
		2444
		2445	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2446
		2447	The current reschedule callback, or C<0>, if this functionality is
		2448	switched off. Can be changed any time, but changes only take effect when
		2449	the periodic timer fires or C<ev_periodic_again> is being called.
		2450
584	=back	2451	=back
585		2452
		2453	=head3 Examples
		2454
		2455	Example: Call a callback every hour, or, more precisely, whenever the
		2456	system time is divisible by 3600. The callback invocation times have
		2457	potentially a lot of jitter, but good long-term stability.
		2458
		2459	static void
		2460	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2461	{
		2462	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2463	}
		2464
		2465	ev_periodic hourly_tick;
		2466	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2467	ev_periodic_start (loop, &hourly_tick);
		2468
		2469	Example: The same as above, but use a reschedule callback to do it:
		2470
		2471	#include <math.h>
		2472
		2473	static ev_tstamp
		2474	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2475	{
		2476	return now + (3600. - fmod (now, 3600.));
		2477	}
		2478
		2479	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2480
		2481	Example: Call a callback every hour, starting now:
		2482
		2483	ev_periodic hourly_tick;
		2484	ev_periodic_init (&hourly_tick, clock_cb,
		2485	fmod (ev_now (loop), 3600.), 3600., 0);
		2486	ev_periodic_start (loop, &hourly_tick);
		2487
		2488
586	=head2 ev_signal - signal me when a signal gets signalled	2489	=head2 C<ev_signal> - signal me when a signal gets signalled!
587		2490
588	Signal watchers will trigger an event when the process receives a specific	2491	Signal watchers will trigger an event when the process receives a specific
589	signal one or more times. Even though signals are very asynchronous, libev	2492	signal one or more times. Even though signals are very asynchronous, libev
590	will try its best to deliver signals synchronously, i.e. as part of the	2493	will try its best to deliver signals synchronously, i.e. as part of the
591	normal event processing, like any other event.	2494	normal event processing, like any other event.
592		2495
		2496	If you want signals to be delivered truly asynchronously, just use
		2497	C<sigaction> as you would do without libev and forget about sharing
		2498	the signal. You can even use C<ev_async> from a signal handler to
		2499	synchronously wake up an event loop.
		2500
593	You cna configure as many watchers as you like per signal. Only when the	2501	You can configure as many watchers as you like for the same signal, but
594	first watcher gets started will libev actually register a signal watcher	2502	only within the same loop, i.e. you can watch for C<SIGINT> in your
595	with the kernel (thus it coexists with your own signal handlers as long	2503	default loop and for C<SIGIO> in another loop, but you cannot watch for
596	as you don't register any with libev). Similarly, when the last signal	2504	C<SIGINT> in both the default loop and another loop at the same time. At
597	watcher for a signal is stopped libev will reset the signal handler to	2505	the moment, C<SIGCHLD> is permanently tied to the default loop.
598	SIG_DFL (regardless of what it was set to before).	2506
		2507	Only after the first watcher for a signal is started will libev actually
		2508	register something with the kernel. It thus coexists with your own signal
		2509	handlers as long as you don't register any with libev for the same signal.
		2510
		2511	If possible and supported, libev will install its handlers with
		2512	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2513	not be unduly interrupted. If you have a problem with system calls getting
		2514	interrupted by signals you can block all signals in an C<ev_check> watcher
		2515	and unblock them in an C<ev_prepare> watcher.
		2516
		2517	=head3 The special problem of inheritance over fork/execve/pthread_create
		2518
		2519	Both the signal mask (C<sigprocmask>) and the signal disposition
		2520	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2521	stopping it again), that is, libev might or might not block the signal,
		2522	and might or might not set or restore the installed signal handler (but
		2523	see C<EVFLAG_NOSIGMASK>).
		2524
		2525	While this does not matter for the signal disposition (libev never
		2526	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2527	C<execve>), this matters for the signal mask: many programs do not expect
		2528	certain signals to be blocked.
		2529
		2530	This means that before calling C<exec> (from the child) you should reset
		2531	the signal mask to whatever "default" you expect (all clear is a good
		2532	choice usually).
		2533
		2534	The simplest way to ensure that the signal mask is reset in the child is
		2535	to install a fork handler with C<pthread_atfork> that resets it. That will
		2536	catch fork calls done by libraries (such as the libc) as well.
		2537
		2538	In current versions of libev, the signal will not be blocked indefinitely
		2539	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2540	the window of opportunity for problems, it will not go away, as libev
		2541	I<has> to modify the signal mask, at least temporarily.
		2542
		2543	So I can't stress this enough: I<If you do not reset your signal mask when
		2544	you expect it to be empty, you have a race condition in your code>. This
		2545	is not a libev-specific thing, this is true for most event libraries.
		2546
		2547	=head3 The special problem of threads signal handling
		2548
		2549	POSIX threads has problematic signal handling semantics, specifically,
		2550	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2551	threads in a process block signals, which is hard to achieve.
		2552
		2553	When you want to use sigwait (or mix libev signal handling with your own
		2554	for the same signals), you can tackle this problem by globally blocking
		2555	all signals before creating any threads (or creating them with a fully set
		2556	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2557	loops. Then designate one thread as "signal receiver thread" which handles
		2558	these signals. You can pass on any signals that libev might be interested
		2559	in by calling C<ev_feed_signal>.
		2560
		2561	=head3 Watcher-Specific Functions and Data Members
599		2562
600	=over 4	2563	=over 4
601		2564
602	=item ev_signal_init (ev_signal *, callback, int signum)	2565	=item ev_signal_init (ev_signal *, callback, int signum)
603		2566
604	=item ev_signal_set (ev_signal *, int signum)	2567	=item ev_signal_set (ev_signal *, int signum)
605		2568
606	Configures the watcher to trigger on the given signal number (usually one	2569	Configures the watcher to trigger on the given signal number (usually one
607	of the C<SIGxxx> constants).	2570	of the C<SIGxxx> constants).
608		2571
		2572	=item int signum [read-only]
		2573
		2574	The signal the watcher watches out for.
		2575
609	=back	2576	=back
610		2577
		2578	=head3 Examples
		2579
		2580	Example: Try to exit cleanly on SIGINT.
		2581
		2582	static void
		2583	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2584	{
		2585	ev_break (loop, EVBREAK_ALL);
		2586	}
		2587
		2588	ev_signal signal_watcher;
		2589	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2590	ev_signal_start (loop, &signal_watcher);
		2591
		2592
611	=head2 ev_child - wait for pid status changes	2593	=head2 C<ev_child> - watch out for process status changes
612		2594
613	Child watchers trigger when your process receives a SIGCHLD in response to	2595	Child watchers trigger when your process receives a SIGCHLD in response to
614	some child status changes (most typically when a child of yours dies).	2596	some child status changes (most typically when a child of yours dies or
		2597	exits). It is permissible to install a child watcher I<after> the child
		2598	has been forked (which implies it might have already exited), as long
		2599	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2600	forking and then immediately registering a watcher for the child is fine,
		2601	but forking and registering a watcher a few event loop iterations later or
		2602	in the next callback invocation is not.
		2603
		2604	Only the default event loop is capable of handling signals, and therefore
		2605	you can only register child watchers in the default event loop.
		2606
		2607	Due to some design glitches inside libev, child watchers will always be
		2608	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2609	libev)
		2610
		2611	=head3 Process Interaction
		2612
		2613	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2614	initialised. This is necessary to guarantee proper behaviour even if the
		2615	first child watcher is started after the child exits. The occurrence
		2616	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2617	synchronously as part of the event loop processing. Libev always reaps all
		2618	children, even ones not watched.
		2619
		2620	=head3 Overriding the Built-In Processing
		2621
		2622	Libev offers no special support for overriding the built-in child
		2623	processing, but if your application collides with libev's default child
		2624	handler, you can override it easily by installing your own handler for
		2625	C<SIGCHLD> after initialising the default loop, and making sure the
		2626	default loop never gets destroyed. You are encouraged, however, to use an
		2627	event-based approach to child reaping and thus use libev's support for
		2628	that, so other libev users can use C<ev_child> watchers freely.
		2629
		2630	=head3 Stopping the Child Watcher
		2631
		2632	Currently, the child watcher never gets stopped, even when the
		2633	child terminates, so normally one needs to stop the watcher in the
		2634	callback. Future versions of libev might stop the watcher automatically
		2635	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2636	problem).
		2637
		2638	=head3 Watcher-Specific Functions and Data Members
615		2639
616	=over 4	2640	=over 4
617		2641
618	=item ev_child_init (ev_child *, callback, int pid)	2642	=item ev_child_init (ev_child *, callback, int pid, int trace)
619		2643
620	=item ev_child_set (ev_child *, int pid)	2644	=item ev_child_set (ev_child *, int pid, int trace)
621		2645
622	Configures the watcher to wait for status changes of process C<pid> (or	2646	Configures the watcher to wait for status changes of process C<pid> (or
623	I<any> process if C<pid> is specified as C<0>). The callback can look	2647	I<any> process if C<pid> is specified as C<0>). The callback can look
624	at the C<rstatus> member of the C<ev_child> watcher structure to see	2648	at the C<rstatus> member of the C<ev_child> watcher structure to see
625	the status word (use the macros from C<sys/wait.h>). The C<rpid> member	2649	the status word (use the macros from C<sys/wait.h> and see your systems
626	contains the pid of the process causing the status change.	2650	C<waitpid> documentation). The C<rpid> member contains the pid of the
		2651	process causing the status change. C<trace> must be either C<0> (only
		2652	activate the watcher when the process terminates) or C<1> (additionally
		2653	activate the watcher when the process is stopped or continued).
		2654
		2655	=item int pid [read-only]
		2656
		2657	The process id this watcher watches out for, or C<0>, meaning any process id.
		2658
		2659	=item int rpid [read-write]
		2660
		2661	The process id that detected a status change.
		2662
		2663	=item int rstatus [read-write]
		2664
		2665	The process exit/trace status caused by C<rpid> (see your systems
		2666	C<waitpid> and C<sys/wait.h> documentation for details).
627		2667
628	=back	2668	=back
629		2669
		2670	=head3 Examples
		2671
		2672	Example: C<fork()> a new process and install a child handler to wait for
		2673	its completion.
		2674
		2675	ev_child cw;
		2676
		2677	static void
		2678	child_cb (EV_P_ ev_child *w, int revents)
		2679	{
		2680	ev_child_stop (EV_A_ w);
		2681	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2682	}
		2683
		2684	pid_t pid = fork ();
		2685
		2686	if (pid < 0)
		2687	// error
		2688	else if (pid == 0)
		2689	{
		2690	// the forked child executes here
		2691	exit (1);
		2692	}
		2693	else
		2694	{
		2695	ev_child_init (&cw, child_cb, pid, 0);
		2696	ev_child_start (EV_DEFAULT_ &cw);
		2697	}
		2698
		2699
		2700	=head2 C<ev_stat> - did the file attributes just change?
		2701
		2702	This watches a file system path for attribute changes. That is, it calls
		2703	C<stat> on that path in regular intervals (or when the OS says it changed)
		2704	and sees if it changed compared to the last time, invoking the callback
		2705	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2706	happen after the watcher has been started will be reported.
		2707
		2708	The path does not need to exist: changing from "path exists" to "path does
		2709	not exist" is a status change like any other. The condition "path does not
		2710	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2711	C<st_nlink> field being zero (which is otherwise always forced to be at
		2712	least one) and all the other fields of the stat buffer having unspecified
		2713	contents.
		2714
		2715	The path I<must not> end in a slash or contain special components such as
		2716	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2717	your working directory changes, then the behaviour is undefined.
		2718
		2719	Since there is no portable change notification interface available, the
		2720	portable implementation simply calls C<stat(2)> regularly on the path
		2721	to see if it changed somehow. You can specify a recommended polling
		2722	interval for this case. If you specify a polling interval of C<0> (highly
		2723	recommended!) then a I<suitable, unspecified default> value will be used
		2724	(which you can expect to be around five seconds, although this might
		2725	change dynamically). Libev will also impose a minimum interval which is
		2726	currently around C<0.1>, but that's usually overkill.
		2727
		2728	This watcher type is not meant for massive numbers of stat watchers,
		2729	as even with OS-supported change notifications, this can be
		2730	resource-intensive.
		2731
		2732	At the time of this writing, the only OS-specific interface implemented
		2733	is the Linux inotify interface (implementing kqueue support is left as an
		2734	exercise for the reader. Note, however, that the author sees no way of
		2735	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2736
		2737	=head3 ABI Issues (Largefile Support)
		2738
		2739	Libev by default (unless the user overrides this) uses the default
		2740	compilation environment, which means that on systems with large file
		2741	support disabled by default, you get the 32 bit version of the stat
		2742	structure. When using the library from programs that change the ABI to
		2743	use 64 bit file offsets the programs will fail. In that case you have to
		2744	compile libev with the same flags to get binary compatibility. This is
		2745	obviously the case with any flags that change the ABI, but the problem is
		2746	most noticeably displayed with ev_stat and large file support.
		2747
		2748	The solution for this is to lobby your distribution maker to make large
		2749	file interfaces available by default (as e.g. FreeBSD does) and not
		2750	optional. Libev cannot simply switch on large file support because it has
		2751	to exchange stat structures with application programs compiled using the
		2752	default compilation environment.
		2753
		2754	=head3 Inotify and Kqueue
		2755
		2756	When C<inotify (7)> support has been compiled into libev and present at
		2757	runtime, it will be used to speed up change detection where possible. The
		2758	inotify descriptor will be created lazily when the first C<ev_stat>
		2759	watcher is being started.
		2760
		2761	Inotify presence does not change the semantics of C<ev_stat> watchers
		2762	except that changes might be detected earlier, and in some cases, to avoid
		2763	making regular C<stat> calls. Even in the presence of inotify support
		2764	there are many cases where libev has to resort to regular C<stat> polling,
		2765	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2766	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2767	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2768	xfs are fully working) libev usually gets away without polling.
		2769
		2770	There is no support for kqueue, as apparently it cannot be used to
		2771	implement this functionality, due to the requirement of having a file
		2772	descriptor open on the object at all times, and detecting renames, unlinks
		2773	etc. is difficult.
		2774
		2775	=head3 C<stat ()> is a synchronous operation
		2776
		2777	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2778	the process. The exception are C<ev_stat> watchers - those call C<stat
		2779	()>, which is a synchronous operation.
		2780
		2781	For local paths, this usually doesn't matter: unless the system is very
		2782	busy or the intervals between stat's are large, a stat call will be fast,
		2783	as the path data is usually in memory already (except when starting the
		2784	watcher).
		2785
		2786	For networked file systems, calling C<stat ()> can block an indefinite
		2787	time due to network issues, and even under good conditions, a stat call
		2788	often takes multiple milliseconds.
