ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.16 by root, Mon Nov 12 08:47:14 2007 UTC vs.
Revision 1.353 by root, Mon Jan 10 14:36:44 2011 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
9	=head1 DESCRIPTION	9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
		18	ev_io stdin_watcher;
		19	ev_timer timeout_watcher;
		20
		21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
		23	static void
		24	stdin_cb (EV_P_ ev_io *w, int revents)
		25	{
		26	puts ("stdin ready");
		27	// for one-shot events, one must manually stop the watcher
		28	// with its corresponding stop function.
		29	ev_io_stop (EV_A_ w);
		30
		31	// this causes all nested ev_run's to stop iterating
		32	ev_break (EV_A_ EVBREAK_ALL);
		33	}
		34
		35	// another callback, this time for a time-out
		36	static void
		37	timeout_cb (EV_P_ ev_timer *w, int revents)
		38	{
		39	puts ("timeout");
		40	// this causes the innermost ev_run to stop iterating
		41	ev_break (EV_A_ EVBREAK_ONE);
		42	}
		43
		44	int
		45	main (void)
		46	{
		47	// use the default event loop unless you have special needs
		48	struct ev_loop *loop = EV_DEFAULT;
		49
		50	// initialise an io watcher, then start it
		51	// this one will watch for stdin to become readable
		52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		53	ev_io_start (loop, &stdin_watcher);
		54
		55	// initialise a timer watcher, then start it
		56	// simple non-repeating 5.5 second timeout
		57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		58	ev_timer_start (loop, &timeout_watcher);
		59
		60	// now wait for events to arrive
		61	ev_run (loop, 0);
		62
		63	// unloop was called, so exit
		64	return 0;
		65	}
		66
		67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
		70
		71	The newest version of this document is also available as an html-formatted
		72	web page you might find easier to navigate when reading it for the first
		73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familiarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 WHAT TO READ WHEN IN A HURRY
		84
		85	This manual tries to be very detailed, but unfortunately, this also makes
		86	it very long. If you just want to know the basics of libev, I suggest
		87	reading L<ANATOMY OF A WATCHER>, then the L<EXAMPLE PROGRAM> above and
		88	look up the missing functions in L<GLOBAL FUNCTIONS> and the C<ev_io> and
		89	C<ev_timer> sections in L<WATCHER TYPES>.
		90
		91	=head1 ABOUT LIBEV
10		92
11	Libev is an event loop: you register interest in certain events (such as a	93	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	94	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	95	these event sources and provide your program with events.
14		96
15	To do this, it must take more or less complete control over your process	97	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	98	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	99	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	101	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	102	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	103	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	104	watcher.
23		105
24	=head1 FEATURES	106	=head2 FEATURES
25		107
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	108	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	109	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	110	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		111	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		112	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		113	timers (C<ev_timer>), absolute timers with customised rescheduling
		114	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	115	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	116	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		117	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		118	limited support for fork events (C<ev_fork>).
		119
		120	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	121	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	122	for example).
33		123
34	=head1 CONVENTIONS	124	=head2 CONVENTIONS
35		125
36	Libev is very configurable. In this manual the default configuration	126	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	127	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	128	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	129	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	130	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 *>)	131	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	132	this argument.
43		133
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	134	=head2 TIME REPRESENTATION
45		135
46	Libev represents time as a single floating point number, representing the	136	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	137	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	138	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	139	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	140	too. It usually aliases to the C<double> type in C. When you need to do
		141	any calculations on it, you should treat it as some floating point value.
		142
		143	Unlike the name component C<stamp> might indicate, it is also used for
		144	time differences (e.g. delays) throughout libev.
		145
		146	=head1 ERROR HANDLING
		147
		148	Libev knows three classes of errors: operating system errors, usage errors
		149	and internal errors (bugs).
		150
		151	When libev catches an operating system error it cannot handle (for example
		152	a system call indicating a condition libev cannot fix), it calls the callback
		153	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		154	abort. The default is to print a diagnostic message and to call C<abort
		155	()>.
		156
		157	When libev detects a usage error such as a negative timer interval, then
		158	it will print a diagnostic message and abort (via the C<assert> mechanism,
		159	so C<NDEBUG> will disable this checking): these are programming errors in
		160	the libev caller and need to be fixed there.
		161
		162	Libev also has a few internal error-checking C<assert>ions, and also has
		163	extensive consistency checking code. These do not trigger under normal
		164	circumstances, as they indicate either a bug in libev or worse.
		165
		166
		167	=head1 GLOBAL FUNCTIONS
		168
		169	These functions can be called anytime, even before initialising the
		170	library in any way.
51		171
52	=over 4	172	=over 4
53		173
54	=item ev_tstamp ev_time ()	174	=item ev_tstamp ev_time ()
55		175
56	Returns the current time as libev would use it.	176	Returns the current time as libev would use it. Please note that the
		177	C<ev_now> function is usually faster and also often returns the timestamp
		178	you actually want to know. Also interesting is the combination of
		179	C<ev_update_now> and C<ev_now>.
		180
		181	=item ev_sleep (ev_tstamp interval)
		182
		183	Sleep for the given interval: The current thread will be blocked until
		184	either it is interrupted or the given time interval has passed. Basically
		185	this is a sub-second-resolution C<sleep ()>.
57		186
58	=item int ev_version_major ()	187	=item int ev_version_major ()
59		188
60	=item int ev_version_minor ()	189	=item int ev_version_minor ()
61		190
62	You can find out the major and minor version numbers of the library	191	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	192	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	193	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	194	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
66	version of the library your program was compiled against.	195	version of the library your program was compiled against.
67		196
		197	These version numbers refer to the ABI version of the library, not the
		198	release version.
		199
68	Usually, it's a good idea to terminate if the major versions mismatch,	200	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	201	as this indicates an incompatible change. Minor versions are usually
70	compatible to older versions, so a larger minor version alone is usually	202	compatible to older versions, so a larger minor version alone is usually
71	not a problem.	203	not a problem.
72		204
		205	Example: Make sure we haven't accidentally been linked against the wrong
		206	version (note, however, that this will not detect other ABI mismatches,
		207	such as LFS or reentrancy).
		208
		209	assert (("libev version mismatch",
		210	ev_version_major () == EV_VERSION_MAJOR
		211	&& ev_version_minor () >= EV_VERSION_MINOR));
		212
		213	=item unsigned int ev_supported_backends ()
		214
		215	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		216	value) compiled into this binary of libev (independent of their
		217	availability on the system you are running on). See C<ev_default_loop> for
		218	a description of the set values.
		219
		220	Example: make sure we have the epoll method, because yeah this is cool and
		221	a must have and can we have a torrent of it please!!!11
		222
		223	assert (("sorry, no epoll, no sex",
		224	ev_supported_backends () & EVBACKEND_EPOLL));
		225
		226	=item unsigned int ev_recommended_backends ()
		227
		228	Return the set of all backends compiled into this binary of libev and
		229	also recommended for this platform, meaning it will work for most file
		230	descriptor types. This set is often smaller than the one returned by
		231	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		232	and will not be auto-detected unless you explicitly request it (assuming
		233	you know what you are doing). This is the set of backends that libev will
		234	probe for if you specify no backends explicitly.
		235
		236	=item unsigned int ev_embeddable_backends ()
		237
		238	Returns the set of backends that are embeddable in other event loops. This
		239	value is platform-specific but can include backends not available on the
		240	current system. To find which embeddable backends might be supported on
		241	the current system, you would need to look at C<ev_embeddable_backends ()
		242	& ev_supported_backends ()>, likewise for recommended ones.
		243
		244	See the description of C<ev_embed> watchers for more info.
		245
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	246	=item ev_set_allocator (void *(*cb)(void *ptr, long size))
74		247
75	Sets the allocation function to use (the prototype is similar to the	248	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	249	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	250	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	251	when memory needs to be allocated (C<size != 0>), the library might abort
79	destructive action. The default is your system realloc function.	252	or take some potentially destructive action.
		253
		254	Since some systems (at least OpenBSD and Darwin) fail to implement
		255	correct C<realloc> semantics, libev will use a wrapper around the system
		256	C<realloc> and C<free> functions by default.
80		257
81	You could override this function in high-availability programs to, say,	258	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,	259	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.	260	or even to sleep a while and retry until some memory is available.
84		261
		262	Example: Replace the libev allocator with one that waits a bit and then
		263	retries (example requires a standards-compliant C<realloc>).
		264
		265	static void *
		266	persistent_realloc (void *ptr, size_t size)
		267	{
		268	for (;;)
		269	{
		270	void *newptr = realloc (ptr, size);
		271
		272	if (newptr)
		273	return newptr;
		274
		275	sleep (60);
		276	}
		277	}
		278
		279	...
		280	ev_set_allocator (persistent_realloc);
		281
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	282	=item ev_set_syserr_cb (void (*cb)(const char *msg))
86		283
87	Set the callback function to call on a retryable syscall error (such	284	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	285	as failed select, poll, epoll_wait). The message is a printable string
89	indicating the system call or subsystem causing the problem. If this	286	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	287	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	288	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	289	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	290	(such as abort).
94		291
		292	Example: This is basically the same thing that libev does internally, too.
		293
		294	static void
		295	fatal_error (const char *msg)
		296	{
		297	perror (msg);
		298	abort ();
		299	}
		300
		301	...
		302	ev_set_syserr_cb (fatal_error);
		303
		304	=item ev_feed_signal (int signum)
		305
		306	This function can be used to "simulate" a signal receive. It is completely
		307	safe to call this function at any time, from any context, including signal
		308	handlers or random threads.
		309
		310	Its main use is to customise signal handling in your process, especially
		311	in the presence of threads. For example, you could block signals
		312	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		313	creating any loops), and in one thread, use C<sigwait> or any other
		314	mechanism to wait for signals, then "deliver" them to libev by calling
		315	C<ev_feed_signal>.
		316
95	=back	317	=back
96		318
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	319	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
98		320
99	An event loop is described by a C<struct ev_loop *>. The library knows two	321	An event loop is described by a C<struct ev_loop *> (the C<struct> is
100	types of such loops, the I<default> loop, which supports signals and child	322	I<not> optional in this case unless libev 3 compatibility is disabled, as
101	events, and dynamically created loops which do not.	323	libev 3 had an C<ev_loop> function colliding with the struct name).
102		324
103	If you use threads, a common model is to run the default event loop	325	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	326	supports child process events, and dynamically created event loops which
105	create, you also create another event loop. Libev itself does no locking	327	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 it's hideous and inefficient).