		2789
		2790	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2791	paths, although this is fully supported by libev.
		2792
		2793	=head3 The special problem of stat time resolution
		2794
		2795	The C<stat ()> system call only supports full-second resolution portably,
		2796	and even on systems where the resolution is higher, most file systems
		2797	still only support whole seconds.
		2798
		2799	That means that, if the time is the only thing that changes, you can
		2800	easily miss updates: on the first update, C<ev_stat> detects a change and
		2801	calls your callback, which does something. When there is another update
		2802	within the same second, C<ev_stat> will be unable to detect unless the
		2803	stat data does change in other ways (e.g. file size).
		2804
		2805	The solution to this is to delay acting on a change for slightly more
		2806	than a second (or till slightly after the next full second boundary), using
		2807	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2808	ev_timer_again (loop, w)>).
		2809
		2810	The C<.02> offset is added to work around small timing inconsistencies
		2811	of some operating systems (where the second counter of the current time
		2812	might be be delayed. One such system is the Linux kernel, where a call to
		2813	C<gettimeofday> might return a timestamp with a full second later than
		2814	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2815	update file times then there will be a small window where the kernel uses
		2816	the previous second to update file times but libev might already execute
		2817	the timer callback).
		2818
		2819	=head3 Watcher-Specific Functions and Data Members
		2820
		2821	=over 4
		2822
		2823	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2824
		2825	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2826
		2827	Configures the watcher to wait for status changes of the given
		2828	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2829	be detected and should normally be specified as C<0> to let libev choose
		2830	a suitable value. The memory pointed to by C<path> must point to the same
		2831	path for as long as the watcher is active.
		2832
		2833	The callback will receive an C<EV_STAT> event when a change was detected,
		2834	relative to the attributes at the time the watcher was started (or the
		2835	last change was detected).
		2836
		2837	=item ev_stat_stat (loop, ev_stat *)
		2838
		2839	Updates the stat buffer immediately with new values. If you change the
		2840	watched path in your callback, you could call this function to avoid
		2841	detecting this change (while introducing a race condition if you are not
		2842	the only one changing the path). Can also be useful simply to find out the
		2843	new values.
		2844
		2845	=item ev_statdata attr [read-only]
		2846
		2847	The most-recently detected attributes of the file. Although the type is
		2848	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2849	suitable for your system, but you can only rely on the POSIX-standardised
		2850	members to be present. If the C<st_nlink> member is C<0>, then there was
		2851	some error while C<stat>ing the file.
		2852
		2853	=item ev_statdata prev [read-only]
		2854
		2855	The previous attributes of the file. The callback gets invoked whenever
		2856	C<prev> != C<attr>, or, more precisely, one or more of these members
		2857	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2858	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2859
		2860	=item ev_tstamp interval [read-only]
		2861
		2862	The specified interval.
		2863
		2864	=item const char *path [read-only]
		2865
		2866	The file system path that is being watched.
		2867
		2868	=back
		2869
		2870	=head3 Examples
		2871
		2872	Example: Watch C</etc/passwd> for attribute changes.
		2873
		2874	static void
		2875	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2876	{
		2877	/* /etc/passwd changed in some way */
		2878	if (w->attr.st_nlink)
		2879	{
		2880	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2881	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2882	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2883	}
		2884	else
		2885	/* you shalt not abuse printf for puts */
		2886	puts ("wow, /etc/passwd is not there, expect problems. "
		2887	"if this is windows, they already arrived\n");
		2888	}
		2889
		2890	...
		2891	ev_stat passwd;
		2892
		2893	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2894	ev_stat_start (loop, &passwd);
		2895
		2896	Example: Like above, but additionally use a one-second delay so we do not
		2897	miss updates (however, frequent updates will delay processing, too, so
		2898	one might do the work both on C<ev_stat> callback invocation I<and> on
		2899	C<ev_timer> callback invocation).
		2900
		2901	static ev_stat passwd;
		2902	static ev_timer timer;
		2903
		2904	static void
		2905	timer_cb (EV_P_ ev_timer *w, int revents)
		2906	{
		2907	ev_timer_stop (EV_A_ w);
		2908
		2909	/* now it's one second after the most recent passwd change */
		2910	}
		2911
		2912	static void
		2913	stat_cb (EV_P_ ev_stat *w, int revents)
		2914	{
		2915	/* reset the one-second timer */
		2916	ev_timer_again (EV_A_ &timer);
		2917	}
		2918
		2919	...
		2920	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2921	ev_stat_start (loop, &passwd);
		2922	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2923
		2924
630	=head2 ev_idle - when you've got nothing better to do	2925	=head2 C<ev_idle> - when you've got nothing better to do...
631		2926
632	Idle watchers trigger events when there are no other I/O or timer (or	2927	Idle watchers trigger events when no other events of the same or higher
633	periodic) events pending. That is, as long as your process is busy	2928	priority are pending (prepare, check and other idle watchers do not count
634	handling sockets or timeouts it will not be called. But when your process	2929	as receiving "events").
635	is idle all idle watchers are being called again and again - until	2930
		2931	That is, as long as your process is busy handling sockets or timeouts
		2932	(or even signals, imagine) of the same or higher priority it will not be
		2933	triggered. But when your process is idle (or only lower-priority watchers
		2934	are pending), the idle watchers are being called once per event loop
636	stopped, that is, or your process receives more events.	2935	iteration - until stopped, that is, or your process receives more events
		2936	and becomes busy again with higher priority stuff.
637		2937
638	The most noteworthy effect is that as long as any idle watchers are	2938	The most noteworthy effect is that as long as any idle watchers are
639	active, the process will not block when waiting for new events.	2939	active, the process will not block when waiting for new events.
640		2940
641	Apart from keeping your process non-blocking (which is a useful	2941	Apart from keeping your process non-blocking (which is a useful
642	effect on its own sometimes), idle watchers are a good place to do	2942	effect on its own sometimes), idle watchers are a good place to do
643	"pseudo-background processing", or delay processing stuff to after the	2943	"pseudo-background processing", or delay processing stuff to after the
644	event loop has handled all outstanding events.	2944	event loop has handled all outstanding events.
645		2945
		2946	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2947
		2948	As long as there is at least one active idle watcher, libev will never
		2949	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2950	For this to work, the idle watcher doesn't need to be invoked at all - the
		2951	lowest priority will do.
		2952
		2953	This mode of operation can be useful together with an C<ev_check> watcher,
		2954	to do something on each event loop iteration - for example to balance load
		2955	between different connections.
		2956
		2957	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2958	example.
		2959
		2960	=head3 Watcher-Specific Functions and Data Members
		2961
646	=over 4	2962	=over 4
647		2963
648	=item ev_idle_init (ev_signal *, callback)	2964	=item ev_idle_init (ev_idle *, callback)
649		2965
650	Initialises and configures the idle watcher - it has no parameters of any	2966	Initialises and configures the idle watcher - it has no parameters of any
651	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2967	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
652	believe me.	2968	believe me.
653		2969
654	=back	2970	=back
655		2971
656	=head2 prepare and check - your hooks into the event loop	2972	=head3 Examples
657		2973
		2974	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2975	callback, free it. Also, use no error checking, as usual.
		2976
		2977	static void
		2978	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2979	{
		2980	// stop the watcher
		2981	ev_idle_stop (loop, w);
		2982
		2983	// now we can free it
		2984	free (w);
		2985
		2986	// now do something you wanted to do when the program has
		2987	// no longer anything immediate to do.
		2988	}
		2989
		2990	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2991	ev_idle_init (idle_watcher, idle_cb);
		2992	ev_idle_start (loop, idle_watcher);
		2993
		2994
		2995	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
		2996
658	Prepare and check watchers usually (but not always) are used in	2997	Prepare and check watchers are often (but not always) used in pairs:
659	tandom. Prepare watchers get invoked before the process blocks and check	2998	prepare watchers get invoked before the process blocks and check watchers
660	watchers afterwards.	2999	afterwards.
661		3000
		3001	You I<must not> call C<ev_run> (or similar functions that enter the
		3002	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
		3003	C<ev_check> watchers. Other loops than the current one are fine,
		3004	however. The rationale behind this is that you do not need to check
		3005	for recursion in those watchers, i.e. the sequence will always be
		3006	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
		3007	kind they will always be called in pairs bracketing the blocking call.
		3008
662	Their main purpose is to integrate other event mechanisms into libev. This	3009	Their main purpose is to integrate other event mechanisms into libev and
663	could be used, for example, to track variable changes, implement your own	3010	their use is somewhat advanced. They could be used, for example, to track
664	watchers, integrate net-snmp or a coroutine library and lots more.	3011	variable changes, implement your own watchers, integrate net-snmp or a
		3012	coroutine library and lots more. They are also occasionally useful if
		3013	you cache some data and want to flush it before blocking (for example,
		3014	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		3015	watcher).
665		3016
666	This is done by examining in each prepare call which file descriptors need	3017	This is done by examining in each prepare call which file descriptors
667	to be watched by the other library, registering ev_io watchers for them	3018	need to be watched by the other library, registering C<ev_io> watchers
668	and starting an ev_timer watcher for any timeouts (many libraries provide	3019	for them and starting an C<ev_timer> watcher for any timeouts (many
669	just this functionality). Then, in the check watcher you check for any	3020	libraries provide exactly this functionality). Then, in the check watcher,
670	events that occured (by making your callbacks set soem flags for example)	3021	you check for any events that occurred (by checking the pending status
671	and call back into the library.	3022	of all watchers and stopping them) and call back into the library. The
		3023	I/O and timer callbacks will never actually be called (but must be valid
		3024	nevertheless, because you never know, you know?).
672		3025
673	As another example, the perl Coro module uses these hooks to integrate	3026	As another example, the Perl Coro module uses these hooks to integrate
674	coroutines into libev programs, by yielding to other active coroutines	3027	coroutines into libev programs, by yielding to other active coroutines
675	during each prepare and only letting the process block if no coroutines	3028	during each prepare and only letting the process block if no coroutines
676	are ready to run.	3029	are ready to run (it's actually more complicated: it only runs coroutines
		3030	with priority higher than or equal to the event loop and one coroutine
		3031	of lower priority, but only once, using idle watchers to keep the event
		3032	loop from blocking if lower-priority coroutines are active, thus mapping
		3033	low-priority coroutines to idle/background tasks).
		3034
		3035	When used for this purpose, it is recommended to give C<ev_check> watchers
		3036	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
		3037	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		3038	watchers).
		3039
		3040	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		3041	activate ("feed") events into libev. While libev fully supports this, they
		3042	might get executed before other C<ev_check> watchers did their job. As
		3043	C<ev_check> watchers are often used to embed other (non-libev) event
		3044	loops those other event loops might be in an unusable state until their
		3045	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		3046	others).
		3047
		3048	=head3 Abusing an C<ev_check> watcher for its side-effect
		3049
		3050	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		3051	useful because they are called once per event loop iteration. For
		3052	example, if you want to handle a large number of connections fairly, you
		3053	normally only do a bit of work for each active connection, and if there
		3054	is more work to do, you wait for the next event loop iteration, so other
		3055	connections have a chance of making progress.
		3056
		3057	Using an C<ev_check> watcher is almost enough: it will be called on the
		3058	next event loop iteration. However, that isn't as soon as possible -
		3059	without external events, your C<ev_check> watcher will not be invoked.
		3060
		3061	This is where C<ev_idle> watchers come in handy - all you need is a
		3062	single global idle watcher that is active as long as you have one active
		3063	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		3064	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		3065	invoked. Neither watcher alone can do that.
		3066
		3067	=head3 Watcher-Specific Functions and Data Members
677		3068
678	=over 4	3069	=over 4
679		3070
680	=item ev_prepare_init (ev_prepare *, callback)	3071	=item ev_prepare_init (ev_prepare *, callback)
681		3072
682	=item ev_check_init (ev_check *, callback)	3073	=item ev_check_init (ev_check *, callback)
683		3074
684	Initialises and configures the prepare or check watcher - they have no	3075	Initialises and configures the prepare or check watcher - they have no
685	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	3076	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
686	macros, but using them is utterly, utterly pointless.	3077	macros, but using them is utterly, utterly, utterly and completely
		3078	pointless.
687		3079
688	=back	3080	=back
689		3081
		3082	=head3 Examples
		3083
		3084	There are a number of principal ways to embed other event loops or modules
		3085	into libev. Here are some ideas on how to include libadns into libev
		3086	(there is a Perl module named C<EV::ADNS> that does this, which you could
		3087	use as a working example. Another Perl module named C<EV::Glib> embeds a
		3088	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		3089	Glib event loop).
		3090
		3091	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		3092	and in a check watcher, destroy them and call into libadns. What follows
		3093	is pseudo-code only of course. This requires you to either use a low
		3094	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		3095	the callbacks for the IO/timeout watchers might not have been called yet.
		3096
		3097	static ev_io iow [nfd];
		3098	static ev_timer tw;
		3099
		3100	static void
		3101	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		3102	{
		3103	}
		3104
		3105	// create io watchers for each fd and a timer before blocking
		3106	static void
		3107	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		3108	{
		3109	int timeout = 3600000;
		3110	struct pollfd fds [nfd];
		3111	// actual code will need to loop here and realloc etc.
		3112	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		3113
		3114	/* the callback is illegal, but won't be called as we stop during check */
		3115	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		3116	ev_timer_start (loop, &tw);
		3117
		3118	// create one ev_io per pollfd
		3119	for (int i = 0; i < nfd; ++i)
		3120	{
		3121	ev_io_init (iow + i, io_cb, fds [i].fd,
		3122	((fds [i].events & POLLIN ? EV_READ : 0)
		3123	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		3124
		3125	fds [i].revents = 0;
		3126	ev_io_start (loop, iow + i);
		3127	}
		3128	}
		3129
		3130	// stop all watchers after blocking
		3131	static void
		3132	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		3133	{
		3134	ev_timer_stop (loop, &tw);
		3135
		3136	for (int i = 0; i < nfd; ++i)
		3137	{
		3138	// set the relevant poll flags
		3139	// could also call adns_processreadable etc. here
		3140	struct pollfd *fd = fds + i;
		3141	int revents = ev_clear_pending (iow + i);
		3142	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		3143	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		3144
		3145	// now stop the watcher
		3146	ev_io_stop (loop, iow + i);
		3147	}
		3148
		3149	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		3150	}
		3151
		3152	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		3153	in the prepare watcher and would dispose of the check watcher.
		3154
		3155	Method 3: If the module to be embedded supports explicit event
		3156	notification (libadns does), you can also make use of the actual watcher
		3157	callbacks, and only destroy/create the watchers in the prepare watcher.
		3158
		3159	static void
		3160	timer_cb (EV_P_ ev_timer *w, int revents)
		3161	{
		3162	adns_state ads = (adns_state)w->data;
		3163	update_now (EV_A);
		3164
		3165	adns_processtimeouts (ads, &tv_now);
		3166	}
		3167
		3168	static void
		3169	io_cb (EV_P_ ev_io *w, int revents)
		3170	{
		3171	adns_state ads = (adns_state)w->data;
		3172	update_now (EV_A);
		3173
		3174	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		3175	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		3176	}
		3177
		3178	// do not ever call adns_afterpoll
		3179
		3180	Method 4: Do not use a prepare or check watcher because the module you
		3181	want to embed is not flexible enough to support it. Instead, you can
		3182	override their poll function. The drawback with this solution is that the
		3183	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		3184	this approach, effectively embedding EV as a client into the horrible
		3185	libglib event loop.
		3186
		3187	static gint
		3188	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		3189	{
		3190	int got_events = 0;
		3191
		3192	for (n = 0; n < nfds; ++n)
		3193	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		3194
		3195	if (timeout >= 0)
		3196	// create/start timer
		3197
		3198	// poll
		3199	ev_run (EV_A_ 0);
		3200
		3201	// stop timer again
		3202	if (timeout >= 0)
		3203	ev_timer_stop (EV_A_ &to);
		3204
		3205	// stop io watchers again - their callbacks should have set
		3206	for (n = 0; n < nfds; ++n)
		3207	ev_io_stop (EV_A_ iow [n]);
		3208
		3209	return got_events;
		3210	}
		3211
		3212
		3213	=head2 C<ev_embed> - when one backend isn't enough...
		3214
		3215	This is a rather advanced watcher type that lets you embed one event loop
		3216	into another (currently only C<ev_io> events are supported in the embedded
		3217	loop, other types of watchers might be handled in a delayed or incorrect
		3218	fashion and must not be used).
		3219
		3220	There are primarily two reasons you would want that: work around bugs and
		3221	prioritise I/O.
		3222
		3223	As an example for a bug workaround, the kqueue backend might only support
		3224	sockets on some platform, so it is unusable as generic backend, but you
		3225	still want to make use of it because you have many sockets and it scales
		3226	so nicely. In this case, you would create a kqueue-based loop and embed
		3227	it into your default loop (which might use e.g. poll). Overall operation
		3228	will be a bit slower because first libev has to call C<poll> and then
		3229	C<kevent>, but at least you can use both mechanisms for what they are
		3230	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		3231
		3232	As for prioritising I/O: under rare circumstances you have the case where
		3233	some fds have to be watched and handled very quickly (with low latency),
		3234	and even priorities and idle watchers might have too much overhead. In
		3235	this case you would put all the high priority stuff in one loop and all
		3236	the rest in a second one, and embed the second one in the first.
		3237
		3238	As long as the watcher is active, the callback will be invoked every
		3239	time there might be events pending in the embedded loop. The callback
		3240	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		3241	sweep and invoke their callbacks (the callback doesn't need to invoke the
		3242	C<ev_embed_sweep> function directly, it could also start an idle watcher
		3243	to give the embedded loop strictly lower priority for example).
		3244
		3245	You can also set the callback to C<0>, in which case the embed watcher
		3246	will automatically execute the embedded loop sweep whenever necessary.
		3247
		3248	Fork detection will be handled transparently while the C<ev_embed> watcher
		3249	is active, i.e., the embedded loop will automatically be forked when the
		3250	embedding loop forks. In other cases, the user is responsible for calling
		3251	C<ev_loop_fork> on the embedded loop.
		3252
		3253	Unfortunately, not all backends are embeddable: only the ones returned by
		3254	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		3255	portable one.
		3256
		3257	So when you want to use this feature you will always have to be prepared
		3258	that you cannot get an embeddable loop. The recommended way to get around
		3259	this is to have a separate variables for your embeddable loop, try to
		3260	create it, and if that fails, use the normal loop for everything.
		3261
		3262	=head3 C<ev_embed> and fork
		3263
		3264	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3265	automatically be applied to the embedded loop as well, so no special
		3266	fork handling is required in that case. When the watcher is not running,
		3267	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3268	as applicable.
		3269
		3270	=head3 Watcher-Specific Functions and Data Members
		3271
		3272	=over 4
		3273
		3274	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		3275
		3276	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
		3277
		3278	Configures the watcher to embed the given loop, which must be
		3279	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		3280	invoked automatically, otherwise it is the responsibility of the callback
		3281	to invoke it (it will continue to be called until the sweep has been done,
		3282	if you do not want that, you need to temporarily stop the embed watcher).
		3283
		3284	=item ev_embed_sweep (loop, ev_embed *)
		3285
		3286	Make a single, non-blocking sweep over the embedded loop. This works
		3287	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3288	appropriate way for embedded loops.
		3289
		3290	=item struct ev_loop *other [read-only]
		3291
		3292	The embedded event loop.
		3293
		3294	=back
		3295
		3296	=head3 Examples
		3297
		3298	Example: Try to get an embeddable event loop and embed it into the default
		3299	event loop. If that is not possible, use the default loop. The default
		3300	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3301	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3302	used).
		3303
		3304	struct ev_loop *loop_hi = ev_default_init (0);
		3305	struct ev_loop *loop_lo = 0;
		3306	ev_embed embed;
		3307
		3308	// see if there is a chance of getting one that works
		3309	// (remember that a flags value of 0 means autodetection)
		3310	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3311	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3312	: 0;
		3313
		3314	// if we got one, then embed it, otherwise default to loop_hi
		3315	if (loop_lo)
		3316	{
		3317	ev_embed_init (&embed, 0, loop_lo);
		3318	ev_embed_start (loop_hi, &embed);
		3319	}
		3320	else
		3321	loop_lo = loop_hi;
		3322
		3323	Example: Check if kqueue is available but not recommended and create
		3324	a kqueue backend for use with sockets (which usually work with any
		3325	kqueue implementation). Store the kqueue/socket-only event loop in
		3326	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3327
		3328	struct ev_loop *loop = ev_default_init (0);
		3329	struct ev_loop *loop_socket = 0;
		3330	ev_embed embed;
		3331
		3332	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3333	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3334	{
		3335	ev_embed_init (&embed, 0, loop_socket);
		3336	ev_embed_start (loop, &embed);
		3337	}
		3338
		3339	if (!loop_socket)
		3340	loop_socket = loop;
		3341
		3342	// now use loop_socket for all sockets, and loop for everything else
		3343
		3344
		3345	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3346
		3347	Fork watchers are called when a C<fork ()> was detected (usually because
		3348	whoever is a good citizen cared to tell libev about it by calling
		3349	C<ev_loop_fork>). The invocation is done before the event loop blocks next
		3350	and before C<ev_check> watchers are being called, and only in the child
		3351	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
		3352	and calls it in the wrong process, the fork handlers will be invoked, too,
		3353	of course.
		3354
		3355	=head3 The special problem of life after fork - how is it possible?
		3356
		3357	Most uses of C<fork ()> consist of forking, then some simple calls to set
		3358	up/change the process environment, followed by a call to C<exec()>. This
		3359	sequence should be handled by libev without any problems.
		3360
		3361	This changes when the application actually wants to do event handling
		3362	in the child, or both parent in child, in effect "continuing" after the
		3363	fork.
		3364
		3365	The default mode of operation (for libev, with application help to detect
		3366	forks) is to duplicate all the state in the child, as would be expected
		3367	when I<either> the parent I<or> the child process continues.
		3368
		3369	When both processes want to continue using libev, then this is usually the
		3370	wrong result. In that case, usually one process (typically the parent) is
		3371	supposed to continue with all watchers in place as before, while the other
		3372	process typically wants to start fresh, i.e. without any active watchers.
		3373
		3374	The cleanest and most efficient way to achieve that with libev is to
		3375	simply create a new event loop, which of course will be "empty", and
		3376	use that for new watchers. This has the advantage of not touching more
		3377	memory than necessary, and thus avoiding the copy-on-write, and the
		3378	disadvantage of having to use multiple event loops (which do not support
		3379	signal watchers).
		3380
		3381	When this is not possible, or you want to use the default loop for
		3382	other reasons, then in the process that wants to start "fresh", call
		3383	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3384	Destroying the default loop will "orphan" (not stop) all registered
		3385	watchers, so you have to be careful not to execute code that modifies
		3386	those watchers. Note also that in that case, you have to re-register any
		3387	signal watchers.
		3388
		3389	=head3 Watcher-Specific Functions and Data Members
		3390
		3391	=over 4
		3392
		3393	=item ev_fork_init (ev_fork *, callback)
		3394
		3395	Initialises and configures the fork watcher - it has no parameters of any
		3396	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3397	really.
		3398
		3399	=back
		3400
		3401
		3402	=head2 C<ev_cleanup> - even the best things end
		3403
		3404	Cleanup watchers are called just before the event loop is being destroyed
		3405	by a call to C<ev_loop_destroy>.
		3406
		3407	While there is no guarantee that the event loop gets destroyed, cleanup
		3408	watchers provide a convenient method to install cleanup hooks for your
		3409	program, worker threads and so on - you just to make sure to destroy the
		3410	loop when you want them to be invoked.
		3411
		3412	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3413	all other watchers, they do not keep a reference to the event loop (which
		3414	makes a lot of sense if you think about it). Like all other watchers, you
		3415	can call libev functions in the callback, except C<ev_cleanup_start>.
		3416
		3417	=head3 Watcher-Specific Functions and Data Members
		3418
		3419	=over 4
		3420
		3421	=item ev_cleanup_init (ev_cleanup *, callback)
		3422
		3423	Initialises and configures the cleanup watcher - it has no parameters of
		3424	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3425	pointless, I assure you.
		3426
		3427	=back
		3428
		3429	Example: Register an atexit handler to destroy the default loop, so any
		3430	cleanup functions are called.
		3431
		3432	static void
		3433	program_exits (void)
		3434	{
		3435	ev_loop_destroy (EV_DEFAULT_UC);
		3436	}
		3437
		3438	...
		3439	atexit (program_exits);
		3440
		3441
		3442	=head2 C<ev_async> - how to wake up an event loop
		3443
		3444	In general, you cannot use an C<ev_loop> from multiple threads or other
		3445	asynchronous sources such as signal handlers (as opposed to multiple event
		3446	loops - those are of course safe to use in different threads).
		3447
		3448	Sometimes, however, you need to wake up an event loop you do not control,
		3449	for example because it belongs to another thread. This is what C<ev_async>
		3450	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3451	it by calling C<ev_async_send>, which is thread- and signal safe.
		3452
		3453	This functionality is very similar to C<ev_signal> watchers, as signals,
		3454	too, are asynchronous in nature, and signals, too, will be compressed
		3455	(i.e. the number of callback invocations may be less than the number of
		3456	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
		3457	of "global async watchers" by using a watcher on an otherwise unused
		3458	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3459	even without knowing which loop owns the signal.
		3460
		3461	=head3 Queueing
		3462
		3463	C<ev_async> does not support queueing of data in any way. The reason
		3464	is that the author does not know of a simple (or any) algorithm for a
		3465	multiple-writer-single-reader queue that works in all cases and doesn't
		3466	need elaborate support such as pthreads or unportable memory access
		3467	semantics.
		3468
		3469	That means that if you want to queue data, you have to provide your own
		3470	queue. But at least I can tell you how to implement locking around your
		3471	queue:
		3472
		3473	=over 4
		3474
		3475	=item queueing from a signal handler context
		3476
		3477	To implement race-free queueing, you simply add to the queue in the signal
		3478	handler but you block the signal handler in the watcher callback. Here is
		3479	an example that does that for some fictitious SIGUSR1 handler:
		3480
		3481	static ev_async mysig;
		3482
		3483	static void
		3484	sigusr1_handler (void)
		3485	{
		3486	sometype data;
		3487
		3488	// no locking etc.
		3489	queue_put (data);
		3490	ev_async_send (EV_DEFAULT_ &mysig);
		3491	}
		3492
		3493	static void
		3494	mysig_cb (EV_P_ ev_async *w, int revents)
		3495	{
		3496	sometype data;
		3497	sigset_t block, prev;
		3498
		3499	sigemptyset (&block);
		3500	sigaddset (&block, SIGUSR1);
		3501	sigprocmask (SIG_BLOCK, &block, &prev);
		3502
		3503	while (queue_get (&data))
		3504	process (data);
		3505
		3506	if (sigismember (&prev, SIGUSR1)
		3507	sigprocmask (SIG_UNBLOCK, &block, 0);
		3508	}
		3509
		3510	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3511	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3512	either...).
		3513
		3514	=item queueing from a thread context
		3515
		3516	The strategy for threads is different, as you cannot (easily) block
		3517	threads but you can easily preempt them, so to queue safely you need to
		3518	employ a traditional mutex lock, such as in this pthread example:
		3519
		3520	static ev_async mysig;
		3521	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3522
		3523	static void
		3524	otherthread (void)
		3525	{
		3526	// only need to lock the actual queueing operation
		3527	pthread_mutex_lock (&mymutex);
		3528	queue_put (data);
		3529	pthread_mutex_unlock (&mymutex);
		3530
		3531	ev_async_send (EV_DEFAULT_ &mysig);
		3532	}
		3533
		3534	static void
		3535	mysig_cb (EV_P_ ev_async *w, int revents)
		3536	{
		3537	pthread_mutex_lock (&mymutex);
		3538
		3539	while (queue_get (&data))
		3540	process (data);
		3541
		3542	pthread_mutex_unlock (&mymutex);
		3543	}
		3544
		3545	=back
		3546
		3547
		3548	=head3 Watcher-Specific Functions and Data Members
		3549
		3550	=over 4
		3551
		3552	=item ev_async_init (ev_async *, callback)
		3553
		3554	Initialises and configures the async watcher - it has no parameters of any
		3555	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3556	trust me.
		3557
		3558	=item ev_async_send (loop, ev_async *)
		3559
		3560	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3561	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3562	returns.
		3563
		3564	Unlike C<ev_feed_event>, this call is safe to do from other threads,
		3565	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
		3566	embedding section below on what exactly this means).
		3567
		3568	Note that, as with other watchers in libev, multiple events might get
		3569	compressed into a single callback invocation (another way to look at
		3570	this is that C<ev_async> watchers are level-triggered: they are set on
		3571	C<ev_async_send>, reset when the event loop detects that).
		3572
		3573	This call incurs the overhead of at most one extra system call per event
		3574	loop iteration, if the event loop is blocked, and no syscall at all if
		3575	the event loop (or your program) is processing events. That means that
		3576	repeated calls are basically free (there is no need to avoid calls for
		3577	performance reasons) and that the overhead becomes smaller (typically
		3578	zero) under load.
		3579
		3580	=item bool = ev_async_pending (ev_async *)
		3581
		3582	Returns a non-zero value when C<ev_async_send> has been called on the
		3583	watcher but the event has not yet been processed (or even noted) by the
		3584	event loop.