109		328
110	=over 4	329	=over 4
111		330
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	331	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		332
114	This will initialise the default event loop if it hasn't been initialised	333	This returns the "default" event loop object, which is what you should
115	yet and return it. If the default loop could not be initialised, returns	334	normally use when you just need "the event loop". Event loop objects and
116	false. If it already was initialised it simply returns it (and ignores the	335	the C<flags> parameter are described in more detail in the entry for
117	flags).	336	C<ev_loop_new>.
		337
		338	If the default loop is already initialised then this function simply
		339	returns it (and ignores the flags. If that is troubling you, check
		340	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		341	flags, which should almost always be C<0>, unless the caller is also the
		342	one calling C<ev_run> or otherwise qualifies as "the main program".
118		343
119	If you don't know what event loop to use, use the one returned from this	344	If you don't know what event loop to use, use the one returned from this
120	function.	345	function (or via the C<EV_DEFAULT> macro).
		346
		347	Note that this function is I<not> thread-safe, so if you want to use it
		348	from multiple threads, you have to employ some kind of mutex (note also
		349	that this case is unlikely, as loops cannot be shared easily between
		350	threads anyway).
		351
		352	The default loop is the only loop that can handle C<ev_child> watchers,
		353	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		354	a problem for your application you can either create a dynamic loop with
		355	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		356	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		357
		358	Example: This is the most typical usage.
		359
		360	if (!ev_default_loop (0))
		361	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		362
		363	Example: Restrict libev to the select and poll backends, and do not allow
		364	environment settings to be taken into account:
		365
		366	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		367
		368	=item struct ev_loop *ev_loop_new (unsigned int flags)
		369
		370	This will create and initialise a new event loop object. If the loop
		371	could not be initialised, returns false.
		372
		373	This function is thread-safe, and one common way to use libev with
		374	threads is indeed to create one loop per thread, and using the default
		375	loop in the "main" or "initial" thread.
121		376
122	The flags argument can be used to specify special behaviour or specific	377	The flags argument can be used to specify special behaviour or specific
123	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	378	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
124		379
125	It supports the following flags:	380	The following flags are supported:
126		381
127	=over 4	382	=over 4
128		383
129	=item C<EVFLAG_AUTO>	384	=item C<EVFLAG_AUTO>
130		385
131	The default flags value. Use this if you have no clue (it's the right	386	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	387	thing, believe me).
133		388
134	=item C<EVFLAG_NOENV>	389	=item C<EVFLAG_NOENV>
135		390
136	If this flag bit is ored into the flag value (or the program runs setuid	391	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	392	or setgid) then libev will I<not> look at the environment variable
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	393	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	394	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	395	useful to try out specific backends to test their performance, or to work
141	around bugs.	396	around bugs.
142		397
		398	=item C<EVFLAG_FORKCHECK>
		399
		400	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		401	make libev check for a fork in each iteration by enabling this flag.
		402
		403	This works by calling C<getpid ()> on every iteration of the loop,
		404	and thus this might slow down your event loop if you do a lot of loop
		405	iterations and little real work, but is usually not noticeable (on my
		406	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		407	without a system call and thus I<very> fast, but my GNU/Linux system also has
		408	C<pthread_atfork> which is even faster).
		409
		410	The big advantage of this flag is that you can forget about fork (and
		411	forget about forgetting to tell libev about forking) when you use this
		412	flag.
		413
		414	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		415	environment variable.
		416
		417	=item C<EVFLAG_NOINOTIFY>
		418
		419	When this flag is specified, then libev will not attempt to use the
		420	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		421	testing, this flag can be useful to conserve inotify file descriptors, as
		422	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		423
		424	=item C<EVFLAG_SIGNALFD>
		425
		426	When this flag is specified, then libev will attempt to use the
		427	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		428	delivers signals synchronously, which makes it both faster and might make
		429	it possible to get the queued signal data. It can also simplify signal
		430	handling with threads, as long as you properly block signals in your
		431	threads that are not interested in handling them.
		432
		433	Signalfd will not be used by default as this changes your signal mask, and
		434	there are a lot of shoddy libraries and programs (glib's threadpool for
		435	example) that can't properly initialise their signal masks.
		436
		437	=item C<EVFLAG_NOSIGMASK>
		438
		439	When this flag is specified, then libev will avoid to modify the signal
		440	mask. Specifically, this means you ahve to make sure signals are unblocked
		441	when you want to receive them.
		442
		443	This behaviour is useful when you want to do your own signal handling, or
		444	want to handle signals only in specific threads and want to avoid libev
		445	unblocking the signals.
		446
		447	This flag's behaviour will become the default in future versions of libev.
		448
143	=item C<EVMETHOD_SELECT> (portable select backend)	449	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		450
		451	This is your standard select(2) backend. Not I<completely> standard, as
		452	libev tries to roll its own fd_set with no limits on the number of fds,
		453	but if that fails, expect a fairly low limit on the number of fds when
		454	using this backend. It doesn't scale too well (O(highest_fd)), but its
		455	usually the fastest backend for a low number of (low-numbered :) fds.
		456
		457	To get good performance out of this backend you need a high amount of
		458	parallelism (most of the file descriptors should be busy). If you are
		459	writing a server, you should C<accept ()> in a loop to accept as many
		460	connections as possible during one iteration. You might also want to have
		461	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		462	readiness notifications you get per iteration.
		463
		464	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		465	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		466	C<exceptfds> set on that platform).
		467
145	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	468	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		469
147	=item C<EVMETHOD_EPOLL> (linux only)	470	And this is your standard poll(2) backend. It's more complicated
		471	than select, but handles sparse fds better and has no artificial
		472	limit on the number of fds you can use (except it will slow down
		473	considerably with a lot of inactive fds). It scales similarly to select,
		474	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		475	performance tips.
148		476
149	=item C<EVMETHOD_KQUEUE> (some bsds only)	477	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		478	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
150		479
151	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	480	=item C<EVBACKEND_EPOLL> (value 4, Linux)
152		481
153	=item C<EVMETHOD_PORT> (solaris 10 only)	482	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		483	kernels).
154		484
155	If one or more of these are ored into the flags value, then only these	485	For few fds, this backend is a bit little slower than poll and select,
156	backends will be tried (in the reverse order as given here). If one are	486	but it scales phenomenally better. While poll and select usually scale
157	specified, any backend will do.	487	like O(total_fds) where n is the total number of fds (or the highest fd),
		488	epoll scales either O(1) or O(active_fds).
		489
		490	The epoll mechanism deserves honorable mention as the most misdesigned
		491	of the more advanced event mechanisms: mere annoyances include silently
		492	dropping file descriptors, requiring a system call per change per file
		493	descriptor (and unnecessary guessing of parameters), problems with dup,
		494	returning before the timeout value, resulting in additional iterations
		495	(and only giving 5ms accuracy while select on the same platform gives
		496	0.1ms) and so on. The biggest issue is fork races, however - if a program
		497	forks then I<both> parent and child process have to recreate the epoll
		498	set, which can take considerable time (one syscall per file descriptor)
		499	and is of course hard to detect.
		500
		501	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		502	of course I<doesn't>, and epoll just loves to report events for totally
		503	I<different> file descriptors (even already closed ones, so one cannot
		504	even remove them from the set) than registered in the set (especially
		505	on SMP systems). Libev tries to counter these spurious notifications by
		506	employing an additional generation counter and comparing that against the
		507	events to filter out spurious ones, recreating the set when required. Last
		508	not least, it also refuses to work with some file descriptors which work
		509	perfectly fine with C<select> (files, many character devices...).
		510
		511	Epoll is truly the train wreck analog among event poll mechanisms.
		512
		513	While stopping, setting and starting an I/O watcher in the same iteration
		514	will result in some caching, there is still a system call per such
		515	incident (because the same I<file descriptor> could point to a different
		516	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		517	file descriptors might not work very well if you register events for both
		518	file descriptors.
		519
		520	Best performance from this backend is achieved by not unregistering all
		521	watchers for a file descriptor until it has been closed, if possible,
		522	i.e. keep at least one watcher active per fd at all times. Stopping and
		523	starting a watcher (without re-setting it) also usually doesn't cause
		524	extra overhead. A fork can both result in spurious notifications as well
		525	as in libev having to destroy and recreate the epoll object, which can
		526	take considerable time and thus should be avoided.
		527
		528	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		529	faster than epoll for maybe up to a hundred file descriptors, depending on
		530	the usage. So sad.
		531
		532	While nominally embeddable in other event loops, this feature is broken in
		533	all kernel versions tested so far.
		534
		535	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		536	C<EVBACKEND_POLL>.
		537
		538	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		539
		540	Kqueue deserves special mention, as at the time of this writing, it
		541	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		542	with anything but sockets and pipes, except on Darwin, where of course
		543	it's completely useless). Unlike epoll, however, whose brokenness
		544	is by design, these kqueue bugs can (and eventually will) be fixed
		545	without API changes to existing programs. For this reason it's not being
		546	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		547	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		548	system like NetBSD.
		549
		550	You still can embed kqueue into a normal poll or select backend and use it
		551	only for sockets (after having made sure that sockets work with kqueue on
		552	the target platform). See C<ev_embed> watchers for more info.
		553
		554	It scales in the same way as the epoll backend, but the interface to the
		555	kernel is more efficient (which says nothing about its actual speed, of
		556	course). While stopping, setting and starting an I/O watcher does never
		557	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		558	two event changes per incident. Support for C<fork ()> is very bad (but
		559	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		560	cases
		561
		562	This backend usually performs well under most conditions.
		563
		564	While nominally embeddable in other event loops, this doesn't work
		565	everywhere, so you might need to test for this. And since it is broken
		566	almost everywhere, you should only use it when you have a lot of sockets
		567	(for which it usually works), by embedding it into another event loop
		568	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		569	also broken on OS X)) and, did I mention it, using it only for sockets.
		570
		571	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		572	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		573	C<NOTE_EOF>.
		574
		575	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		576
		577	This is not implemented yet (and might never be, unless you send me an
		578	implementation). According to reports, C</dev/poll> only supports sockets
		579	and is not embeddable, which would limit the usefulness of this backend
		580	immensely.
		581
		582	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		583
		584	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		585	it's really slow, but it still scales very well (O(active_fds)).
		586
		587	While this backend scales well, it requires one system call per active
		588	file descriptor per loop iteration. For small and medium numbers of file
		589	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		590	might perform better.
		591
		592	On the positive side, this backend actually performed fully to
		593	specification in all tests and is fully embeddable, which is a rare feat
		594	among the OS-specific backends (I vastly prefer correctness over speed
		595	hacks).
		596
		597	On the negative side, the interface is I<bizarre> - so bizarre that
		598	even sun itself gets it wrong in their code examples: The event polling
		599	function sometimes returning events to the caller even though an error
		600	occurred, but with no indication whether it has done so or not (yes, it's
		601	even documented that way) - deadly for edge-triggered interfaces where
		602	you absolutely have to know whether an event occurred or not because you
		603	have to re-arm the watcher.