		3585
		3586	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3587	the loop iterates next and checks for the watcher to have become active,
		3588	it will reset the flag again. C<ev_async_pending> can be used to very
		3589	quickly check whether invoking the loop might be a good idea.
		3590
		3591	Not that this does I<not> check whether the watcher itself is pending,
		3592	only whether it has been requested to make this watcher pending: there
		3593	is a time window between the event loop checking and resetting the async
		3594	notification, and the callback being invoked.
		3595
		3596	=back
		3597
		3598
690	=head1 OTHER FUNCTIONS	3599	=head1 OTHER FUNCTIONS
691		3600
692	There are some other fucntions of possible interest. Described. Here. Now.	3601	There are some other functions of possible interest. Described. Here. Now.
693		3602
694	=over 4	3603	=over 4
695		3604
696	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3605	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
697		3606
698	This function combines a simple timer and an I/O watcher, calls your	3607	This function combines a simple timer and an I/O watcher, calls your
699	callback on whichever event happens first and automatically stop both	3608	callback on whichever event happens first and automatically stops both
700	watchers. This is useful if you want to wait for a single event on an fd	3609	watchers. This is useful if you want to wait for a single event on an fd
701	or timeout without havign to allocate/configure/start/stop/free one or	3610	or timeout without having to allocate/configure/start/stop/free one or
702	more watchers yourself.	3611	more watchers yourself.
703		3612
704	If C<fd> is less than 0, then no I/O watcher will be started and events is	3613	If C<fd> is less than 0, then no I/O watcher will be started and the
705	ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set	3614	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
706	will be craeted and started.	3615	the given C<fd> and C<events> set will be created and started.
707		3616
708	If C<timeout> is less than 0, then no timeout watcher will be	3617	If C<timeout> is less than 0, then no timeout watcher will be
709	started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat	3618	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
710	= 0) will be started.	3619	repeat = 0) will be started. C<0> is a valid timeout.
711		3620
712	The callback has the type C<void (*cb)(int revents, void *arg)> and	3621	The callback has the type C<void (*cb)(int revents, void *arg)> and is
713	gets passed an events set (normally a combination of EV_ERROR, EV_READ,	3622	passed an C<revents> set like normal event callbacks (a combination of
714	EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>:	3623	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
		3624	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3625	a timeout and an io event at the same time - you probably should give io
		3626	events precedence.
715		3627
		3628	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3629
716	static void stdin_ready (int revents, void *arg)	3630	static void stdin_ready (int revents, void *arg)
		3631	{
		3632	if (revents & EV_READ)
		3633	/* stdin might have data for us, joy! */;
		3634	else if (revents & EV_TIMER)
		3635	/* doh, nothing entered */;
		3636	}
		3637
		3638	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
		3639
		3640	=item ev_feed_fd_event (loop, int fd, int revents)
		3641
		3642	Feed an event on the given fd, as if a file descriptor backend detected
		3643	the given events.
		3644
		3645	=item ev_feed_signal_event (loop, int signum)
		3646
		3647	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3648	which is async-safe.
		3649
		3650	=back
		3651
		3652
		3653	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3654
		3655	This section explains some common idioms that are not immediately
		3656	obvious. Note that examples are sprinkled over the whole manual, and this
		3657	section only contains stuff that wouldn't fit anywhere else.
		3658
		3659	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3660
		3661	Each watcher has, by default, a C<void *data> member that you can read
		3662	or modify at any time: libev will completely ignore it. This can be used
		3663	to associate arbitrary data with your watcher. If you need more data and
		3664	don't want to allocate memory separately and store a pointer to it in that
		3665	data member, you can also "subclass" the watcher type and provide your own
		3666	data:
		3667
		3668	struct my_io
		3669	{
		3670	ev_io io;
		3671	int otherfd;
		3672	void *somedata;
		3673	struct whatever *mostinteresting;
		3674	};
		3675
		3676	...
		3677	struct my_io w;
		3678	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3679
		3680	And since your callback will be called with a pointer to the watcher, you
		3681	can cast it back to your own type:
		3682
		3683	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3684	{
		3685	struct my_io *w = (struct my_io *)w_;
		3686	...
		3687	}
		3688
		3689	More interesting and less C-conformant ways of casting your callback
		3690	function type instead have been omitted.
		3691
		3692	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3693
		3694	Another common scenario is to use some data structure with multiple
		3695	embedded watchers, in effect creating your own watcher that combines
		3696	multiple libev event sources into one "super-watcher":
		3697
		3698	struct my_biggy
		3699	{
		3700	int some_data;
		3701	ev_timer t1;
		3702	ev_timer t2;
		3703	}
		3704
		3705	In this case getting the pointer to C<my_biggy> is a bit more
		3706	complicated: Either you store the address of your C<my_biggy> struct in
		3707	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3708	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3709	real programmers):
		3710
		3711	#include <stddef.h>
		3712
		3713	static void
		3714	t1_cb (EV_P_ ev_timer *w, int revents)
		3715	{
		3716	struct my_biggy big = (struct my_biggy *)
		3717	(((char *)w) - offsetof (struct my_biggy, t1));
		3718	}
		3719
		3720	static void
		3721	t2_cb (EV_P_ ev_timer *w, int revents)
		3722	{
		3723	struct my_biggy big = (struct my_biggy *)
		3724	(((char *)w) - offsetof (struct my_biggy, t2));
		3725	}
		3726
		3727	=head2 AVOIDING FINISHING BEFORE RETURNING
		3728
		3729	Often you have structures like this in event-based programs:
		3730
		3731	callback ()
717	{	3732	{
718	if (revents & EV_TIMEOUT)	3733	free (request);
719	/* doh, nothing entered */
720	else if (revents & EV_READ)
721	/* stdin might have data for us, joy! */
722	}	3734	}
723		3735
724	ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0);	3736	request = start_new_request (..., callback);
725		3737
726	=item ev_feed_event (loop, watcher, int events)	3738	The intent is to start some "lengthy" operation. The C<request> could be
		3739	used to cancel the operation, or do other things with it.
727		3740
728	Feeds the given event set into the event loop, as if the specified event	3741	It's not uncommon to have code paths in C<start_new_request> that
729	has happened for the specified watcher (which must be a pointer to an	3742	immediately invoke the callback, for example, to report errors. Or you add
730	initialised but not necessarily active event watcher).	3743	some caching layer that finds that it can skip the lengthy aspects of the
		3744	operation and simply invoke the callback with the result.
731		3745
732	=item ev_feed_fd_event (loop, int fd, int revents)	3746	The problem here is that this will happen I<before> C<start_new_request>
		3747	has returned, so C<request> is not set.
733		3748
734	Feed an event on the given fd, as if a file descriptor backend detected it.	3749	Even if you pass the request by some safer means to the callback, you
		3750	might want to do something to the request after starting it, such as
		3751	canceling it, which probably isn't working so well when the callback has
		3752	already been invoked.
735		3753
736	=item ev_feed_signal_event (loop, int signum)	3754	A common way around all these issues is to make sure that
		3755	C<start_new_request> I<always> returns before the callback is invoked. If
		3756	C<start_new_request> immediately knows the result, it can artificially
		3757	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3758	example, or more sneakily, by reusing an existing (stopped) watcher and
		3759	pushing it into the pending queue:
737		3760
738	Feed an event as if the given signal occured (loop must be the default loop!).	3761	ev_set_cb (watcher, callback);
		3762	ev_feed_event (EV_A_ watcher, 0);
		3763
		3764	This way, C<start_new_request> can safely return before the callback is
		3765	invoked, while not delaying callback invocation too much.
		3766
		3767	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3768
		3769	Often (especially in GUI toolkits) there are places where you have
		3770	I<modal> interaction, which is most easily implemented by recursively
		3771	invoking C<ev_run>.
		3772
		3773	This brings the problem of exiting - a callback might want to finish the
		3774	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3775	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3776	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3777	other combination: In these cases, a simple C<ev_break> will not work.
		3778
		3779	The solution is to maintain "break this loop" variable for each C<ev_run>
		3780	invocation, and use a loop around C<ev_run> until the condition is
		3781	triggered, using C<EVRUN_ONCE>:
		3782
		3783	// main loop
		3784	int exit_main_loop = 0;
		3785
		3786	while (!exit_main_loop)
		3787	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3788
		3789	// in a modal watcher
		3790	int exit_nested_loop = 0;
		3791
		3792	while (!exit_nested_loop)
		3793	ev_run (EV_A_ EVRUN_ONCE);
		3794
		3795	To exit from any of these loops, just set the corresponding exit variable:
		3796
		3797	// exit modal loop
		3798	exit_nested_loop = 1;
		3799
		3800	// exit main program, after modal loop is finished
		3801	exit_main_loop = 1;
		3802
		3803	// exit both
		3804	exit_main_loop = exit_nested_loop = 1;
		3805
		3806	=head2 THREAD LOCKING EXAMPLE
		3807
		3808	Here is a fictitious example of how to run an event loop in a different
		3809	thread from where callbacks are being invoked and watchers are
		3810	created/added/removed.
		3811
		3812	For a real-world example, see the C<EV::Loop::Async> perl module,
		3813	which uses exactly this technique (which is suited for many high-level
		3814	languages).
		3815
		3816	The example uses a pthread mutex to protect the loop data, a condition
		3817	variable to wait for callback invocations, an async watcher to notify the
		3818	event loop thread and an unspecified mechanism to wake up the main thread.
		3819
		3820	First, you need to associate some data with the event loop:
		3821
		3822	typedef struct {
		3823	mutex_t lock; /* global loop lock */
		3824	ev_async async_w;
		3825	thread_t tid;
		3826	cond_t invoke_cv;
		3827	} userdata;
		3828
		3829	void prepare_loop (EV_P)
		3830	{
		3831	// for simplicity, we use a static userdata struct.
		3832	static userdata u;
		3833
		3834	ev_async_init (&u->async_w, async_cb);
		3835	ev_async_start (EV_A_ &u->async_w);
		3836
		3837	pthread_mutex_init (&u->lock, 0);
		3838	pthread_cond_init (&u->invoke_cv, 0);
		3839
		3840	// now associate this with the loop
		3841	ev_set_userdata (EV_A_ u);
		3842	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3843	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3844
		3845	// then create the thread running ev_run
		3846	pthread_create (&u->tid, 0, l_run, EV_A);
		3847	}
		3848
		3849	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3850	solely to wake up the event loop so it takes notice of any new watchers
		3851	that might have been added:
		3852
		3853	static void
		3854	async_cb (EV_P_ ev_async *w, int revents)
		3855	{
		3856	// just used for the side effects
		3857	}
		3858
		3859	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3860	protecting the loop data, respectively.
		3861
		3862	static void
		3863	l_release (EV_P)
		3864	{
		3865	userdata *u = ev_userdata (EV_A);
		3866	pthread_mutex_unlock (&u->lock);
		3867	}
		3868
		3869	static void
		3870	l_acquire (EV_P)
		3871	{
		3872	userdata *u = ev_userdata (EV_A);
		3873	pthread_mutex_lock (&u->lock);
		3874	}
		3875
		3876	The event loop thread first acquires the mutex, and then jumps straight
		3877	into C<ev_run>:
		3878
		3879	void *
		3880	l_run (void *thr_arg)
		3881	{
		3882	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3883
		3884	l_acquire (EV_A);
		3885	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3886	ev_run (EV_A_ 0);
		3887	l_release (EV_A);
		3888
		3889	return 0;
		3890	}
		3891
		3892	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3893	signal the main thread via some unspecified mechanism (signals? pipe
		3894	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3895	have been called (in a while loop because a) spurious wakeups are possible
		3896	and b) skipping inter-thread-communication when there are no pending
		3897	watchers is very beneficial):
		3898
		3899	static void
		3900	l_invoke (EV_P)
		3901	{
		3902	userdata *u = ev_userdata (EV_A);
		3903
		3904	while (ev_pending_count (EV_A))
		3905	{
		3906	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3907	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3908	}
		3909	}
		3910
		3911	Now, whenever the main thread gets told to invoke pending watchers, it
		3912	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3913	thread to continue:
		3914
		3915	static void
		3916	real_invoke_pending (EV_P)
		3917	{
		3918	userdata *u = ev_userdata (EV_A);
		3919
		3920	pthread_mutex_lock (&u->lock);
		3921	ev_invoke_pending (EV_A);
		3922	pthread_cond_signal (&u->invoke_cv);
		3923	pthread_mutex_unlock (&u->lock);
		3924	}
		3925
		3926	Whenever you want to start/stop a watcher or do other modifications to an
		3927	event loop, you will now have to lock:
		3928
		3929	ev_timer timeout_watcher;
		3930	userdata *u = ev_userdata (EV_A);
		3931
		3932	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3933
		3934	pthread_mutex_lock (&u->lock);
		3935	ev_timer_start (EV_A_ &timeout_watcher);
		3936	ev_async_send (EV_A_ &u->async_w);
		3937	pthread_mutex_unlock (&u->lock);
		3938
		3939	Note that sending the C<ev_async> watcher is required because otherwise
		3940	an event loop currently blocking in the kernel will have no knowledge
		3941	about the newly added timer. By waking up the loop it will pick up any new
		3942	watchers in the next event loop iteration.
		3943
		3944	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3945
		3946	While the overhead of a callback that e.g. schedules a thread is small, it
		3947	is still an overhead. If you embed libev, and your main usage is with some
		3948	kind of threads or coroutines, you might want to customise libev so that
		3949	doesn't need callbacks anymore.
		3950
		3951	Imagine you have coroutines that you can switch to using a function
		3952	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3953	and that due to some magic, the currently active coroutine is stored in a
		3954	global called C<current_coro>. Then you can build your own "wait for libev
		3955	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3956	the differing C<;> conventions):
		3957
		3958	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3959	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3960
		3961	That means instead of having a C callback function, you store the
		3962	coroutine to switch to in each watcher, and instead of having libev call
		3963	your callback, you instead have it switch to that coroutine.
		3964
		3965	A coroutine might now wait for an event with a function called
		3966	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3967	matter when, or whether the watcher is active or not when this function is
		3968	called):
		3969
		3970	void
		3971	wait_for_event (ev_watcher *w)
		3972	{
		3973	ev_set_cb (w, current_coro);
		3974	switch_to (libev_coro);
		3975	}
		3976
		3977	That basically suspends the coroutine inside C<wait_for_event> and
		3978	continues the libev coroutine, which, when appropriate, switches back to
		3979	this or any other coroutine.
		3980
		3981	You can do similar tricks if you have, say, threads with an event queue -
		3982	instead of storing a coroutine, you store the queue object and instead of
		3983	switching to a coroutine, you push the watcher onto the queue and notify
		3984	any waiters.
		3985
		3986	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		3987	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3988
		3989	// my_ev.h
		3990	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3991	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3992	#include "../libev/ev.h"
		3993
		3994	// my_ev.c
		3995	#define EV_H "my_ev.h"
		3996	#include "../libev/ev.c"
		3997
		3998	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3999	F<my_ev.c> into your project. When properly specifying include paths, you
		4000	can even use F<ev.h> as header file name directly.
		4001
		4002
		4003	=head1 LIBEVENT EMULATION
		4004
		4005	Libev offers a compatibility emulation layer for libevent. It cannot
		4006	emulate the internals of libevent, so here are some usage hints:
		4007
		4008	=over 4
		4009
		4010	=item * Only the libevent-1.4.1-beta API is being emulated.
		4011
		4012	This was the newest libevent version available when libev was implemented,
		4013	and is still mostly unchanged in 2010.
		4014
		4015	=item * Use it by including <event.h>, as usual.
		4016
		4017	=item * The following members are fully supported: ev_base, ev_callback,
		4018	ev_arg, ev_fd, ev_res, ev_events.
		4019
		4020	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		4021	maintained by libev, it does not work exactly the same way as in libevent (consider
		4022	it a private API).
		4023
		4024	=item * Priorities are not currently supported. Initialising priorities
		4025	will fail and all watchers will have the same priority, even though there
		4026	is an ev_pri field.
		4027
		4028	=item * In libevent, the last base created gets the signals, in libev, the
		4029	base that registered the signal gets the signals.
		4030
		4031	=item * Other members are not supported.
		4032
		4033	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		4034	to use the libev header file and library.
739		4035
740	=back	4036	=back
741		4037
		4038	=head1 C++ SUPPORT
		4039
		4040	=head2 C API
		4041
		4042	The normal C API should work fine when used from C++: both ev.h and the
		4043	libev sources can be compiled as C++. Therefore, code that uses the C API
		4044	will work fine.
		4045
		4046	Proper exception specifications might have to be added to callbacks passed
		4047	to libev: exceptions may be thrown only from watcher callbacks, all other
		4048	callbacks (allocator, syserr, loop acquire/release and periodic reschedule
		4049	callbacks) must not throw exceptions, and might need a C<noexcept>
		4050	specification. If you have code that needs to be compiled as both C and
		4051	C++ you can use the C<EV_NOEXCEPT> macro for this:
		4052
		4053	static void
		4054	fatal_error (const char *msg) EV_NOEXCEPT
		4055	{
		4056	perror (msg);
		4057	abort ();
		4058	}
		4059
		4060	...
		4061	ev_set_syserr_cb (fatal_error);
		4062
		4063	The only API functions that can currently throw exceptions are C<ev_run>,
		4064	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		4065	because it runs cleanup watchers).
		4066
		4067	Throwing exceptions in watcher callbacks is only supported if libev itself
		4068	is compiled with a C++ compiler or your C and C++ environments allow
		4069	throwing exceptions through C libraries (most do).
		4070
		4071	=head2 C++ API
		4072
		4073	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		4074	you to use some convenience methods to start/stop watchers and also change
		4075	the callback model to a model using method callbacks on objects.
		4076
		4077	To use it,
		4078
		4079	#include <ev++.h>
		4080
		4081	This automatically includes F<ev.h> and puts all of its definitions (many
		4082	of them macros) into the global namespace. All C++ specific things are
		4083	put into the C<ev> namespace. It should support all the same embedding
		4084	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		4085
		4086	Care has been taken to keep the overhead low. The only data member the C++
		4087	classes add (compared to plain C-style watchers) is the event loop pointer
		4088	that the watcher is associated with (or no additional members at all if
		4089	you disable C<EV_MULTIPLICITY> when embedding libev).
		4090
		4091	Currently, functions, static and non-static member functions and classes
		4092	with C<operator ()> can be used as callbacks. Other types should be easy
		4093	to add as long as they only need one additional pointer for context. If
		4094	you need support for other types of functors please contact the author
		4095	(preferably after implementing it).
		4096
		4097	For all this to work, your C++ compiler either has to use the same calling
		4098	conventions as your C compiler (for static member functions), or you have
		4099	to embed libev and compile libev itself as C++.
		4100
		4101	Here is a list of things available in the C<ev> namespace:
		4102
		4103	=over 4
		4104
		4105	=item C<ev::READ>, C<ev::WRITE> etc.
		4106
		4107	These are just enum values with the same values as the C<EV_READ> etc.
		4108	macros from F<ev.h>.
		4109
		4110	=item C<ev::tstamp>, C<ev::now>
		4111
		4112	Aliases to the same types/functions as with the C<ev_> prefix.
		4113
		4114	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		4115
		4116	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		4117	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		4118	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		4119	defined by many implementations.
		4120
		4121	All of those classes have these methods:
		4122
		4123	=over 4
		4124
		4125	=item ev::TYPE::TYPE ()
		4126
		4127	=item ev::TYPE::TYPE (loop)
		4128
		4129	=item ev::TYPE::~TYPE
		4130
		4131	The constructor (optionally) takes an event loop to associate the watcher
		4132	with. If it is omitted, it will use C<EV_DEFAULT>.
		4133
		4134	The constructor calls C<ev_init> for you, which means you have to call the
		4135	C<set> method before starting it.
		4136
		4137	It will not set a callback, however: You have to call the templated C<set>
		4138	method to set a callback before you can start the watcher.
		4139
		4140	(The reason why you have to use a method is a limitation in C++ which does
		4141	not allow explicit template arguments for constructors).
		4142
		4143	The destructor automatically stops the watcher if it is active.
		4144
		4145	=item w->set<class, &class::method> (object *)
		4146
		4147	This method sets the callback method to call. The method has to have a
		4148	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		4149	first argument and the C<revents> as second. The object must be given as
		4150	parameter and is stored in the C<data> member of the watcher.
		4151
		4152	This method synthesizes efficient thunking code to call your method from
		4153	the C callback that libev requires. If your compiler can inline your
		4154	callback (i.e. it is visible to it at the place of the C<set> call and
		4155	your compiler is good :), then the method will be fully inlined into the
		4156	thunking function, making it as fast as a direct C callback.
		4157
		4158	Example: simple class declaration and watcher initialisation
		4159
		4160	struct myclass
		4161	{
		4162	void io_cb (ev::io &w, int revents) { }
		4163	}
		4164
		4165	myclass obj;
		4166	ev::io iow;
		4167	iow.set <myclass, &myclass::io_cb> (&obj);
		4168
		4169	=item w->set (object *)
		4170
		4171	This is a variation of a method callback - leaving out the method to call
		4172	will default the method to C<operator ()>, which makes it possible to use
		4173	functor objects without having to manually specify the C<operator ()> all
		4174	the time. Incidentally, you can then also leave out the template argument
		4175	list.
		4176
		4177	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		4178	int revents)>.
		4179
		4180	See the method-C<set> above for more details.
		4181
		4182	Example: use a functor object as callback.
		4183
		4184	struct myfunctor
		4185	{
		4186	void operator() (ev::io &w, int revents)
		4187	{
		4188	...
		4189	}
		4190	}
		4191
		4192	myfunctor f;
		4193
		4194	ev::io w;
		4195	w.set (&f);
		4196
		4197	=item w->set<function> (void *data = 0)
		4198
		4199	Also sets a callback, but uses a static method or plain function as
		4200	callback. The optional C<data> argument will be stored in the watcher's
		4201	C<data> member and is free for you to use.
		4202
		4203	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		4204
		4205	See the method-C<set> above for more details.
		4206
		4207	Example: Use a plain function as callback.
		4208
		4209	static void io_cb (ev::io &w, int revents) { }
		4210	iow.set <io_cb> ();
		4211
		4212	=item w->set (loop)
		4213
		4214	Associates a different C<struct ev_loop> with this watcher. You can only
		4215	do this when the watcher is inactive (and not pending either).
		4216
		4217	=item w->set ([arguments])
		4218
		4219	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
		4220	with the same arguments. Either this method or a suitable start method
		4221	must be called at least once. Unlike the C counterpart, an active watcher
		4222	gets automatically stopped and restarted when reconfiguring it with this
		4223	method.
		4224
		4225	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4226	clashing with the C<set (loop)> method.
		4227
		4228	=item w->start ()
		4229
		4230	Starts the watcher. Note that there is no C<loop> argument, as the
		4231	constructor already stores the event loop.
		4232
		4233	=item w->start ([arguments])
		4234
		4235	Instead of calling C<set> and C<start> methods separately, it is often
		4236	convenient to wrap them in one call. Uses the same type of arguments as
		4237	the configure C<set> method of the watcher.
		4238
		4239	=item w->stop ()
		4240
		4241	Stops the watcher if it is active. Again, no C<loop> argument.
		4242
		4243	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		4244
		4245	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		4246	C<ev_TYPE_again> function.
		4247
		4248	=item w->sweep () (C<ev::embed> only)
		4249
		4250	Invokes C<ev_embed_sweep>.
		4251
		4252	=item w->update () (C<ev::stat> only)
		4253
		4254	Invokes C<ev_stat_stat>.
		4255
		4256	=back
		4257
		4258	=back
		4259
		4260	Example: Define a class with two I/O and idle watchers, start the I/O
		4261	watchers in the constructor.
		4262
		4263	class myclass
		4264	{
		4265	ev::io io ; void io_cb (ev::io &w, int revents);
		4266	ev::io io2 ; void io2_cb (ev::io &w, int revents);
		4267	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		4268
		4269	myclass (int fd)
		4270	{
		4271	io .set <myclass, &myclass::io_cb > (this);
		4272	io2 .set <myclass, &myclass::io2_cb > (this);
		4273	idle.set <myclass, &myclass::idle_cb> (this);
		4274
		4275	io.set (fd, ev::WRITE); // configure the watcher
		4276	io.start (); // start it whenever convenient
		4277
		4278	io2.start (fd, ev::READ); // set + start in one call
		4279	}
		4280	};
		4281
		4282
		4283	=head1 OTHER LANGUAGE BINDINGS
		4284
		4285	Libev does not offer other language bindings itself, but bindings for a
		4286	number of languages exist in the form of third-party packages. If you know
		4287	any interesting language binding in addition to the ones listed here, drop
		4288	me a note.
		4289
		4290	=over 4
		4291
		4292	=item Perl
		4293
		4294	The EV module implements the full libev API and is actually used to test
		4295	libev. EV is developed together with libev. Apart from the EV core module,
		4296	there are additional modules that implement libev-compatible interfaces
		4297	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		4298	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		4299	and C<EV::Glib>).
		4300
		4301	It can be found and installed via CPAN, its homepage is at
		4302	L<http://software.schmorp.de/pkg/EV>.
		4303
		4304	=item Python
		4305
		4306	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		4307	seems to be quite complete and well-documented.
		4308
		4309	=item Ruby
		4310
		4311	Tony Arcieri has written a ruby extension that offers access to a subset
		4312	of the libev API and adds file handle abstractions, asynchronous DNS and
		4313	more on top of it. It can be found via gem servers. Its homepage is at
		4314	L<http://rev.rubyforge.org/>.
		4315
		4316	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		4317	makes rev work even on mingw.
		4318
		4319	=item Haskell
		4320
		4321	A haskell binding to libev is available at
		4322	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		4323
		4324	=item D
		4325
		4326	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		4327	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
		4328
		4329	=item Ocaml
		4330
		4331	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4332	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4333
		4334	=item Lua
		4335
		4336	Brian Maher has written a partial interface to libev for lua (at the
		4337	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4338	L<http://github.com/brimworks/lua-ev>.
		4339
		4340	=item Javascript
		4341
		4342	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4343
		4344	=item Others
		4345
		4346	There are others, and I stopped counting.
		4347
		4348	=back
		4349
		4350
		4351	=head1 MACRO MAGIC
		4352
		4353	Libev can be compiled with a variety of options, the most fundamental
		4354	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		4355	functions and callbacks have an initial C<struct ev_loop *> argument.
		4356
		4357	To make it easier to write programs that cope with either variant, the
		4358	following macros are defined:
		4359
		4360	=over 4
		4361
		4362	=item C<EV_A>, C<EV_A_>
		4363
		4364	This provides the loop I<argument> for functions, if one is required ("ev
		4365	loop argument"). The C<EV_A> form is used when this is the sole argument,
		4366	C<EV_A_> is used when other arguments are following. Example:
		4367
		4368	ev_unref (EV_A);
		4369	ev_timer_add (EV_A_ watcher);
		4370	ev_run (EV_A_ 0);
		4371
		4372	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		4373	which is often provided by the following macro.
		4374
		4375	=item C<EV_P>, C<EV_P_>
		4376
		4377	This provides the loop I<parameter> for functions, if one is required ("ev
		4378	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		4379	C<EV_P_> is used when other parameters are following. Example:
		4380
		4381	// this is how ev_unref is being declared
		4382	static void ev_unref (EV_P);
		4383
		4384	// this is how you can declare your typical callback
		4385	static void cb (EV_P_ ev_timer *w, int revents)
		4386
		4387	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		4388	suitable for use with C<EV_A>.
		4389
		4390	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		4391
		4392	Similar to the other two macros, this gives you the value of the default
		4393	loop, if multiple loops are supported ("ev loop default"). The default loop
		4394	will be initialised if it isn't already initialised.
		4395
		4396	For non-multiplicity builds, these macros do nothing, so you always have
		4397	to initialise the loop somewhere.
		4398
		4399	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		4400
		4401	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		4402	default loop has been initialised (C<UC> == unchecked). Their behaviour
		4403	is undefined when the default loop has not been initialised by a previous
		4404	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		4405
		4406	It is often prudent to use C<EV_DEFAULT> when initialising the first
		4407	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		4408
		4409	=back
		4410
		4411	Example: Declare and initialise a check watcher, utilising the above
		4412	macros so it will work regardless of whether multiple loops are supported
		4413	or not.
		4414
		4415	static void
		4416	check_cb (EV_P_ ev_timer *w, int revents)
		4417	{
		4418	ev_check_stop (EV_A_ w);
		4419	}
		4420
		4421	ev_check check;
		4422	ev_check_init (&check, check_cb);
		4423	ev_check_start (EV_DEFAULT_ &check);
		4424	ev_run (EV_DEFAULT_ 0);
		4425
		4426	=head1 EMBEDDING
		4427
		4428	Libev can (and often is) directly embedded into host
		4429	applications. Examples of applications that embed it include the Deliantra
		4430	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		4431	and rxvt-unicode.
		4432
		4433	The goal is to enable you to just copy the necessary files into your
		4434	source directory without having to change even a single line in them, so
		4435	you can easily upgrade by simply copying (or having a checked-out copy of
		4436	libev somewhere in your source tree).
		4437
		4438	=head2 FILESETS
		4439
		4440	Depending on what features you need you need to include one or more sets of files
		4441	in your application.