		604
		605	Fortunately libev seems to be able to work around these idiocies.
		606
		607	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		608	C<EVBACKEND_POLL>.
		609
		610	=item C<EVBACKEND_ALL>
		611
		612	Try all backends (even potentially broken ones that wouldn't be tried
		613	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		614	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		615
		616	It is definitely not recommended to use this flag, use whatever
		617	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		618	at all.
		619
		620	=item C<EVBACKEND_MASK>
		621
		622	Not a backend at all, but a mask to select all backend bits from a
		623	C<flags> value, in case you want to mask out any backends from a flags
		624	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
158		625
159	=back	626	=back
160		627
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	628	If one or more of the backend flags are or'ed into the flags value,
		629	then only these backends will be tried (in the reverse order as listed
		630	here). If none are specified, all backends in C<ev_recommended_backends
		631	()> will be tried.
162		632
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	633	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		634
168	=item ev_default_destroy ()	635	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		636	if (!epoller)
		637	fatal ("no epoll found here, maybe it hides under your chair");
169		638
170	Destroys the default loop again (frees all memory and kernel state	639	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	640	used if available.
172	any way whatsoever, although you cannot rely on this :).	641
		642	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
173		643
174	=item ev_loop_destroy (loop)	644	=item ev_loop_destroy (loop)
175		645
176	Like C<ev_default_destroy>, but destroys an event loop created by an	646	Destroys an event loop object (frees all memory and kernel state
177	earlier call to C<ev_loop_new>.	647	etc.). None of the active event watchers will be stopped in the normal
		648	sense, so e.g. C<ev_is_active> might still return true. It is your
		649	responsibility to either stop all watchers cleanly yourself I<before>
		650	calling this function, or cope with the fact afterwards (which is usually
		651	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		652	for example).
178		653
179	=item ev_default_fork ()	654	Note that certain global state, such as signal state (and installed signal
		655	handlers), will not be freed by this function, and related watchers (such
		656	as signal and child watchers) would need to be stopped manually.
180		657
		658	This function is normally used on loop objects allocated by
		659	C<ev_loop_new>, but it can also be used on the default loop returned by
		660	C<ev_default_loop>, in which case it is not thread-safe.
		661
		662	Note that it is not advisable to call this function on the default loop
		663	except in the rare occasion where you really need to free its resources.
		664	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		665	and C<ev_loop_destroy>.
		666
		667	=item ev_loop_fork (loop)
		668
		669	This function sets a flag that causes subsequent C<ev_run> iterations to
181	This function reinitialises the kernel state for backends that have	670	reinitialise the kernel state for backends that have one. Despite the
182	one. Despite the name, you can call it anytime, but it makes most sense	671	name, you can call it anytime, but it makes most sense after forking, in
183	after forking, in either the parent or child process (or both, but that	672	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
184	again makes little sense).	673	child before resuming or calling C<ev_run>.
185		674
186	You I<must> call this function after forking if and only if you want to	675	Again, you I<have> to call it on I<any> loop that you want to re-use after
187	use the event library in both processes. If you just fork+exec, you don't	676	a fork, I<even if you do not plan to use the loop in the parent>. This is
188	have to call it.	677	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		678	during fork.
		679
		680	On the other hand, you only need to call this function in the child
		681	process if and only if you want to use the event loop in the child. If
		682	you just fork+exec or create a new loop in the child, you don't have to
		683	call it at all (in fact, C<epoll> is so badly broken that it makes a
		684	difference, but libev will usually detect this case on its own and do a
		685	costly reset of the backend).
189		686
190	The function itself is quite fast and it's usually not a problem to call	687	The function itself is quite fast and it's usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in	688	it just in case after a fork.
192	quite nicely into a call to C<pthread_atfork>:
193		689
		690	Example: Automate calling C<ev_loop_fork> on the default loop when
		691	using pthreads.
		692
		693	static void
		694	post_fork_child (void)
		695	{
		696	ev_loop_fork (EV_DEFAULT);
		697	}
		698
		699	...
194	pthread_atfork (0, 0, ev_default_fork);	700	pthread_atfork (0, 0, post_fork_child);
195		701
196	=item ev_loop_fork (loop)	702	=item int ev_is_default_loop (loop)
197		703
198	Like C<ev_default_fork>, but acts on an event loop created by	704	Returns true when the given loop is, in fact, the default loop, and false
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	705	otherwise.
200	after fork, and how you do this is entirely your own problem.
201		706
		707	=item unsigned int ev_iteration (loop)
		708
		709	Returns the current iteration count for the event loop, which is identical
		710	to the number of times libev did poll for new events. It starts at C<0>
		711	and happily wraps around with enough iterations.
		712
		713	This value can sometimes be useful as a generation counter of sorts (it
		714	"ticks" the number of loop iterations), as it roughly corresponds with
		715	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		716	prepare and check phases.
		717
202	=item unsigned int ev_method (loop)	718	=item unsigned int ev_depth (loop)
203		719
		720	Returns the number of times C<ev_run> was entered minus the number of
		721	times C<ev_run> was exited normally, in other words, the recursion depth.
		722
		723	Outside C<ev_run>, this number is zero. In a callback, this number is
		724	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		725	in which case it is higher.
		726
		727	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		728	throwing an exception etc.), doesn't count as "exit" - consider this
		729	as a hint to avoid such ungentleman-like behaviour unless it's really
		730	convenient, in which case it is fully supported.
		731
		732	=item unsigned int ev_backend (loop)
		733
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	734	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	735	use.
206		736
207	=item ev_tstamp ev_now (loop)	737	=item ev_tstamp ev_now (loop)
208		738
209	Returns the current "event loop time", which is the time the event loop	739	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	740	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	741	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	742	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).	743	event occurring (or more correctly, libev finding out about it).
214		744
		745	=item ev_now_update (loop)
		746
		747	Establishes the current time by querying the kernel, updating the time
		748	returned by C<ev_now ()> in the progress. This is a costly operation and
		749	is usually done automatically within C<ev_run ()>.
		750
		751	This function is rarely useful, but when some event callback runs for a
		752	very long time without entering the event loop, updating libev's idea of
		753	the current time is a good idea.
		754
		755	See also L<The special problem of time updates> in the C<ev_timer> section.
		756
		757	=item ev_suspend (loop)
		758
		759	=item ev_resume (loop)
		760
		761	These two functions suspend and resume an event loop, for use when the
		762	loop is not used for a while and timeouts should not be processed.
		763
		764	A typical use case would be an interactive program such as a game: When
		765	the user presses C<^Z> to suspend the game and resumes it an hour later it
		766	would be best to handle timeouts as if no time had actually passed while
		767	the program was suspended. This can be achieved by calling C<ev_suspend>
		768	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		769	C<ev_resume> directly afterwards to resume timer processing.
		770
		771	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		772	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		773	will be rescheduled (that is, they will lose any events that would have
		774	occurred while suspended).
		775
		776	After calling C<ev_suspend> you B<must not> call I<any> function on the
		777	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		778	without a previous call to C<ev_suspend>.
		779
		780	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		781	event loop time (see C<ev_now_update>).
		782
215	=item ev_loop (loop, int flags)	783	=item ev_run (loop, int flags)
216		784
217	Finally, this is it, the event handler. This function usually is called	785	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	786	after you have initialised all your watchers and you want to start
219	events.	787	handling events. It will ask the operating system for any new events, call
		788	the watcher callbacks, an then repeat the whole process indefinitely: This
		789	is why event loops are called I<loops>.
220		790
221	If the flags argument is specified as 0, it will not return until either	791	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.	792	until either no event watchers are active anymore or C<ev_break> was
		793	called.
223		794
		795	Please note that an explicit C<ev_break> is usually better than
		796	relying on all watchers to be stopped when deciding when a program has
		797	finished (especially in interactive programs), but having a program
		798	that automatically loops as long as it has to and no longer by virtue
		799	of relying on its watchers stopping correctly, that is truly a thing of
		800	beauty.
		801
		802	This function is also I<mostly> exception-safe - you can break out of
		803	a C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		804	exception and so on. This does not decrement the C<ev_depth> value, nor
		805	will it clear any outstanding C<EVBREAK_ONE> breaks.
		806
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	807	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	808	those events and any already outstanding ones, but will not wait and
226	case there are no events and will return after one iteration of the loop.	809	block your process in case there are no events and will return after one
		810	iteration of the loop. This is sometimes useful to poll and handle new
		811	events while doing lengthy calculations, to keep the program responsive.
227		812
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	813	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	814	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives, and will return after	815	will block your process until at least one new event arrives (which could
		816	be an event internal to libev itself, so there is no guarantee that a
		817	user-registered callback will be called), and will return after one
231	one iteration of the loop.	818	iteration of the loop.
232		819
233	This flags value could be used to implement alternative looping	820	This is useful if you are waiting for some external event in conjunction
234	constructs, but the C<prepare> and C<check> watchers provide a better and	821	with something not expressible using other libev watchers (i.e. "roll your
235	more generic mechanism.	822	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		823	usually a better approach for this kind of thing.
236		824
		825	Here are the gory details of what C<ev_run> does:
		826
		827	- Increment loop depth.
		828	- Reset the ev_break status.
		829	- Before the first iteration, call any pending watchers.
		830	LOOP:
		831	- If EVFLAG_FORKCHECK was used, check for a fork.
		832	- If a fork was detected (by any means), queue and call all fork watchers.
		833	- Queue and call all prepare watchers.
		834	- If ev_break was called, goto FINISH.
		835	- If we have been forked, detach and recreate the kernel state
		836	as to not disturb the other process.
		837	- Update the kernel state with all outstanding changes.
		838	- Update the "event loop time" (ev_now ()).
		839	- Calculate for how long to sleep or block, if at all
		840	(active idle watchers, EVRUN_NOWAIT or not having
		841	any active watchers at all will result in not sleeping).
		842	- Sleep if the I/O and timer collect interval say so.
		843	- Increment loop iteration counter.
		844	- Block the process, waiting for any events.
		845	- Queue all outstanding I/O (fd) events.
		846	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		847	- Queue all expired timers.
		848	- Queue all expired periodics.
		849	- Queue all idle watchers with priority higher than that of pending events.
		850	- Queue all check watchers.
		851	- Call all queued watchers in reverse order (i.e. check watchers first).
		852	Signals and child watchers are implemented as I/O watchers, and will
		853	be handled here by queueing them when their watcher gets executed.
		854	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		855	were used, or there are no active watchers, goto FINISH, otherwise
		856	continue with step LOOP.
		857	FINISH:
		858	- Reset the ev_break status iff it was EVBREAK_ONE.
		859	- Decrement the loop depth.
		860	- Return.
		861
		862	Example: Queue some jobs and then loop until no events are outstanding
		863	anymore.
		864
		865	... queue jobs here, make sure they register event watchers as long
		866	... as they still have work to do (even an idle watcher will do..)
		867	ev_run (my_loop, 0);
		868	... jobs done or somebody called unloop. yeah!
		869
237	=item ev_unloop (loop, how)	870	=item ev_break (loop, how)
238		871
239	Can be used to make a call to C<ev_loop> return early (but only after it	872	Can be used to make a call to C<ev_run> return early (but only after it
240	has processed all outstanding events). The C<how> argument must be either	873	has processed all outstanding events). The C<how> argument must be either
241	C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or	874	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
242	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	875	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		876
		877	This "break state" will be cleared on the next call to C<ev_run>.
		878
		879	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		880	which case it will have no effect.
243		881
244	=item ev_ref (loop)	882	=item ev_ref (loop)
245		883
246	=item ev_unref (loop)	884	=item ev_unref (loop)
247		885
248	Ref/unref can be used to add or remove a reference count on the event	886	Ref/unref can be used to add or remove a reference count on the event
249	loop: Every watcher keeps one reference, and as long as the reference	887	loop: Every watcher keeps one reference, and as long as the reference
250	count is nonzero, C<ev_loop> will not return on its own. If you have	888	count is nonzero, C<ev_run> will not return on its own.
251	a watcher you never unregister that should not keep C<ev_loop> from	889
252	returning, ev_unref() after starting, and ev_ref() before stopping it. For	890	This is useful when you have a watcher that you never intend to
		891	unregister, but that nevertheless should not keep C<ev_run> from
		892	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		893	before stopping it.
		894
253	example, libev itself uses this for its internal signal pipe: It is not	895	As an example, libev itself uses this for its internal signal pipe: It
254	visible to the libev user and should not keep C<ev_loop> from exiting if	896	is not visible to the libev user and should not keep C<ev_run> from
255	no event watchers registered by it are active. It is also an excellent	897	exiting if no event watchers registered by it are active. It is also an
256	way to do this for generic recurring timers or from within third-party	898	excellent way to do this for generic recurring timers or from within
257	libraries. Just remember to I<unref after start> and I<ref before stop>.	899	third-party libraries. Just remember to I<unref after start> and I<ref
		900	before stop> (but only if the watcher wasn't active before, or was active
		901	before, respectively. Note also that libev might stop watchers itself
		902	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		903	in the callback).
		904
		905	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		906	running when nothing else is active.
		907
		908	ev_signal exitsig;
		909	ev_signal_init (&exitsig, sig_cb, SIGINT);
		910	ev_signal_start (loop, &exitsig);
		911	ev_unref (loop);
		912
		913	Example: For some weird reason, unregister the above signal handler again.
		914
		915	ev_ref (loop);
		916	ev_signal_stop (loop, &exitsig);
		917
		918	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		919
		920	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		921
		922	These advanced functions influence the time that libev will spend waiting
		923	for events. Both time intervals are by default C<0>, meaning that libev
		924	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		925	latency.
		926
		927	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		928	allows libev to delay invocation of I/O and timer/periodic callbacks
		929	to increase efficiency of loop iterations (or to increase power-saving
		930	opportunities).
		931
		932	The idea is that sometimes your program runs just fast enough to handle
		933	one (or very few) event(s) per loop iteration. While this makes the
		934	program responsive, it also wastes a lot of CPU time to poll for new
		935	events, especially with backends like C<select ()> which have a high
		936	overhead for the actual polling but can deliver many events at once.
		937
		938	By setting a higher I<io collect interval> you allow libev to spend more
		939	time collecting I/O events, so you can handle more events per iteration,
		940	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		941	C<ev_timer>) will be not affected. Setting this to a non-null value will
		942	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		943	sleep time ensures that libev will not poll for I/O events more often then
		944	once per this interval, on average.
		945
		946	Likewise, by setting a higher I<timeout collect interval> you allow libev
		947	to spend more time collecting timeouts, at the expense of increased
		948	latency/jitter/inexactness (the watcher callback will be called
		949	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		950	value will not introduce any overhead in libev.
		951
		952	Many (busy) programs can usually benefit by setting the I/O collect
		953	interval to a value near C<0.1> or so, which is often enough for
		954	interactive servers (of course not for games), likewise for timeouts. It
		955	usually doesn't make much sense to set it to a lower value than C<0.01>,
		956	as this approaches the timing granularity of most systems. Note that if
		957	you do transactions with the outside world and you can't increase the
		958	parallelity, then this setting will limit your transaction rate (if you
		959	need to poll once per transaction and the I/O collect interval is 0.01,
		960	then you can't do more than 100 transactions per second).
		961
		962	Setting the I<timeout collect interval> can improve the opportunity for
		963	saving power, as the program will "bundle" timer callback invocations that
		964	are "near" in time together, by delaying some, thus reducing the number of
		965	times the process sleeps and wakes up again. Another useful technique to
		966	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		967	they fire on, say, one-second boundaries only.
		968
		969	Example: we only need 0.1s timeout granularity, and we wish not to poll
		970	more often than 100 times per second:
		971
		972	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		973	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		974
		975	=item ev_invoke_pending (loop)
		976
		977	This call will simply invoke all pending watchers while resetting their
		978	pending state. Normally, C<ev_run> does this automatically when required,
		979	but when overriding the invoke callback this call comes handy. This
		980	function can be invoked from a watcher - this can be useful for example
		981	when you want to do some lengthy calculation and want to pass further
		982	event handling to another thread (you still have to make sure only one
		983	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		984
		985	=item int ev_pending_count (loop)
		986
		987	Returns the number of pending watchers - zero indicates that no watchers
		988	are pending.
		989
		990	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		991
		992	This overrides the invoke pending functionality of the loop: Instead of
		993	invoking all pending watchers when there are any, C<ev_run> will call
		994	this callback instead. This is useful, for example, when you want to
		995	invoke the actual watchers inside another context (another thread etc.).
		996
		997	If you want to reset the callback, use C<ev_invoke_pending> as new
		998	callback.
		999
		1000	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		1001
		1002	Sometimes you want to share the same loop between multiple threads. This
		1003	can be done relatively simply by putting mutex_lock/unlock calls around
		1004	each call to a libev function.
		1005
		1006	However, C<ev_run> can run an indefinite time, so it is not feasible
		1007	to wait for it to return. One way around this is to wake up the event
		1008	loop via C<ev_break> and C<av_async_send>, another way is to set these
		1009	I<release> and I<acquire> callbacks on the loop.
		1010
		1011	When set, then C<release> will be called just before the thread is
		1012	suspended waiting for new events, and C<acquire> is called just
		1013	afterwards.
		1014
		1015	Ideally, C<release> will just call your mutex_unlock function, and
		1016	C<acquire> will just call the mutex_lock function again.
		1017
		1018	While event loop modifications are allowed between invocations of
		1019	C<release> and C<acquire> (that's their only purpose after all), no
		1020	modifications done will affect the event loop, i.e. adding watchers will
		1021	have no effect on the set of file descriptors being watched, or the time
		1022	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1023	to take note of any changes you made.
		1024
		1025	In theory, threads executing C<ev_run> will be async-cancel safe between
		1026	invocations of C<release> and C<acquire>.
		1027
		1028	See also the locking example in the C<THREADS> section later in this
		1029	document.
		1030
		1031	=item ev_set_userdata (loop, void *data)
		1032
		1033	=item void *ev_userdata (loop)
		1034
		1035	Set and retrieve a single C<void *> associated with a loop. When
		1036	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1037	C<0>.
		1038
		1039	These two functions can be used to associate arbitrary data with a loop,
		1040	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1041	C<acquire> callbacks described above, but of course can be (ab-)used for
		1042	any other purpose as well.
		1043
		1044	=item ev_verify (loop)
		1045
		1046	This function only does something when C<EV_VERIFY> support has been
		1047	compiled in, which is the default for non-minimal builds. It tries to go
		1048	through all internal structures and checks them for validity. If anything
		1049	is found to be inconsistent, it will print an error message to standard
		1050	error and call C<abort ()>.
		1051
		1052	This can be used to catch bugs inside libev itself: under normal
		1053	circumstances, this function will never abort as of course libev keeps its
		1054	data structures consistent.
258		1055
259	=back	1056	=back
260		1057
		1058
261	=head1 ANATOMY OF A WATCHER	1059	=head1 ANATOMY OF A WATCHER
262		1060
		1061	In the following description, uppercase C<TYPE> in names stands for the
		1062	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1063	watchers and C<ev_io_start> for I/O watchers.
		1064
263	A watcher is a structure that you create and register to record your	1065	A watcher is an opaque structure that you allocate and register to record
264	interest in some event. For instance, if you want to wait for STDIN to	1066	your interest in some event. To make a concrete example, imagine you want
265	become readable, you would create an C<ev_io> watcher for that:	1067	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1068	for that:
266		1069
267	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1070	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
268	{	1071	{
269	ev_io_stop (w);	1072	ev_io_stop (w);
270	ev_unloop (loop, EVUNLOOP_ALL);	1073	ev_break (loop, EVBREAK_ALL);
271	}	1074	}
272		1075
273	struct ev_loop *loop = ev_default_loop (0);	1076	struct ev_loop *loop = ev_default_loop (0);
		1077
274	struct ev_io stdin_watcher;	1078	ev_io stdin_watcher;
		1079
275	ev_init (&stdin_watcher, my_cb);	1080	ev_init (&stdin_watcher, my_cb);
276	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1081	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
277	ev_io_start (loop, &stdin_watcher);	1082	ev_io_start (loop, &stdin_watcher);
		1083
278	ev_loop (loop, 0);	1084	ev_run (loop, 0);
279		1085
280	As you can see, you are responsible for allocating the memory for your	1086	As you can see, you are responsible for allocating the memory for your
281	watcher structures (and it is usually a bad idea to do this on the stack,	1087	watcher structures (and it is I<usually> a bad idea to do this on the
282	although this can sometimes be quite valid).	1088	stack).
283		1089
		1090	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1091	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1092
284	Each watcher structure must be initialised by a call to C<ev_init	1093	Each watcher structure must be initialised by a call to C<ev_init (watcher
285	(watcher *, callback)>, which expects a callback to be provided. This	1094	*, callback)>, which expects a callback to be provided. This callback is
286	callback gets invoked each time the event occurs (or, in the case of io	1095	invoked each time the event occurs (or, in the case of I/O watchers, each
287	watchers, each time the event loop detects that the file descriptor given	1096	time the event loop detects that the file descriptor given is readable
288	is readable and/or writable).	1097	and/or writable).