		4442
		4443	=head3 CORE EVENT LOOP
		4444
		4445	To include only the libev core (all the C<ev_*> functions), with manual
		4446	configuration (no autoconf):
		4447
		4448	#define EV_STANDALONE 1
		4449	#include "ev.c"
		4450
		4451	This will automatically include F<ev.h>, too, and should be done in a
		4452	single C source file only to provide the function implementations. To use
		4453	it, do the same for F<ev.h> in all files wishing to use this API (best
		4454	done by writing a wrapper around F<ev.h> that you can include instead and
		4455	where you can put other configuration options):
		4456
		4457	#define EV_STANDALONE 1
		4458	#include "ev.h"
		4459
		4460	Both header files and implementation files can be compiled with a C++
		4461	compiler (at least, that's a stated goal, and breakage will be treated
		4462	as a bug).
		4463
		4464	You need the following files in your source tree, or in a directory
		4465	in your include path (e.g. in libev/ when using -Ilibev):
		4466
		4467	ev.h
		4468	ev.c
		4469	ev_vars.h
		4470	ev_wrap.h
		4471
		4472	ev_win32.c required on win32 platforms only
		4473
		4474	ev_select.c only when select backend is enabled
		4475	ev_poll.c only when poll backend is enabled
		4476	ev_epoll.c only when the epoll backend is enabled
		4477	ev_linuxaio.c only when the linux aio backend is enabled
		4478	ev_kqueue.c only when the kqueue backend is enabled
		4479	ev_port.c only when the solaris port backend is enabled
		4480
		4481	F<ev.c> includes the backend files directly when enabled, so you only need
		4482	to compile this single file.
		4483
		4484	=head3 LIBEVENT COMPATIBILITY API
		4485
		4486	To include the libevent compatibility API, also include:
		4487
		4488	#include "event.c"
		4489
		4490	in the file including F<ev.c>, and:
		4491
		4492	#include "event.h"
		4493
		4494	in the files that want to use the libevent API. This also includes F<ev.h>.
		4495
		4496	You need the following additional files for this:
		4497
		4498	event.h
		4499	event.c
		4500
		4501	=head3 AUTOCONF SUPPORT
		4502
		4503	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		4504	whatever way you want, you can also C<m4_include([libev.m4])> in your
		4505	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		4506	include F<config.h> and configure itself accordingly.
		4507
		4508	For this of course you need the m4 file:
		4509
		4510	libev.m4
		4511
		4512	=head2 PREPROCESSOR SYMBOLS/MACROS
		4513
		4514	Libev can be configured via a variety of preprocessor symbols you have to
		4515	define before including (or compiling) any of its files. The default in
		4516	the absence of autoconf is documented for every option.
		4517
		4518	Symbols marked with "(h)" do not change the ABI, and can have different
		4519	values when compiling libev vs. including F<ev.h>, so it is permissible
		4520	to redefine them before including F<ev.h> without breaking compatibility
		4521	to a compiled library. All other symbols change the ABI, which means all
		4522	users of libev and the libev code itself must be compiled with compatible
		4523	settings.
		4524
		4525	=over 4
		4526
		4527	=item EV_COMPAT3 (h)
		4528
		4529	Backwards compatibility is a major concern for libev. This is why this
		4530	release of libev comes with wrappers for the functions and symbols that
		4531	have been renamed between libev version 3 and 4.
		4532
		4533	You can disable these wrappers (to test compatibility with future
		4534	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4535	sources. This has the additional advantage that you can drop the C<struct>
		4536	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4537	typedef in that case.
		4538
		4539	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4540	and in some even more future version the compatibility code will be
		4541	removed completely.
		4542
		4543	=item EV_STANDALONE (h)
		4544
		4545	Must always be C<1> if you do not use autoconf configuration, which
		4546	keeps libev from including F<config.h>, and it also defines dummy
		4547	implementations for some libevent functions (such as logging, which is not
		4548	supported). It will also not define any of the structs usually found in
		4549	F<event.h> that are not directly supported by the libev core alone.
		4550
		4551	In standalone mode, libev will still try to automatically deduce the
		4552	configuration, but has to be more conservative.
		4553
		4554	=item EV_USE_FLOOR
		4555
		4556	If defined to be C<1>, libev will use the C<floor ()> function for its
		4557	periodic reschedule calculations, otherwise libev will fall back on a
		4558	portable (slower) implementation. If you enable this, you usually have to
		4559	link against libm or something equivalent. Enabling this when the C<floor>
		4560	function is not available will fail, so the safe default is to not enable
		4561	this.
		4562
		4563	=item EV_USE_MONOTONIC
		4564
		4565	If defined to be C<1>, libev will try to detect the availability of the
		4566	monotonic clock option at both compile time and runtime. Otherwise no
		4567	use of the monotonic clock option will be attempted. If you enable this,
		4568	you usually have to link against librt or something similar. Enabling it
		4569	when the functionality isn't available is safe, though, although you have
		4570	to make sure you link against any libraries where the C<clock_gettime>
		4571	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		4572
		4573	=item EV_USE_REALTIME
		4574
		4575	If defined to be C<1>, libev will try to detect the availability of the
		4576	real-time clock option at compile time (and assume its availability
		4577	at runtime if successful). Otherwise no use of the real-time clock
		4578	option will be attempted. This effectively replaces C<gettimeofday>
		4579	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		4580	correctness. See the note about libraries in the description of
		4581	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4582	C<EV_USE_CLOCK_SYSCALL>.
		4583
		4584	=item EV_USE_CLOCK_SYSCALL
		4585
		4586	If defined to be C<1>, libev will try to use a direct syscall instead
		4587	of calling the system-provided C<clock_gettime> function. This option
		4588	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4589	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4590	programs needlessly. Using a direct syscall is slightly slower (in
		4591	theory), because no optimised vdso implementation can be used, but avoids
		4592	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4593	higher, as it simplifies linking (no need for C<-lrt>).
		4594
		4595	=item EV_USE_NANOSLEEP
		4596
		4597	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4598	and will use it for delays. Otherwise it will use C<select ()>.
		4599
		4600	=item EV_USE_EVENTFD
		4601
		4602	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4603	available and will probe for kernel support at runtime. This will improve
		4604	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4605	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4606	2.7 or newer, otherwise disabled.
		4607
		4608	=item EV_USE_SELECT
		4609
		4610	If undefined or defined to be C<1>, libev will compile in support for the
		4611	C<select>(2) backend. No attempt at auto-detection will be done: if no
		4612	other method takes over, select will be it. Otherwise the select backend
		4613	will not be compiled in.
		4614
		4615	=item EV_SELECT_USE_FD_SET
		4616
		4617	If defined to C<1>, then the select backend will use the system C<fd_set>
		4618	structure. This is useful if libev doesn't compile due to a missing
		4619	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		4620	on exotic systems. This usually limits the range of file descriptors to
		4621	some low limit such as 1024 or might have other limitations (winsocket
		4622	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		4623	configures the maximum size of the C<fd_set>.
		4624
		4625	=item EV_SELECT_IS_WINSOCKET
		4626
		4627	When defined to C<1>, the select backend will assume that
		4628	select/socket/connect etc. don't understand file descriptors but
		4629	wants osf handles on win32 (this is the case when the select to
		4630	be used is the winsock select). This means that it will call
		4631	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		4632	it is assumed that all these functions actually work on fds, even
		4633	on win32. Should not be defined on non-win32 platforms.
		4634
		4635	=item EV_FD_TO_WIN32_HANDLE(fd)
		4636
		4637	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4638	file descriptors to socket handles. When not defining this symbol (the
		4639	default), then libev will call C<_get_osfhandle>, which is usually
		4640	correct. In some cases, programs use their own file descriptor management,
		4641	in which case they can provide this function to map fds to socket handles.
		4642
		4643	=item EV_WIN32_HANDLE_TO_FD(handle)
		4644
		4645	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4646	using the standard C<_open_osfhandle> function. For programs implementing
		4647	their own fd to handle mapping, overwriting this function makes it easier
		4648	to do so. This can be done by defining this macro to an appropriate value.
		4649
		4650	=item EV_WIN32_CLOSE_FD(fd)
		4651
		4652	If programs implement their own fd to handle mapping on win32, then this
		4653	macro can be used to override the C<close> function, useful to unregister
		4654	file descriptors again. Note that the replacement function has to close
		4655	the underlying OS handle.
		4656
		4657	=item EV_USE_WSASOCKET
		4658
		4659	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4660	communication socket, which works better in some environments. Otherwise,
		4661	the normal C<socket> function will be used, which works better in other
		4662	environments.
		4663
		4664	=item EV_USE_POLL
		4665
		4666	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		4667	backend. Otherwise it will be enabled on non-win32 platforms. It
		4668	takes precedence over select.
		4669
		4670	=item EV_USE_EPOLL
		4671
		4672	If defined to be C<1>, libev will compile in support for the Linux
		4673	C<epoll>(7) backend. Its availability will be detected at runtime,
		4674	otherwise another method will be used as fallback. This is the preferred
		4675	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4676	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4677
		4678	=item EV_USE_LINUXAIO
		4679
		4680	If defined to be C<1>, libev will compile in support for the Linux
		4681	aio backend. Due to it's currenbt limitations it has to be requested
		4682	explicitly. If undefined, it will be enabled on linux, otherwise
		4683	disabled.
		4684
		4685	=item EV_USE_KQUEUE
		4686
		4687	If defined to be C<1>, libev will compile in support for the BSD style
		4688	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		4689	otherwise another method will be used as fallback. This is the preferred
		4690	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		4691	supports some types of fds correctly (the only platform we found that
		4692	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		4693	not be used unless explicitly requested. The best way to use it is to find
		4694	out whether kqueue supports your type of fd properly and use an embedded
		4695	kqueue loop.
		4696
		4697	=item EV_USE_PORT
		4698
		4699	If defined to be C<1>, libev will compile in support for the Solaris
		4700	10 port style backend. Its availability will be detected at runtime,
		4701	otherwise another method will be used as fallback. This is the preferred
		4702	backend for Solaris 10 systems.
		4703
		4704	=item EV_USE_DEVPOLL
		4705
		4706	Reserved for future expansion, works like the USE symbols above.
		4707
		4708	=item EV_USE_INOTIFY
		4709
		4710	If defined to be C<1>, libev will compile in support for the Linux inotify
		4711	interface to speed up C<ev_stat> watchers. Its actual availability will
		4712	be detected at runtime. If undefined, it will be enabled if the headers
		4713	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4714
		4715	=item EV_NO_SMP
		4716
		4717	If defined to be C<1>, libev will assume that memory is always coherent
		4718	between threads, that is, threads can be used, but threads never run on
		4719	different cpus (or different cpu cores). This reduces dependencies
		4720	and makes libev faster.
		4721
		4722	=item EV_NO_THREADS
		4723
		4724	If defined to be C<1>, libev will assume that it will never be called from
		4725	different threads (that includes signal handlers), which is a stronger
		4726	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4727	libev faster.
		4728
		4729	=item EV_ATOMIC_T
		4730
		4731	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4732	access is atomic with respect to other threads or signal contexts. No
		4733	such type is easily found in the C language, so you can provide your own
		4734	type that you know is safe for your purposes. It is used both for signal
		4735	handler "locking" as well as for signal and thread safety in C<ev_async>
		4736	watchers.
		4737
		4738	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4739	(from F<signal.h>), which is usually good enough on most platforms.
		4740
		4741	=item EV_H (h)
		4742
		4743	The name of the F<ev.h> header file used to include it. The default if
		4744	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4745	used to virtually rename the F<ev.h> header file in case of conflicts.
		4746
		4747	=item EV_CONFIG_H (h)
		4748
		4749	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4750	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4751	C<EV_H>, above.
		4752
		4753	=item EV_EVENT_H (h)
		4754
		4755	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4756	of how the F<event.h> header can be found, the default is C<"event.h">.
		4757
		4758	=item EV_PROTOTYPES (h)
		4759
		4760	If defined to be C<0>, then F<ev.h> will not define any function
		4761	prototypes, but still define all the structs and other symbols. This is
		4762	occasionally useful if you want to provide your own wrapper functions
		4763	around libev functions.
		4764
		4765	=item EV_MULTIPLICITY
		4766
		4767	If undefined or defined to C<1>, then all event-loop-specific functions
		4768	will have the C<struct ev_loop *> as first argument, and you can create
		4769	additional independent event loops. Otherwise there will be no support
		4770	for multiple event loops and there is no first event loop pointer
		4771	argument. Instead, all functions act on the single default loop.
		4772
		4773	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4774	default loop when multiplicity is switched off - you always have to
		4775	initialise the loop manually in this case.
		4776
		4777	=item EV_MINPRI
		4778
		4779	=item EV_MAXPRI
		4780
		4781	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4782	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4783	provide for more priorities by overriding those symbols (usually defined
		4784	to be C<-2> and C<2>, respectively).
		4785
		4786	When doing priority-based operations, libev usually has to linearly search
		4787	all the priorities, so having many of them (hundreds) uses a lot of space
		4788	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4789	fine.
		4790
		4791	If your embedding application does not need any priorities, defining these
		4792	both to C<0> will save some memory and CPU.
		4793
		4794	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4795	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4796	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4797
		4798	If undefined or defined to be C<1> (and the platform supports it), then
		4799	the respective watcher type is supported. If defined to be C<0>, then it
		4800	is not. Disabling watcher types mainly saves code size.
		4801
		4802	=item EV_FEATURES
		4803
		4804	If you need to shave off some kilobytes of code at the expense of some
		4805	speed (but with the full API), you can define this symbol to request
		4806	certain subsets of functionality. The default is to enable all features
		4807	that can be enabled on the platform.
		4808
		4809	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4810	with some broad features you want) and then selectively re-enable
		4811	additional parts you want, for example if you want everything minimal,
		4812	but multiple event loop support, async and child watchers and the poll
		4813	backend, use this:
		4814
		4815	#define EV_FEATURES 0
		4816	#define EV_MULTIPLICITY 1
		4817	#define EV_USE_POLL 1
		4818	#define EV_CHILD_ENABLE 1
		4819	#define EV_ASYNC_ENABLE 1
		4820
		4821	The actual value is a bitset, it can be a combination of the following
		4822	values (by default, all of these are enabled):
		4823
		4824	=over 4
		4825
		4826	=item C<1> - faster/larger code
		4827
		4828	Use larger code to speed up some operations.
		4829
		4830	Currently this is used to override some inlining decisions (enlarging the
		4831	code size by roughly 30% on amd64).
		4832
		4833	When optimising for size, use of compiler flags such as C<-Os> with
		4834	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4835	assertions.
		4836
		4837	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4838	(e.g. gcc with C<-Os>).
		4839
		4840	=item C<2> - faster/larger data structures
		4841
		4842	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4843	hash table sizes and so on. This will usually further increase code size
		4844	and can additionally have an effect on the size of data structures at
		4845	runtime.