289		1098
290	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1099	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
291	with arguments specific to this watcher type. There is also a macro	1100	macro to configure it, with arguments specific to the watcher type. There
292	to combine initialisation and setting in one call: C<< ev_<type>_init	1101	is also a macro to combine initialisation and setting in one call: C<<
293	(watcher *, callback, ...) >>.	1102	ev_TYPE_init (watcher *, callback, ...) >>.
294		1103
295	To make the watcher actually watch out for events, you have to start it	1104	To make the watcher actually watch out for events, you have to start it
296	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1105	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
297	*) >>), and you can stop watching for events at any time by calling the	1106	*) >>), and you can stop watching for events at any time by calling the
298	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1107	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
299		1108
300	As long as your watcher is active (has been started but not stopped) you	1109	As long as your watcher is active (has been started but not stopped) you
301	must not touch the values stored in it. Most specifically you must never	1110	must not touch the values stored in it. Most specifically you must never
302	reinitialise it or call its set method.	1111	reinitialise it or call its C<ev_TYPE_set> macro.
303
304	You can check whether an event is active by calling the C<ev_is_active
305	(watcher *)> macro. To see whether an event is outstanding (but the
306	callback for it has not been called yet) you can use the C<ev_is_pending
307	(watcher *)> macro.
308		1112
309	Each and every callback receives the event loop pointer as first, the	1113	Each and every callback receives the event loop pointer as first, the
310	registered watcher structure as second, and a bitset of received events as	1114	registered watcher structure as second, and a bitset of received events as
311	third argument.	1115	third argument.
312		1116
…		…
321	=item C<EV_WRITE>	1125	=item C<EV_WRITE>
322		1126
323	The file descriptor in the C<ev_io> watcher has become readable and/or	1127	The file descriptor in the C<ev_io> watcher has become readable and/or
324	writable.	1128	writable.
325		1129
326	=item C<EV_TIMEOUT>	1130	=item C<EV_TIMER>
327		1131
328	The C<ev_timer> watcher has timed out.	1132	The C<ev_timer> watcher has timed out.
329		1133
330	=item C<EV_PERIODIC>	1134	=item C<EV_PERIODIC>
331		1135
…		…
337		1141
338	=item C<EV_CHILD>	1142	=item C<EV_CHILD>
339		1143
340	The pid specified in the C<ev_child> watcher has received a status change.	1144	The pid specified in the C<ev_child> watcher has received a status change.
341		1145
		1146	=item C<EV_STAT>
		1147
		1148	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1149
342	=item C<EV_IDLE>	1150	=item C<EV_IDLE>
343		1151
344	The C<ev_idle> watcher has determined that you have nothing better to do.	1152	The C<ev_idle> watcher has determined that you have nothing better to do.
345		1153
346	=item C<EV_PREPARE>	1154	=item C<EV_PREPARE>
347		1155
348	=item C<EV_CHECK>	1156	=item C<EV_CHECK>
349		1157
350	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1158	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
351	to gather new events, and all C<ev_check> watchers are invoked just after	1159	to gather new events, and all C<ev_check> watchers are invoked just after
352	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1160	C<ev_run> has gathered them, but before it invokes any callbacks for any
353	received events. Callbacks of both watcher types can start and stop as	1161	received events. Callbacks of both watcher types can start and stop as
354	many watchers as they want, and all of them will be taken into account	1162	many watchers as they want, and all of them will be taken into account
355	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1163	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
356	C<ev_loop> from blocking).	1164	C<ev_run> from blocking).
		1165
		1166	=item C<EV_EMBED>
		1167
		1168	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1169
		1170	=item C<EV_FORK>
		1171
		1172	The event loop has been resumed in the child process after fork (see
		1173	C<ev_fork>).
		1174
		1175	=item C<EV_CLEANUP>
		1176
		1177	The event loop is about to be destroyed (see C<ev_cleanup>).
		1178
		1179	=item C<EV_ASYNC>
		1180
		1181	The given async watcher has been asynchronously notified (see C<ev_async>).
		1182
		1183	=item C<EV_CUSTOM>
		1184
		1185	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1186	by libev users to signal watchers (e.g. via C<ev_feed_event>).
357		1187
358	=item C<EV_ERROR>	1188	=item C<EV_ERROR>
359		1189
360	An unspecified error has occured, the watcher has been stopped. This might	1190	An unspecified error has occurred, the watcher has been stopped. This might
361	happen because the watcher could not be properly started because libev	1191	happen because the watcher could not be properly started because libev
362	ran out of memory, a file descriptor was found to be closed or any other	1192	ran out of memory, a file descriptor was found to be closed or any other
		1193	problem. Libev considers these application bugs.
		1194
363	problem. You best act on it by reporting the problem and somehow coping	1195	You best act on it by reporting the problem and somehow coping with the
364	with the watcher being stopped.	1196	watcher being stopped. Note that well-written programs should not receive
		1197	an error ever, so when your watcher receives it, this usually indicates a
		1198	bug in your program.
365		1199
366	Libev will usually signal a few "dummy" events together with an error,	1200	Libev will usually signal a few "dummy" events together with an error, for
367	for example it might indicate that a fd is readable or writable, and if	1201	example it might indicate that a fd is readable or writable, and if your
368	your callbacks is well-written it can just attempt the operation and cope	1202	callbacks is well-written it can just attempt the operation and cope with
369	with the error from read() or write(). This will not work in multithreaded	1203	the error from read() or write(). This will not work in multi-threaded
370	programs, though, so beware.	1204	programs, though, as the fd could already be closed and reused for another
		1205	thing, so beware.
371		1206
372	=back	1207	=back
373		1208
		1209	=head2 GENERIC WATCHER FUNCTIONS
		1210
		1211	=over 4
		1212
		1213	=item C<ev_init> (ev_TYPE *watcher, callback)
		1214
		1215	This macro initialises the generic portion of a watcher. The contents
		1216	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1217	the generic parts of the watcher are initialised, you I<need> to call
		1218	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1219	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1220	which rolls both calls into one.
		1221
		1222	You can reinitialise a watcher at any time as long as it has been stopped
		1223	(or never started) and there are no pending events outstanding.
		1224
		1225	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1226	int revents)>.
		1227
		1228	Example: Initialise an C<ev_io> watcher in two steps.
		1229
		1230	ev_io w;
		1231	ev_init (&w, my_cb);
		1232	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1233
		1234	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1235
		1236	This macro initialises the type-specific parts of a watcher. You need to
		1237	call C<ev_init> at least once before you call this macro, but you can
		1238	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1239	macro on a watcher that is active (it can be pending, however, which is a
		1240	difference to the C<ev_init> macro).
		1241
		1242	Although some watcher types do not have type-specific arguments
		1243	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1244
		1245	See C<ev_init>, above, for an example.
		1246
		1247	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1248
		1249	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1250	calls into a single call. This is the most convenient method to initialise
		1251	a watcher. The same limitations apply, of course.
		1252
		1253	Example: Initialise and set an C<ev_io> watcher in one step.
		1254
		1255	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1256
		1257	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1258
		1259	Starts (activates) the given watcher. Only active watchers will receive
		1260	events. If the watcher is already active nothing will happen.
		1261
		1262	Example: Start the C<ev_io> watcher that is being abused as example in this
		1263	whole section.
		1264
		1265	ev_io_start (EV_DEFAULT_UC, &w);
		1266
		1267	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1268
		1269	Stops the given watcher if active, and clears the pending status (whether
		1270	the watcher was active or not).
		1271
		1272	It is possible that stopped watchers are pending - for example,
		1273	non-repeating timers are being stopped when they become pending - but
		1274	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1275	pending. If you want to free or reuse the memory used by the watcher it is
		1276	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1277
		1278	=item bool ev_is_active (ev_TYPE *watcher)
		1279
		1280	Returns a true value iff the watcher is active (i.e. it has been started
		1281	and not yet been stopped). As long as a watcher is active you must not modify
		1282	it.
		1283
		1284	=item bool ev_is_pending (ev_TYPE *watcher)
		1285
		1286	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1287	events but its callback has not yet been invoked). As long as a watcher
		1288	is pending (but not active) you must not call an init function on it (but
		1289	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1290	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1291	it).
		1292
		1293	=item callback ev_cb (ev_TYPE *watcher)
		1294
		1295	Returns the callback currently set on the watcher.
		1296
		1297	=item ev_cb_set (ev_TYPE *watcher, callback)
		1298
		1299	Change the callback. You can change the callback at virtually any time
		1300	(modulo threads).
		1301
		1302	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1303
		1304	=item int ev_priority (ev_TYPE *watcher)
		1305
		1306	Set and query the priority of the watcher. The priority is a small
		1307	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1308	(default: C<-2>). Pending watchers with higher priority will be invoked
		1309	before watchers with lower priority, but priority will not keep watchers
		1310	from being executed (except for C<ev_idle> watchers).
		1311
		1312	If you need to suppress invocation when higher priority events are pending
		1313	you need to look at C<ev_idle> watchers, which provide this functionality.
		1314
		1315	You I<must not> change the priority of a watcher as long as it is active or
		1316	pending.
		1317
		1318	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1319	fine, as long as you do not mind that the priority value you query might
		1320	or might not have been clamped to the valid range.
		1321
		1322	The default priority used by watchers when no priority has been set is
		1323	always C<0>, which is supposed to not be too high and not be too low :).
		1324
		1325	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1326	priorities.
		1327
		1328	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1329
		1330	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1331	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1332	can deal with that fact, as both are simply passed through to the
		1333	callback.
		1334
		1335	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1336
		1337	If the watcher is pending, this function clears its pending status and
		1338	returns its C<revents> bitset (as if its callback was invoked). If the
		1339	watcher isn't pending it does nothing and returns C<0>.
		1340
		1341	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1342	callback to be invoked, which can be accomplished with this function.
		1343
		1344	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1345
		1346	Feeds the given event set into the event loop, as if the specified event
		1347	had happened for the specified watcher (which must be a pointer to an
		1348	initialised but not necessarily started event watcher). Obviously you must
		1349	not free the watcher as long as it has pending events.
		1350
		1351	Stopping the watcher, letting libev invoke it, or calling
		1352	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1353	not started in the first place.
		1354
		1355	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1356	functions that do not need a watcher.
		1357
		1358	=back
		1359
374	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1360	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375		1361
376	Each watcher has, by default, a member C<void *data> that you can change	1362	Each watcher has, by default, a member C<void *data> that you can change
377	and read at any time, libev will completely ignore it. This can be used	1363	and read at any time: libev will completely ignore it. This can be used
378	to associate arbitrary data with your watcher. If you need more data and	1364	to associate arbitrary data with your watcher. If you need more data and
379	don't want to allocate memory and store a pointer to it in that data	1365	don't want to allocate memory and store a pointer to it in that data
380	member, you can also "subclass" the watcher type and provide your own	1366	member, you can also "subclass" the watcher type and provide your own
381	data:	1367	data:
382		1368
383	struct my_io	1369	struct my_io
384	{	1370	{
385	struct ev_io io;	1371	ev_io io;
386	int otherfd;	1372	int otherfd;
387	void *somedata;	1373	void *somedata;
388	struct whatever *mostinteresting;	1374	struct whatever *mostinteresting;
389	}	1375	};
		1376
		1377	...