		4846
		4847	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4848	(e.g. gcc with C<-Os>).
		4849
		4850	=item C<4> - full API configuration
		4851
		4852	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4853	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4854
		4855	=item C<8> - full API
		4856
		4857	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4858	details on which parts of the API are still available without this
		4859	feature, and do not complain if this subset changes over time.
		4860
		4861	=item C<16> - enable all optional watcher types
		4862
		4863	Enables all optional watcher types. If you want to selectively enable
		4864	only some watcher types other than I/O and timers (e.g. prepare,
		4865	embed, async, child...) you can enable them manually by defining
		4866	C<EV_watchertype_ENABLE> to C<1> instead.
		4867
		4868	=item C<32> - enable all backends
		4869
		4870	This enables all backends - without this feature, you need to enable at
		4871	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4872
		4873	=item C<64> - enable OS-specific "helper" APIs
		4874
		4875	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4876	default.
		4877
		4878	=back
		4879
		4880	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4881	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4882	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4883	watchers, timers and monotonic clock support.
		4884
		4885	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4886	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4887	your program might be left out as well - a binary starting a timer and an
		4888	I/O watcher then might come out at only 5Kb.
		4889
		4890	=item EV_API_STATIC
		4891
		4892	If this symbol is defined (by default it is not), then all identifiers
		4893	will have static linkage. This means that libev will not export any
		4894	identifiers, and you cannot link against libev anymore. This can be useful
		4895	when you embed libev, only want to use libev functions in a single file,
		4896	and do not want its identifiers to be visible.
		4897
		4898	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4899	wants to use libev.
		4900
		4901	This option only works when libev is compiled with a C compiler, as C++
		4902	doesn't support the required declaration syntax.
		4903
		4904	=item EV_AVOID_STDIO
		4905
		4906	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4907	functions (printf, scanf, perror etc.). This will increase the code size
		4908	somewhat, but if your program doesn't otherwise depend on stdio and your
		4909	libc allows it, this avoids linking in the stdio library which is quite
		4910	big.
		4911
		4912	Note that error messages might become less precise when this option is
		4913	enabled.
		4914
		4915	=item EV_NSIG
		4916
		4917	The highest supported signal number, +1 (or, the number of
		4918	signals): Normally, libev tries to deduce the maximum number of signals
		4919	automatically, but sometimes this fails, in which case it can be
		4920	specified. Also, using a lower number than detected (C<32> should be
		4921	good for about any system in existence) can save some memory, as libev
		4922	statically allocates some 12-24 bytes per signal number.
		4923
		4924	=item EV_PID_HASHSIZE
		4925
		4926	C<ev_child> watchers use a small hash table to distribute workload by
		4927	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4928	usually more than enough. If you need to manage thousands of children you
		4929	might want to increase this value (I<must> be a power of two).
		4930
		4931	=item EV_INOTIFY_HASHSIZE
		4932
		4933	C<ev_stat> watchers use a small hash table to distribute workload by
		4934	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4935	disabled), usually more than enough. If you need to manage thousands of
		4936	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4937	power of two).
		4938
		4939	=item EV_USE_4HEAP
		4940
		4941	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4942	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4943	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4944	faster performance with many (thousands) of watchers.
		4945
		4946	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4947	will be C<0>.
		4948
		4949	=item EV_HEAP_CACHE_AT
		4950
		4951	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4952	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4953	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4954	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4955	but avoids random read accesses on heap changes. This improves performance
		4956	noticeably with many (hundreds) of watchers.
		4957
		4958	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4959	will be C<0>.
		4960
		4961	=item EV_VERIFY
		4962
		4963	Controls how much internal verification (see C<ev_verify ()>) will
		4964	be done: If set to C<0>, no internal verification code will be compiled
		4965	in. If set to C<1>, then verification code will be compiled in, but not
		4966	called. If set to C<2>, then the internal verification code will be
		4967	called once per loop, which can slow down libev. If set to C<3>, then the
		4968	verification code will be called very frequently, which will slow down
		4969	libev considerably.
		4970
		4971	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4972	will be C<0>.
		4973
		4974	=item EV_COMMON
		4975
		4976	By default, all watchers have a C<void *data> member. By redefining
		4977	this macro to something else you can include more and other types of
		4978	members. You have to define it each time you include one of the files,
		4979	though, and it must be identical each time.
		4980
		4981	For example, the perl EV module uses something like this:
		4982
		4983	#define EV_COMMON \
		4984	SV *self; /* contains this struct */ \
		4985	SV *cb_sv, *fh /* note no trailing ";" */
		4986
		4987	=item EV_CB_DECLARE (type)
		4988
		4989	=item EV_CB_INVOKE (watcher, revents)
		4990
		4991	=item ev_set_cb (ev, cb)
		4992
		4993	Can be used to change the callback member declaration in each watcher,
		4994	and the way callbacks are invoked and set. Must expand to a struct member
		4995	definition and a statement, respectively. See the F<ev.h> header file for
		4996	their default definitions. One possible use for overriding these is to
		4997	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4998	method calls instead of plain function calls in C++.
		4999
		5000	=back
		5001
		5002	=head2 EXPORTED API SYMBOLS
		5003
		5004	If you need to re-export the API (e.g. via a DLL) and you need a list of
		5005	exported symbols, you can use the provided F<Symbol.*> files which list
		5006	all public symbols, one per line:
		5007
		5008	Symbols.ev for libev proper
		5009	Symbols.event for the libevent emulation
		5010
		5011	This can also be used to rename all public symbols to avoid clashes with
		5012	multiple versions of libev linked together (which is obviously bad in
		5013	itself, but sometimes it is inconvenient to avoid this).
		5014
		5015	A sed command like this will create wrapper C<#define>'s that you need to
		5016	include before including F<ev.h>:
		5017
		5018	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		5019
		5020	This would create a file F<wrap.h> which essentially looks like this:
		5021
		5022	#define ev_backend myprefix_ev_backend
		5023	#define ev_check_start myprefix_ev_check_start
		5024	#define ev_check_stop myprefix_ev_check_stop
		5025	...
		5026
		5027	=head2 EXAMPLES
		5028
		5029	For a real-world example of a program the includes libev
		5030	verbatim, you can have a look at the EV perl module
		5031	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		5032	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		5033	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		5034	will be compiled. It is pretty complex because it provides its own header
		5035	file.
		5036
		5037	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		5038	that everybody includes and which overrides some configure choices:
		5039
		5040	#define EV_FEATURES 8
		5041	#define EV_USE_SELECT 1
		5042	#define EV_PREPARE_ENABLE 1
		5043	#define EV_IDLE_ENABLE 1
		5044	#define EV_SIGNAL_ENABLE 1
		5045	#define EV_CHILD_ENABLE 1
		5046	#define EV_USE_STDEXCEPT 0
		5047	#define EV_CONFIG_H <config.h>
		5048
		5049	#include "ev++.h"
		5050
		5051	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		5052
		5053	#include "ev_cpp.h"
		5054	#include "ev.c"
		5055
		5056	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
		5057
		5058	=head2 THREADS AND COROUTINES
		5059
		5060	=head3 THREADS
		5061
		5062	All libev functions are reentrant and thread-safe unless explicitly
		5063	documented otherwise, but libev implements no locking itself. This means
		5064	that you can use as many loops as you want in parallel, as long as there
		5065	are no concurrent calls into any libev function with the same loop
		5066	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		5067	of course): libev guarantees that different event loops share no data
		5068	structures that need any locking.
		5069
		5070	Or to put it differently: calls with different loop parameters can be done
		5071	concurrently from multiple threads, calls with the same loop parameter
		5072	must be done serially (but can be done from different threads, as long as
		5073	only one thread ever is inside a call at any point in time, e.g. by using
		5074	a mutex per loop).
		5075
		5076	Specifically to support threads (and signal handlers), libev implements
		5077	so-called C<ev_async> watchers, which allow some limited form of
		5078	concurrency on the same event loop, namely waking it up "from the
		5079	outside".
		5080
		5081	If you want to know which design (one loop, locking, or multiple loops
		5082	without or something else still) is best for your problem, then I cannot
		5083	help you, but here is some generic advice:
		5084
		5085	=over 4
		5086
		5087	=item * most applications have a main thread: use the default libev loop
		5088	in that thread, or create a separate thread running only the default loop.
		5089
		5090	This helps integrating other libraries or software modules that use libev
		5091	themselves and don't care/know about threading.
		5092
		5093	=item * one loop per thread is usually a good model.
		5094
		5095	Doing this is almost never wrong, sometimes a better-performance model
		5096	exists, but it is always a good start.
		5097
		5098	=item * other models exist, such as the leader/follower pattern, where one
		5099	loop is handed through multiple threads in a kind of round-robin fashion.
		5100
		5101	Choosing a model is hard - look around, learn, know that usually you can do
		5102	better than you currently do :-)
		5103
		5104	=item * often you need to talk to some other thread which blocks in the
		5105	event loop.
		5106
		5107	C<ev_async> watchers can be used to wake them up from other threads safely
		5108	(or from signal contexts...).
		5109
		5110	An example use would be to communicate signals or other events that only
		5111	work in the default loop by registering the signal watcher with the
		5112	default loop and triggering an C<ev_async> watcher from the default loop
		5113	watcher callback into the event loop interested in the signal.
		5114
		5115	=back
		5116
		5117	See also L</THREAD LOCKING EXAMPLE>.
		5118
		5119	=head3 COROUTINES
		5120
		5121	Libev is very accommodating to coroutines ("cooperative threads"):
		5122	libev fully supports nesting calls to its functions from different
		5123	coroutines (e.g. you can call C<ev_run> on the same loop from two
		5124	different coroutines, and switch freely between both coroutines running
		5125	the loop, as long as you don't confuse yourself). The only exception is
		5126	that you must not do this from C<ev_periodic> reschedule callbacks.
		5127
		5128	Care has been taken to ensure that libev does not keep local state inside
		5129	C<ev_run>, and other calls do not usually allow for coroutine switches as
		5130	they do not call any callbacks.
		5131
		5132	=head2 COMPILER WARNINGS
		5133
		5134	Depending on your compiler and compiler settings, you might get no or a
		5135	lot of warnings when compiling libev code. Some people are apparently
		5136	scared by this.
		5137
		5138	However, these are unavoidable for many reasons. For one, each compiler
		5139	has different warnings, and each user has different tastes regarding
		5140	warning options. "Warn-free" code therefore cannot be a goal except when
		5141	targeting a specific compiler and compiler-version.
		5142
		5143	Another reason is that some compiler warnings require elaborate
		5144	workarounds, or other changes to the code that make it less clear and less
		5145	maintainable.
		5146
		5147	And of course, some compiler warnings are just plain stupid, or simply
		5148	wrong (because they don't actually warn about the condition their message
		5149	seems to warn about). For example, certain older gcc versions had some
		5150	warnings that resulted in an extreme number of false positives. These have
		5151	been fixed, but some people still insist on making code warn-free with
		5152	such buggy versions.
		5153
		5154	While libev is written to generate as few warnings as possible,
		5155	"warn-free" code is not a goal, and it is recommended not to build libev
		5156	with any compiler warnings enabled unless you are prepared to cope with
		5157	them (e.g. by ignoring them). Remember that warnings are just that:
		5158	warnings, not errors, or proof of bugs.
		5159
		5160
		5161	=head2 VALGRIND
		5162
		5163	Valgrind has a special section here because it is a popular tool that is
		5164	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		5165
		5166	If you think you found a bug (memory leak, uninitialised data access etc.)
		5167	in libev, then check twice: If valgrind reports something like:
		5168
		5169	==2274== definitely lost: 0 bytes in 0 blocks.
		5170	==2274== possibly lost: 0 bytes in 0 blocks.
		5171	==2274== still reachable: 256 bytes in 1 blocks.
		5172
		5173	Then there is no memory leak, just as memory accounted to global variables
		5174	is not a memleak - the memory is still being referenced, and didn't leak.
		5175
		5176	Similarly, under some circumstances, valgrind might report kernel bugs
		5177	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		5178	although an acceptable workaround has been found here), or it might be
		5179	confused.
		5180
		5181	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		5182	make it into some kind of religion.
		5183
		5184	If you are unsure about something, feel free to contact the mailing list
		5185	with the full valgrind report and an explanation on why you think this
		5186	is a bug in libev (best check the archives, too :). However, don't be
		5187	annoyed when you get a brisk "this is no bug" answer and take the chance
		5188	of learning how to interpret valgrind properly.
		5189
		5190	If you need, for some reason, empty reports from valgrind for your project
		5191	I suggest using suppression lists.
		5192
		5193
		5194	=head1 PORTABILITY NOTES
		5195
		5196	=head2 GNU/LINUX 32 BIT LIMITATIONS
		5197
		5198	GNU/Linux is the only common platform that supports 64 bit file/large file
		5199	interfaces but I<disables> them by default.
		5200
		5201	That means that libev compiled in the default environment doesn't support
		5202	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		5203
		5204	Unfortunately, many programs try to work around this GNU/Linux issue
		5205	by enabling the large file API, which makes them incompatible with the
		5206	standard libev compiled for their system.
		5207
		5208	Likewise, libev cannot enable the large file API itself as this would
		5209	suddenly make it incompatible to the default compile time environment,
		5210	i.e. all programs not using special compile switches.
		5211
		5212	=head2 OS/X AND DARWIN BUGS
		5213
		5214	The whole thing is a bug if you ask me - basically any system interface
		5215	you touch is broken, whether it is locales, poll, kqueue or even the
		5216	OpenGL drivers.
		5217
		5218	=head3 C<kqueue> is buggy
		5219
		5220	The kqueue syscall is broken in all known versions - most versions support
		5221	only sockets, many support pipes.
		5222
		5223	Libev tries to work around this by not using C<kqueue> by default on this
		5224	rotten platform, but of course you can still ask for it when creating a
		5225	loop - embedding a socket-only kqueue loop into a select-based one is
		5226	probably going to work well.
		5227
		5228	=head3 C<poll> is buggy
		5229
		5230	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		5231	implementation by something calling C<kqueue> internally around the 10.5.6
		5232	release, so now C<kqueue> I<and> C<poll> are broken.
		5233
		5234	Libev tries to work around this by not using C<poll> by default on
		5235	this rotten platform, but of course you can still ask for it when creating
		5236	a loop.
		5237
		5238	=head3 C<select> is buggy
		5239
		5240	All that's left is C<select>, and of course Apple found a way to fuck this
		5241	one up as well: On OS/X, C<select> actively limits the number of file
		5242	descriptors you can pass in to 1024 - your program suddenly crashes when
		5243	you use more.
		5244
		5245	There is an undocumented "workaround" for this - defining
		5246	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		5247	work on OS/X.