		1378	struct my_io w;
		1379	ev_io_init (&w.io, my_cb, fd, EV_READ);
390		1380
391	And since your callback will be called with a pointer to the watcher, you	1381	And since your callback will be called with a pointer to the watcher, you
392	can cast it back to your own type:	1382	can cast it back to your own type:
393		1383
394	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1384	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
395	{	1385	{
396	struct my_io *w = (struct my_io *)w_;	1386	struct my_io *w = (struct my_io *)w_;
397	...	1387	...
398	}	1388	}
399		1389
400	More interesting and less C-conformant ways of catsing your callback type	1390	More interesting and less C-conformant ways of casting your callback type
401	have been omitted....	1391	instead have been omitted.
		1392
		1393	Another common scenario is to use some data structure with multiple
		1394	embedded watchers:
		1395
		1396	struct my_biggy
		1397	{
		1398	int some_data;
		1399	ev_timer t1;
		1400	ev_timer t2;
		1401	}
		1402
		1403	In this case getting the pointer to C<my_biggy> is a bit more
		1404	complicated: Either you store the address of your C<my_biggy> struct
		1405	in the C<data> member of the watcher (for woozies), or you need to use
		1406	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1407	programmers):
		1408
		1409	#include <stddef.h>
		1410
		1411	static void
		1412	t1_cb (EV_P_ ev_timer *w, int revents)
		1413	{
		1414	struct my_biggy big = (struct my_biggy *)
		1415	(((char *)w) - offsetof (struct my_biggy, t1));
		1416	}
		1417
		1418	static void
		1419	t2_cb (EV_P_ ev_timer *w, int revents)
		1420	{
		1421	struct my_biggy big = (struct my_biggy *)
		1422	(((char *)w) - offsetof (struct my_biggy, t2));
		1423	}
		1424
		1425	=head2 WATCHER STATES
		1426
		1427	There are various watcher states mentioned throughout this manual -
		1428	active, pending and so on. In this section these states and the rules to
		1429	transition between them will be described in more detail - and while these
		1430	rules might look complicated, they usually do "the right thing".
		1431
		1432	=over 4
		1433
		1434	=item initialiased
		1435
		1436	Before a watcher can be registered with the event looop it has to be
		1437	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1438	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1439
		1440	In this state it is simply some block of memory that is suitable for use
		1441	in an event loop. It can be moved around, freed, reused etc. at will.
		1442
		1443	=item started/running/active
		1444
		1445	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1446	property of the event loop, and is actively waiting for events. While in
		1447	this state it cannot be accessed (except in a few documented ways), moved,
		1448	freed or anything else - the only legal thing is to keep a pointer to it,
		1449	and call libev functions on it that are documented to work on active watchers.
		1450
		1451	=item pending
		1452
		1453	If a watcher is active and libev determines that an event it is interested
		1454	in has occurred (such as a timer expiring), it will become pending. It will
		1455	stay in this pending state until either it is stopped or its callback is
		1456	about to be invoked, so it is not normally pending inside the watcher
		1457	callback.
		1458
		1459	The watcher might or might not be active while it is pending (for example,
		1460	an expired non-repeating timer can be pending but no longer active). If it
		1461	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1462	but it is still property of the event loop at this time, so cannot be
		1463	moved, freed or reused. And if it is active the rules described in the
		1464	previous item still apply.
		1465
		1466	It is also possible to feed an event on a watcher that is not active (e.g.
		1467	via C<ev_feed_event>), in which case it becomes pending without being
		1468	active.
		1469
		1470	=item stopped
		1471
		1472	A watcher can be stopped implicitly by libev (in which case it might still
		1473	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1474	latter will clear any pending state the watcher might be in, regardless
		1475	of whether it was active or not, so stopping a watcher explicitly before
		1476	freeing it is often a good idea.
		1477
		1478	While stopped (and not pending) the watcher is essentially in the
		1479	initialised state, that is it can be reused, moved, modified in any way
		1480	you wish.
		1481
		1482	=back
		1483
		1484	=head2 WATCHER PRIORITY MODELS
		1485
		1486	Many event loops support I<watcher priorities>, which are usually small
		1487	integers that influence the ordering of event callback invocation
		1488	between watchers in some way, all else being equal.
		1489
		1490	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1491	description for the more technical details such as the actual priority
		1492	range.
		1493
		1494	There are two common ways how these these priorities are being interpreted
		1495	by event loops:
		1496
		1497	In the more common lock-out model, higher priorities "lock out" invocation
		1498	of lower priority watchers, which means as long as higher priority
		1499	watchers receive events, lower priority watchers are not being invoked.
		1500
		1501	The less common only-for-ordering model uses priorities solely to order
		1502	callback invocation within a single event loop iteration: Higher priority
		1503	watchers are invoked before lower priority ones, but they all get invoked
		1504	before polling for new events.
		1505
		1506	Libev uses the second (only-for-ordering) model for all its watchers
		1507	except for idle watchers (which use the lock-out model).
		1508
		1509	The rationale behind this is that implementing the lock-out model for
		1510	watchers is not well supported by most kernel interfaces, and most event
		1511	libraries will just poll for the same events again and again as long as
		1512	their callbacks have not been executed, which is very inefficient in the
		1513	common case of one high-priority watcher locking out a mass of lower
		1514	priority ones.
		1515
		1516	Static (ordering) priorities are most useful when you have two or more
		1517	watchers handling the same resource: a typical usage example is having an
		1518	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1519	timeouts. Under load, data might be received while the program handles
		1520	other jobs, but since timers normally get invoked first, the timeout
		1521	handler will be executed before checking for data. In that case, giving
		1522	the timer a lower priority than the I/O watcher ensures that I/O will be
		1523	handled first even under adverse conditions (which is usually, but not
		1524	always, what you want).
		1525
		1526	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1527	will only be executed when no same or higher priority watchers have
		1528	received events, they can be used to implement the "lock-out" model when
		1529	required.
		1530
		1531	For example, to emulate how many other event libraries handle priorities,
		1532	you can associate an C<ev_idle> watcher to each such watcher, and in
		1533	the normal watcher callback, you just start the idle watcher. The real
		1534	processing is done in the idle watcher callback. This causes libev to
		1535	continuously poll and process kernel event data for the watcher, but when
		1536	the lock-out case is known to be rare (which in turn is rare :), this is
		1537	workable.
		1538
		1539	Usually, however, the lock-out model implemented that way will perform
		1540	miserably under the type of load it was designed to handle. In that case,
		1541	it might be preferable to stop the real watcher before starting the
		1542	idle watcher, so the kernel will not have to process the event in case
		1543	the actual processing will be delayed for considerable time.
		1544
		1545	Here is an example of an I/O watcher that should run at a strictly lower
		1546	priority than the default, and which should only process data when no
		1547	other events are pending:
		1548
		1549	ev_idle idle; // actual processing watcher
		1550	ev_io io; // actual event watcher
		1551
		1552	static void
		1553	io_cb (EV_P_ ev_io *w, int revents)
		1554	{
		1555	// stop the I/O watcher, we received the event, but
		1556	// are not yet ready to handle it.
		1557	ev_io_stop (EV_A_ w);
		1558
		1559	// start the idle watcher to handle the actual event.
		1560	// it will not be executed as long as other watchers
		1561	// with the default priority are receiving events.
		1562	ev_idle_start (EV_A_ &idle);
		1563	}
		1564
		1565	static void
		1566	idle_cb (EV_P_ ev_idle *w, int revents)
		1567	{
		1568	// actual processing
		1569	read (STDIN_FILENO, ...);
		1570
		1571	// have to start the I/O watcher again, as
		1572	// we have handled the event
		1573	ev_io_start (EV_P_ &io);
		1574	}
		1575
		1576	// initialisation
		1577	ev_idle_init (&idle, idle_cb);
		1578	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1579	ev_io_start (EV_DEFAULT_ &io);
		1580
		1581	In the "real" world, it might also be beneficial to start a timer, so that
		1582	low-priority connections can not be locked out forever under load. This
		1583	enables your program to keep a lower latency for important connections
		1584	during short periods of high load, while not completely locking out less
		1585	important ones.
402		1586
403		1587
404	=head1 WATCHER TYPES	1588	=head1 WATCHER TYPES
405		1589
406	This section describes each watcher in detail, but will not repeat	1590	This section describes each watcher in detail, but will not repeat
407	information given in the last section.	1591	information given in the last section. Any initialisation/set macros,
		1592	functions and members specific to the watcher type are explained.
408		1593
		1594	Members are additionally marked with either I<[read-only]>, meaning that,
		1595	while the watcher is active, you can look at the member and expect some
		1596	sensible content, but you must not modify it (you can modify it while the
		1597	watcher is stopped to your hearts content), or I<[read-write]>, which
		1598	means you can expect it to have some sensible content while the watcher
		1599	is active, but you can also modify it. Modifying it may not do something
		1600	sensible or take immediate effect (or do anything at all), but libev will
		1601	not crash or malfunction in any way.
		1602
		1603
409	=head2 C<ev_io> - is this file descriptor readable or writable	1604	=head2 C<ev_io> - is this file descriptor readable or writable?
410		1605
411	I/O watchers check whether a file descriptor is readable or writable	1606	I/O watchers check whether a file descriptor is readable or writable
412	in each iteration of the event loop (This behaviour is called	1607	in each iteration of the event loop, or, more precisely, when reading
413	level-triggering because you keep receiving events as long as the	1608	would not block the process and writing would at least be able to write
414	condition persists. Remember you can stop the watcher if you don't want to	1609	some data. This behaviour is called level-triggering because you keep
415	act on the event and neither want to receive future events).	1610	receiving events as long as the condition persists. Remember you can stop
		1611	the watcher if you don't want to act on the event and neither want to
		1612	receive future events.
416		1613
417	In general you can register as many read and/or write event watchers oer	1614	In general you can register as many read and/or write event watchers per
418	fd as you want (as long as you don't confuse yourself). Setting all file	1615	fd as you want (as long as you don't confuse yourself). Setting all file
419	descriptors to non-blocking mode is also usually a good idea (but not	1616	descriptors to non-blocking mode is also usually a good idea (but not
420	required if you know what you are doing).	1617	required if you know what you are doing).
421		1618
422	You have to be careful with dup'ed file descriptors, though. Some backends	1619	If you cannot use non-blocking mode, then force the use of a
423	(the linux epoll backend is a notable example) cannot handle dup'ed file	1620	known-to-be-good backend (at the time of this writing, this includes only
424	descriptors correctly if you register interest in two or more fds pointing	1621	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
425	to the same file/socket etc. description.	1622	descriptors for which non-blocking operation makes no sense (such as
		1623	files) - libev doesn't guarantee any specific behaviour in that case.
426		1624
427	If you must do this, then force the use of a known-to-be-good backend	1625	Another thing you have to watch out for is that it is quite easy to
428	(at the time of this writing, this includes only EVMETHOD_SELECT and	1626	receive "spurious" readiness notifications, that is your callback might
429	EVMETHOD_POLL).	1627	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1628	because there is no data. Not only are some backends known to create a
		1629	lot of those (for example Solaris ports), it is very easy to get into
		1630	this situation even with a relatively standard program structure. Thus
		1631	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1632	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1633
		1634	If you cannot run the fd in non-blocking mode (for example you should
		1635	not play around with an Xlib connection), then you have to separately
		1636	re-test whether a file descriptor is really ready with a known-to-be good
		1637	interface such as poll (fortunately in our Xlib example, Xlib already
		1638	does this on its own, so its quite safe to use). Some people additionally
		1639	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1640	indefinitely.
		1641
		1642	But really, best use non-blocking mode.
		1643
		1644	=head3 The special problem of disappearing file descriptors
		1645
		1646	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1647	descriptor (either due to calling C<close> explicitly or any other means,
		1648	such as C<dup2>). The reason is that you register interest in some file
		1649	descriptor, but when it goes away, the operating system will silently drop
		1650	this interest. If another file descriptor with the same number then is
		1651	registered with libev, there is no efficient way to see that this is, in
		1652	fact, a different file descriptor.
		1653
		1654	To avoid having to explicitly tell libev about such cases, libev follows
		1655	the following policy: Each time C<ev_io_set> is being called, libev
		1656	will assume that this is potentially a new file descriptor, otherwise
		1657	it is assumed that the file descriptor stays the same. That means that
		1658	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1659	descriptor even if the file descriptor number itself did not change.
		1660
		1661	This is how one would do it normally anyway, the important point is that
		1662	the libev application should not optimise around libev but should leave
		1663	optimisations to libev.
		1664
		1665	=head3 The special problem of dup'ed file descriptors
		1666
		1667	Some backends (e.g. epoll), cannot register events for file descriptors,
		1668	but only events for the underlying file descriptions. That means when you
		1669	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1670	events for them, only one file descriptor might actually receive events.
		1671
		1672	There is no workaround possible except not registering events
		1673	for potentially C<dup ()>'ed file descriptors, or to resort to
		1674	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1675
		1676	=head3 The special problem of fork
		1677
		1678	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1679	useless behaviour. Libev fully supports fork, but needs to be told about
		1680	it in the child.
		1681
		1682	To support fork in your programs, you either have to call
		1683	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1684	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1685	C<EVBACKEND_POLL>.
		1686
		1687	=head3 The special problem of SIGPIPE
		1688
		1689	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1690	when writing to a pipe whose other end has been closed, your program gets
		1691	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1692	this is sensible behaviour, for daemons, this is usually undesirable.
		1693
		1694	So when you encounter spurious, unexplained daemon exits, make sure you
		1695	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1696	somewhere, as that would have given you a big clue).
		1697
		1698	=head3 The special problem of accept()ing when you can't
		1699
		1700	Many implementations of the POSIX C<accept> function (for example,
		1701	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1702	connection from the pending queue in all error cases.
		1703
		1704	For example, larger servers often run out of file descriptors (because
		1705	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1706	rejecting the connection, leading to libev signalling readiness on
		1707	the next iteration again (the connection still exists after all), and
		1708	typically causing the program to loop at 100% CPU usage.
		1709
		1710	Unfortunately, the set of errors that cause this issue differs between
		1711	operating systems, there is usually little the app can do to remedy the
		1712	situation, and no known thread-safe method of removing the connection to
		1713	cope with overload is known (to me).
		1714
		1715	One of the easiest ways to handle this situation is to just ignore it
		1716	- when the program encounters an overload, it will just loop until the
		1717	situation is over. While this is a form of busy waiting, no OS offers an
		1718	event-based way to handle this situation, so it's the best one can do.
		1719
		1720	A better way to handle the situation is to log any errors other than
		1721	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1722	messages, and continue as usual, which at least gives the user an idea of
		1723	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1724	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1725	usage.
		1726
		1727	If your program is single-threaded, then you could also keep a dummy file
		1728	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1729	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1730	close that fd, and create a new dummy fd. This will gracefully refuse
		1731	clients under typical overload conditions.
		1732
		1733	The last way to handle it is to simply log the error and C<exit>, as
		1734	is often done with C<malloc> failures, but this results in an easy
		1735	opportunity for a DoS attack.
		1736
		1737	=head3 Watcher-Specific Functions
430		1738
431	=over 4	1739	=over 4
432		1740
433	=item ev_io_init (ev_io *, callback, int fd, int events)	1741	=item ev_io_init (ev_io *, callback, int fd, int events)
434		1742
435	=item ev_io_set (ev_io *, int fd, int events)	1743	=item ev_io_set (ev_io *, int fd, int events)
436		1744
437	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1745	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
438	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1746	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
439	EV_WRITE> to receive the given events.	1747	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1748
		1749	=item int fd [read-only]
		1750
		1751	The file descriptor being watched.
		1752
		1753	=item int events [read-only]
		1754
		1755	The events being watched.
440		1756
441	=back	1757	=back
442		1758
		1759	=head3 Examples
		1760
		1761	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1762	readable, but only once. Since it is likely line-buffered, you could
		1763	attempt to read a whole line in the callback.
		1764
		1765	static void
		1766	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1767	{
		1768	ev_io_stop (loop, w);
		1769	.. read from stdin here (or from w->fd) and handle any I/O errors
		1770	}
		1771
		1772	...
		1773	struct ev_loop *loop = ev_default_init (0);
		1774	ev_io stdin_readable;
		1775	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1776	ev_io_start (loop, &stdin_readable);
		1777	ev_run (loop, 0);
		1778
		1779
443	=head2 C<ev_timer> - relative and optionally recurring timeouts	1780	=head2 C<ev_timer> - relative and optionally repeating timeouts
444		1781
445	Timer watchers are simple relative timers that generate an event after a	1782	Timer watchers are simple relative timers that generate an event after a
446	given time, and optionally repeating in regular intervals after that.	1783	given time, and optionally repeating in regular intervals after that.
447		1784
448	The timers are based on real time, that is, if you register an event that	1785	The timers are based on real time, that is, if you register an event that
449	times out after an hour and youreset your system clock to last years	1786	times out after an hour and you reset your system clock to January last
450	time, it will still time out after (roughly) and hour. "Roughly" because	1787	year, it will still time out after (roughly) one hour. "Roughly" because
451	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1788	detecting time jumps is hard, and some inaccuracies are unavoidable (the
452	monotonic clock option helps a lot here).	1789	monotonic clock option helps a lot here).
		1790
		1791	The callback is guaranteed to be invoked only I<after> its timeout has
		1792	passed (not I<at>, so on systems with very low-resolution clocks this
		1793	might introduce a small delay). If multiple timers become ready during the
		1794	same loop iteration then the ones with earlier time-out values are invoked
		1795	before ones of the same priority with later time-out values (but this is
		1796	no longer true when a callback calls C<ev_run> recursively).
		1797
		1798	=head3 Be smart about timeouts
		1799
		1800	Many real-world problems involve some kind of timeout, usually for error
		1801	recovery. A typical example is an HTTP request - if the other side hangs,
		1802	you want to raise some error after a while.
		1803
		1804	What follows are some ways to handle this problem, from obvious and
		1805	inefficient to smart and efficient.
		1806
		1807	In the following, a 60 second activity timeout is assumed - a timeout that
		1808	gets reset to 60 seconds each time there is activity (e.g. each time some
		1809	data or other life sign was received).
		1810
		1811	=over 4
		1812
		1813	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1814
		1815	This is the most obvious, but not the most simple way: In the beginning,
		1816	start the watcher:
		1817
		1818	ev_timer_init (timer, callback, 60., 0.);
		1819	ev_timer_start (loop, timer);
		1820
		1821	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1822	and start it again:
		1823
		1824	ev_timer_stop (loop, timer);
		1825	ev_timer_set (timer, 60., 0.);
		1826	ev_timer_start (loop, timer);
		1827
		1828	This is relatively simple to implement, but means that each time there is
		1829	some activity, libev will first have to remove the timer from its internal
		1830	data structure and then add it again. Libev tries to be fast, but it's
		1831	still not a constant-time operation.
		1832
		1833	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1834
		1835	This is the easiest way, and involves using C<ev_timer_again> instead of
		1836	C<ev_timer_start>.
		1837
		1838	To implement this, configure an C<ev_timer> with a C<repeat> value
		1839	of C<60> and then call C<ev_timer_again> at start and each time you
		1840	successfully read or write some data. If you go into an idle state where
		1841	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1842	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1843
		1844	That means you can ignore both the C<ev_timer_start> function and the
		1845	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1846	member and C<ev_timer_again>.
		1847
		1848	At start:
		1849
		1850	ev_init (timer, callback);
		1851	timer->repeat = 60.;
		1852	ev_timer_again (loop, timer);
		1853
		1854	Each time there is some activity:
		1855
		1856	ev_timer_again (loop, timer);
		1857
		1858	It is even possible to change the time-out on the fly, regardless of
		1859	whether the watcher is active or not:
		1860
		1861	timer->repeat = 30.;
		1862	ev_timer_again (loop, timer);
		1863
		1864	This is slightly more efficient then stopping/starting the timer each time
		1865	you want to modify its timeout value, as libev does not have to completely
		1866	remove and re-insert the timer from/into its internal data structure.
		1867
		1868	It is, however, even simpler than the "obvious" way to do it.
		1869
		1870	=item 3. Let the timer time out, but then re-arm it as required.
		1871
		1872	This method is more tricky, but usually most efficient: Most timeouts are
		1873	relatively long compared to the intervals between other activity - in
		1874	our example, within 60 seconds, there are usually many I/O events with
		1875	associated activity resets.