		5248
		5249	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		5250
		5251	=head3 C<errno> reentrancy
		5252
		5253	The default compile environment on Solaris is unfortunately so
		5254	thread-unsafe that you can't even use components/libraries compiled
		5255	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		5256	defined by default. A valid, if stupid, implementation choice.
		5257
		5258	If you want to use libev in threaded environments you have to make sure
		5259	it's compiled with C<_REENTRANT> defined.
		5260
		5261	=head3 Event port backend
		5262
		5263	The scalable event interface for Solaris is called "event
		5264	ports". Unfortunately, this mechanism is very buggy in all major
		5265	releases. If you run into high CPU usage, your program freezes or you get
		5266	a large number of spurious wakeups, make sure you have all the relevant
		5267	and latest kernel patches applied. No, I don't know which ones, but there
		5268	are multiple ones to apply, and afterwards, event ports actually work
		5269	great.
		5270
		5271	If you can't get it to work, you can try running the program by setting
		5272	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		5273	C<select> backends.
		5274
		5275	=head2 AIX POLL BUG
		5276
		5277	AIX unfortunately has a broken C<poll.h> header. Libev works around
		5278	this by trying to avoid the poll backend altogether (i.e. it's not even
		5279	compiled in), which normally isn't a big problem as C<select> works fine
		5280	with large bitsets on AIX, and AIX is dead anyway.
		5281
		5282	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		5283
		5284	=head3 General issues
		5285
		5286	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		5287	requires, and its I/O model is fundamentally incompatible with the POSIX
		5288	model. Libev still offers limited functionality on this platform in
		5289	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		5290	descriptors. This only applies when using Win32 natively, not when using
		5291	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		5292	as every compiler comes with a slightly differently broken/incompatible
		5293	environment.
		5294
		5295	Lifting these limitations would basically require the full
		5296	re-implementation of the I/O system. If you are into this kind of thing,
		5297	then note that glib does exactly that for you in a very portable way (note
		5298	also that glib is the slowest event library known to man).
		5299
		5300	There is no supported compilation method available on windows except
		5301	embedding it into other applications.
		5302
		5303	Sensible signal handling is officially unsupported by Microsoft - libev
		5304	tries its best, but under most conditions, signals will simply not work.
		5305
		5306	Not a libev limitation but worth mentioning: windows apparently doesn't
		5307	accept large writes: instead of resulting in a partial write, windows will
		5308	either accept everything or return C<ENOBUFS> if the buffer is too large,
		5309	so make sure you only write small amounts into your sockets (less than a
		5310	megabyte seems safe, but this apparently depends on the amount of memory
		5311	available).
		5312
		5313	Due to the many, low, and arbitrary limits on the win32 platform and
		5314	the abysmal performance of winsockets, using a large number of sockets
		5315	is not recommended (and not reasonable). If your program needs to use
		5316	more than a hundred or so sockets, then likely it needs to use a totally
		5317	different implementation for windows, as libev offers the POSIX readiness
		5318	notification model, which cannot be implemented efficiently on windows
		5319	(due to Microsoft monopoly games).
		5320
		5321	A typical way to use libev under windows is to embed it (see the embedding
		5322	section for details) and use the following F<evwrap.h> header file instead
		5323	of F<ev.h>:
		5324
		5325	#define EV_STANDALONE /* keeps ev from requiring config.h */
		5326	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		5327
		5328	#include "ev.h"
		5329
		5330	And compile the following F<evwrap.c> file into your project (make sure
		5331	you do I<not> compile the F<ev.c> or any other embedded source files!):
		5332
		5333	#include "evwrap.h"
		5334	#include "ev.c"
		5335
		5336	=head3 The winsocket C<select> function
		5337
		5338	The winsocket C<select> function doesn't follow POSIX in that it
		5339	requires socket I<handles> and not socket I<file descriptors> (it is
		5340	also extremely buggy). This makes select very inefficient, and also
		5341	requires a mapping from file descriptors to socket handles (the Microsoft
		5342	C runtime provides the function C<_open_osfhandle> for this). See the
		5343	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		5344	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		5345
		5346	The configuration for a "naked" win32 using the Microsoft runtime
		5347	libraries and raw winsocket select is:
		5348
		5349	#define EV_USE_SELECT 1
		5350	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		5351
		5352	Note that winsockets handling of fd sets is O(n), so you can easily get a
		5353	complexity in the O(n²) range when using win32.
		5354
		5355	=head3 Limited number of file descriptors
		5356
		5357	Windows has numerous arbitrary (and low) limits on things.
		5358
		5359	Early versions of winsocket's select only supported waiting for a maximum
		5360	of C<64> handles (probably owning to the fact that all windows kernels
		5361	can only wait for C<64> things at the same time internally; Microsoft
		5362	recommends spawning a chain of threads and wait for 63 handles and the
		5363	previous thread in each. Sounds great!).
		5364
		5365	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		5366	to some high number (e.g. C<2048>) before compiling the winsocket select
		5367	call (which might be in libev or elsewhere, for example, perl and many
		5368	other interpreters do their own select emulation on windows).
		5369
		5370	Another limit is the number of file descriptors in the Microsoft runtime
		5371	libraries, which by default is C<64> (there must be a hidden I<64>
		5372	fetish or something like this inside Microsoft). You can increase this
		5373	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		5374	(another arbitrary limit), but is broken in many versions of the Microsoft
		5375	runtime libraries. This might get you to about C<512> or C<2048> sockets
		5376	(depending on windows version and/or the phase of the moon). To get more,
		5377	you need to wrap all I/O functions and provide your own fd management, but
		5378	the cost of calling select (O(n²)) will likely make this unworkable.
		5379
		5380	=head2 PORTABILITY REQUIREMENTS
		5381
		5382	In addition to a working ISO-C implementation and of course the
		5383	backend-specific APIs, libev relies on a few additional extensions:
		5384
		5385	=over 4
		5386
		5387	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		5388	calling conventions regardless of C<ev_watcher_type *>.
		5389
		5390	Libev assumes not only that all watcher pointers have the same internal
		5391	structure (guaranteed by POSIX but not by ISO C for example), but it also
		5392	assumes that the same (machine) code can be used to call any watcher
		5393	callback: The watcher callbacks have different type signatures, but libev
		5394	calls them using an C<ev_watcher *> internally.
		5395
		5396	=item null pointers and integer zero are represented by 0 bytes
		5397
		5398	Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
		5399	relies on this setting pointers and integers to null.
		5400
		5401	=item pointer accesses must be thread-atomic
		5402
		5403	Accessing a pointer value must be atomic, it must both be readable and
		5404	writable in one piece - this is the case on all current architectures.
		5405
		5406	=item C<sig_atomic_t volatile> must be thread-atomic as well
		5407
		5408	The type C<sig_atomic_t volatile> (or whatever is defined as
		5409	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		5410	threads. This is not part of the specification for C<sig_atomic_t>, but is
		5411	believed to be sufficiently portable.
		5412
		5413	=item C<sigprocmask> must work in a threaded environment
		5414
		5415	Libev uses C<sigprocmask> to temporarily block signals. This is not
		5416	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		5417	pthread implementations will either allow C<sigprocmask> in the "main
		5418	thread" or will block signals process-wide, both behaviours would
		5419	be compatible with libev. Interaction between C<sigprocmask> and
		5420	C<pthread_sigmask> could complicate things, however.
		5421
		5422	The most portable way to handle signals is to block signals in all threads
		5423	except the initial one, and run the signal handling loop in the initial
		5424	thread as well.
		5425
		5426	=item C<long> must be large enough for common memory allocation sizes
		5427
		5428	To improve portability and simplify its API, libev uses C<long> internally
		5429	instead of C<size_t> when allocating its data structures. On non-POSIX
		5430	systems (Microsoft...) this might be unexpectedly low, but is still at
		5431	least 31 bits everywhere, which is enough for hundreds of millions of
		5432	watchers.
		5433
		5434	=item C<double> must hold a time value in seconds with enough accuracy
		5435
		5436	The type C<double> is used to represent timestamps. It is required to
		5437	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		5438	good enough for at least into the year 4000 with millisecond accuracy
		5439	(the design goal for libev). This requirement is overfulfilled by
		5440	implementations using IEEE 754, which is basically all existing ones.
		5441
		5442	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5443	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5444	is either obsolete or somebody patched it to use C<long double> or
		5445	something like that, just kidding).
		5446
		5447	=back
		5448
		5449	If you know of other additional requirements drop me a note.
		5450
		5451
		5452	=head1 ALGORITHMIC COMPLEXITIES
		5453
		5454	In this section the complexities of (many of) the algorithms used inside
		5455	libev will be documented. For complexity discussions about backends see
		5456	the documentation for C<ev_default_init>.
		5457
		5458	All of the following are about amortised time: If an array needs to be
		5459	extended, libev needs to realloc and move the whole array, but this
		5460	happens asymptotically rarer with higher number of elements, so O(1) might
		5461	mean that libev does a lengthy realloc operation in rare cases, but on
		5462	average it is much faster and asymptotically approaches constant time.
		5463
		5464	=over 4
		5465
		5466	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		5467
		5468	This means that, when you have a watcher that triggers in one hour and
		5469	there are 100 watchers that would trigger before that, then inserting will
		5470	have to skip roughly seven (C<ld 100>) of these watchers.
		5471
		5472	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		5473
		5474	That means that changing a timer costs less than removing/adding them,
		5475	as only the relative motion in the event queue has to be paid for.
		5476
		5477	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		5478
		5479	These just add the watcher into an array or at the head of a list.
		5480
		5481	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		5482
		5483	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		5484
		5485	These watchers are stored in lists, so they need to be walked to find the
		5486	correct watcher to remove. The lists are usually short (you don't usually
		5487	have many watchers waiting for the same fd or signal: one is typical, two
		5488	is rare).
		5489
		5490	=item Finding the next timer in each loop iteration: O(1)
		5491
		5492	By virtue of using a binary or 4-heap, the next timer is always found at a
		5493	fixed position in the storage array.
		5494
		5495	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5496
		5497	A change means an I/O watcher gets started or stopped, which requires
		5498	libev to recalculate its status (and possibly tell the kernel, depending
		5499	on backend and whether C<ev_io_set> was used).
		5500
		5501	=item Activating one watcher (putting it into the pending state): O(1)
		5502
		5503	=item Priority handling: O(number_of_priorities)
		5504
		5505	Priorities are implemented by allocating some space for each
		5506	priority. When doing priority-based operations, libev usually has to
		5507	linearly search all the priorities, but starting/stopping and activating
		5508	watchers becomes O(1) with respect to priority handling.
		5509
		5510	=item Sending an ev_async: O(1)
		5511
		5512	=item Processing ev_async_send: O(number_of_async_watchers)
		5513
		5514	=item Processing signals: O(max_signal_number)
		5515
		5516	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5517	calls in the current loop iteration and the loop is currently
		5518	blocked. Checking for async and signal events involves iterating over all
		5519	running async watchers or all signal numbers.
		5520
		5521	=back
		5522
		5523
		5524	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5525
		5526	The major version 4 introduced some incompatible changes to the API.
		5527
		5528	At the moment, the C<ev.h> header file provides compatibility definitions
		5529	for all changes, so most programs should still compile. The compatibility
		5530	layer might be removed in later versions of libev, so better update to the
		5531	new API early than late.
		5532
		5533	=over 4
		5534
		5535	=item C<EV_COMPAT3> backwards compatibility mechanism
		5536
		5537	The backward compatibility mechanism can be controlled by
		5538	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
		5539	section.
		5540
		5541	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5542
		5543	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5544
		5545	ev_loop_destroy (EV_DEFAULT_UC);
		5546	ev_loop_fork (EV_DEFAULT);
		5547
		5548	=item function/symbol renames
		5549
		5550	A number of functions and symbols have been renamed:
		5551
		5552	ev_loop => ev_run
		5553	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5554	EVLOOP_ONESHOT => EVRUN_ONCE
		5555
		5556	ev_unloop => ev_break
		5557	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5558	EVUNLOOP_ONE => EVBREAK_ONE
		5559	EVUNLOOP_ALL => EVBREAK_ALL
		5560
		5561	EV_TIMEOUT => EV_TIMER
		5562
		5563	ev_loop_count => ev_iteration
		5564	ev_loop_depth => ev_depth
		5565	ev_loop_verify => ev_verify
		5566
		5567	Most functions working on C<struct ev_loop> objects don't have an
		5568	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5569	associated constants have been renamed to not collide with the C<struct
		5570	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5571	as all other watcher types. Note that C<ev_loop_fork> is still called
		5572	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5573	typedef.
		5574
		5575	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5576
		5577	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5578	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5579	and work, but the library code will of course be larger.
		5580
		5581	=back
		5582
		5583
		5584	=head1 GLOSSARY
		5585
		5586	=over 4
		5587
		5588	=item active
		5589
		5590	A watcher is active as long as it has been started and not yet stopped.
		5591	See L</WATCHER STATES> for details.
		5592
		5593	=item application
		5594
		5595	In this document, an application is whatever is using libev.
		5596
		5597	=item backend
		5598
		5599	The part of the code dealing with the operating system interfaces.
		5600
		5601	=item callback
		5602
		5603	The address of a function that is called when some event has been
		5604	detected. Callbacks are being passed the event loop, the watcher that
		5605	received the event, and the actual event bitset.
		5606
		5607	=item callback/watcher invocation
		5608
		5609	The act of calling the callback associated with a watcher.
		5610
		5611	=item event
		5612
		5613	A change of state of some external event, such as data now being available
		5614	for reading on a file descriptor, time having passed or simply not having
		5615	any other events happening anymore.
		5616
		5617	In libev, events are represented as single bits (such as C<EV_READ> or
		5618	C<EV_TIMER>).
		5619
		5620	=item event library
		5621
		5622	A software package implementing an event model and loop.
		5623
		5624	=item event loop
		5625
		5626	An entity that handles and processes external events and converts them
		5627	into callback invocations.
		5628
		5629	=item event model
		5630
		5631	The model used to describe how an event loop handles and processes
		5632	watchers and events.
		5633
		5634	=item pending
		5635
		5636	A watcher is pending as soon as the corresponding event has been
		5637	detected. See L</WATCHER STATES> for details.
		5638
		5639	=item real time
		5640
		5641	The physical time that is observed. It is apparently strictly monotonic :)
		5642
		5643	=item wall-clock time
		5644
		5645	The time and date as shown on clocks. Unlike real time, it can actually
		5646	be wrong and jump forwards and backwards, e.g. when you adjust your
		5647	clock.
		5648
		5649	=item watcher
		5650
		5651	A data structure that describes interest in certain events. Watchers need
		5652	to be started (attached to an event loop) before they can receive events.
		5653
		5654	=back
		5655
742	=head1 AUTHOR	5656	=head1 AUTHOR
743		5657
744	Marc Lehmann <libev@schmorp.de>.	5658	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5659	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
745		5660

Diff Legend

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