		1876
		1877	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1878	but remember the time of last activity, and check for a real timeout only
		1879	within the callback:
		1880
		1881	ev_tstamp last_activity; // time of last activity
		1882
		1883	static void
		1884	callback (EV_P_ ev_timer *w, int revents)
		1885	{
		1886	ev_tstamp now = ev_now (EV_A);
		1887	ev_tstamp timeout = last_activity + 60.;
		1888
		1889	// if last_activity + 60. is older than now, we did time out
		1890	if (timeout < now)
		1891	{
		1892	// timeout occurred, take action
		1893	}
		1894	else
		1895	{
		1896	// callback was invoked, but there was some activity, re-arm
		1897	// the watcher to fire in last_activity + 60, which is
		1898	// guaranteed to be in the future, so "again" is positive:
		1899	w->repeat = timeout - now;
		1900	ev_timer_again (EV_A_ w);
		1901	}
		1902	}
		1903
		1904	To summarise the callback: first calculate the real timeout (defined
		1905	as "60 seconds after the last activity"), then check if that time has
		1906	been reached, which means something I<did>, in fact, time out. Otherwise
		1907	the callback was invoked too early (C<timeout> is in the future), so
		1908	re-schedule the timer to fire at that future time, to see if maybe we have
		1909	a timeout then.
		1910
		1911	Note how C<ev_timer_again> is used, taking advantage of the
		1912	C<ev_timer_again> optimisation when the timer is already running.
		1913
		1914	This scheme causes more callback invocations (about one every 60 seconds
		1915	minus half the average time between activity), but virtually no calls to
		1916	libev to change the timeout.
		1917
		1918	To start the timer, simply initialise the watcher and set C<last_activity>
		1919	to the current time (meaning we just have some activity :), then call the
		1920	callback, which will "do the right thing" and start the timer:
		1921
		1922	ev_init (timer, callback);
		1923	last_activity = ev_now (loop);
		1924	callback (loop, timer, EV_TIMER);
		1925
		1926	And when there is some activity, simply store the current time in
		1927	C<last_activity>, no libev calls at all:
		1928
		1929	last_activity = ev_now (loop);
		1930
		1931	This technique is slightly more complex, but in most cases where the
		1932	time-out is unlikely to be triggered, much more efficient.
		1933
		1934	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1935	callback :) - just change the timeout and invoke the callback, which will
		1936	fix things for you.
		1937
		1938	=item 4. Wee, just use a double-linked list for your timeouts.
		1939
		1940	If there is not one request, but many thousands (millions...), all
		1941	employing some kind of timeout with the same timeout value, then one can
		1942	do even better:
		1943
		1944	When starting the timeout, calculate the timeout value and put the timeout
		1945	at the I<end> of the list.
		1946
		1947	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1948	the list is expected to fire (for example, using the technique #3).
		1949
		1950	When there is some activity, remove the timer from the list, recalculate
		1951	the timeout, append it to the end of the list again, and make sure to
		1952	update the C<ev_timer> if it was taken from the beginning of the list.
		1953
		1954	This way, one can manage an unlimited number of timeouts in O(1) time for
		1955	starting, stopping and updating the timers, at the expense of a major
		1956	complication, and having to use a constant timeout. The constant timeout
		1957	ensures that the list stays sorted.
		1958
		1959	=back
		1960
		1961	So which method the best?
		1962
		1963	Method #2 is a simple no-brain-required solution that is adequate in most
		1964	situations. Method #3 requires a bit more thinking, but handles many cases
		1965	better, and isn't very complicated either. In most case, choosing either
		1966	one is fine, with #3 being better in typical situations.
		1967
		1968	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1969	rather complicated, but extremely efficient, something that really pays
		1970	off after the first million or so of active timers, i.e. it's usually
		1971	overkill :)
		1972
		1973	=head3 The special problem of time updates
		1974
		1975	Establishing the current time is a costly operation (it usually takes at
		1976	least two system calls): EV therefore updates its idea of the current
		1977	time only before and after C<ev_run> collects new events, which causes a
		1978	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1979	lots of events in one iteration.
453		1980
454	The relative timeouts are calculated relative to the C<ev_now ()>	1981	The relative timeouts are calculated relative to the C<ev_now ()>
455	time. This is usually the right thing as this timestamp refers to the time	1982	time. This is usually the right thing as this timestamp refers to the time
456	of the event triggering whatever timeout you are modifying/starting. If	1983	of the event triggering whatever timeout you are modifying/starting. If
457	you suspect event processing to be delayed and you *need* to base the timeout	1984	you suspect event processing to be delayed and you I<need> to base the
458	ion the current time, use something like this to adjust for this:	1985	timeout on the current time, use something like this to adjust for this:
459		1986
460	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1987	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
461		1988
		1989	If the event loop is suspended for a long time, you can also force an
		1990	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1991	()>.
		1992
		1993	=head3 The special problems of suspended animation
		1994
		1995	When you leave the server world it is quite customary to hit machines that
		1996	can suspend/hibernate - what happens to the clocks during such a suspend?
		1997
		1998	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1999	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2000	to run until the system is suspended, but they will not advance while the
		2001	system is suspended. That means, on resume, it will be as if the program
		2002	was frozen for a few seconds, but the suspend time will not be counted
		2003	towards C<ev_timer> when a monotonic clock source is used. The real time
		2004	clock advanced as expected, but if it is used as sole clocksource, then a
		2005	long suspend would be detected as a time jump by libev, and timers would
		2006	be adjusted accordingly.
		2007
		2008	I would not be surprised to see different behaviour in different between
		2009	operating systems, OS versions or even different hardware.
		2010
		2011	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2012	time jump in the monotonic clocks and the realtime clock. If the program
		2013	is suspended for a very long time, and monotonic clock sources are in use,
		2014	then you can expect C<ev_timer>s to expire as the full suspension time
		2015	will be counted towards the timers. When no monotonic clock source is in
		2016	use, then libev will again assume a timejump and adjust accordingly.
		2017
		2018	It might be beneficial for this latter case to call C<ev_suspend>
		2019	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2020	deterministic behaviour in this case (you can do nothing against
		2021	C<SIGSTOP>).
		2022
		2023	=head3 Watcher-Specific Functions and Data Members
		2024
462	=over 4	2025	=over 4
463		2026
464	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2027	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
465		2028
466	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2029	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
467		2030
468	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2031	Configure the timer to trigger after C<after> seconds. If C<repeat>
469	C<0.>, then it will automatically be stopped. If it is positive, then the	2032	is C<0.>, then it will automatically be stopped once the timeout is
470	timer will automatically be configured to trigger again C<repeat> seconds	2033	reached. If it is positive, then the timer will automatically be
471	later, again, and again, until stopped manually.	2034	configured to trigger again C<repeat> seconds later, again, and again,
		2035	until stopped manually.
472		2036
473	The timer itself will do a best-effort at avoiding drift, that is, if you	2037	The timer itself will do a best-effort at avoiding drift, that is, if
474	configure a timer to trigger every 10 seconds, then it will trigger at	2038	you configure a timer to trigger every 10 seconds, then it will normally
475	exactly 10 second intervals. If, however, your program cannot keep up with	2039	trigger at exactly 10 second intervals. If, however, your program cannot
476	the timer (ecause it takes longer than those 10 seconds to do stuff) the	2040	keep up with the timer (because it takes longer than those 10 seconds to
477	timer will not fire more than once per event loop iteration.	2041	do stuff) the timer will not fire more than once per event loop iteration.
478		2042
479	=item ev_timer_again (loop)	2043	=item ev_timer_again (loop, ev_timer *)
480		2044
481	This will act as if the timer timed out and restart it again if it is	2045	This will act as if the timer timed out and restart it again if it is
482	repeating. The exact semantics are:	2046	repeating. The exact semantics are:
483		2047
		2048	If the timer is pending, its pending status is cleared.
		2049
484	If the timer is started but nonrepeating, stop it.	2050	If the timer is started but non-repeating, stop it (as if it timed out).
485		2051
486	If the timer is repeating, either start it if necessary (with the repeat	2052	If the timer is repeating, either start it if necessary (with the
487	value), or reset the running timer to the repeat value.	2053	C<repeat> value), or reset the running timer to the C<repeat> value.
488		2054
489	This sounds a bit complicated, but here is a useful and typical	2055	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
490	example: Imagine you have a tcp connection and you want a so-called idle	2056	usage example.
491	timeout, that is, you want to be called when there have been, say, 60	2057
492	seconds of inactivity on the socket. The easiest way to do this is to	2058	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
493	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	2059
494	time you successfully read or write some data. If you go into an idle	2060	Returns the remaining time until a timer fires. If the timer is active,
495	state where you do not expect data to travel on the socket, you can stop	2061	then this time is relative to the current event loop time, otherwise it's
496	the timer, and again will automatically restart it if need be.	2062	the timeout value currently configured.
		2063
		2064	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2065	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2066	will return C<4>. When the timer expires and is restarted, it will return
		2067	roughly C<7> (likely slightly less as callback invocation takes some time,
		2068	too), and so on.
		2069
		2070	=item ev_tstamp repeat [read-write]
		2071
		2072	The current C<repeat> value. Will be used each time the watcher times out
		2073	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2074	which is also when any modifications are taken into account.
497		2075
498	=back	2076	=back
499		2077
		2078	=head3 Examples
		2079
		2080	Example: Create a timer that fires after 60 seconds.
		2081
		2082	static void
		2083	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2084	{
		2085	.. one minute over, w is actually stopped right here
		2086	}
		2087
		2088	ev_timer mytimer;
		2089	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2090	ev_timer_start (loop, &mytimer);
		2091
		2092	Example: Create a timeout timer that times out after 10 seconds of
		2093	inactivity.
		2094
		2095	static void
		2096	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2097	{
		2098	.. ten seconds without any activity
		2099	}
		2100
		2101	ev_timer mytimer;
		2102	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2103	ev_timer_again (&mytimer); /* start timer */
		2104	ev_run (loop, 0);
		2105
		2106	// and in some piece of code that gets executed on any "activity":
		2107	// reset the timeout to start ticking again at 10 seconds
		2108	ev_timer_again (&mytimer);
		2109
		2110
500	=head2 C<ev_periodic> - to cron or not to cron	2111	=head2 C<ev_periodic> - to cron or not to cron?
501		2112
502	Periodic watchers are also timers of a kind, but they are very versatile	2113	Periodic watchers are also timers of a kind, but they are very versatile
503	(and unfortunately a bit complex).	2114	(and unfortunately a bit complex).
504		2115
505	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2116	Unlike C<ev_timer>, periodic watchers are not based on real time (or
506	but on wallclock time (absolute time). You can tell a periodic watcher	2117	relative time, the physical time that passes) but on wall clock time
507	to trigger "at" some specific point in time. For example, if you tell a	2118	(absolute time, the thing you can read on your calender or clock). The
508	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2119	difference is that wall clock time can run faster or slower than real
509	+ 10.>) and then reset your system clock to the last year, then it will	2120	time, and time jumps are not uncommon (e.g. when you adjust your
510	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2121	wrist-watch).
511	roughly 10 seconds later and of course not if you reset your system time
512	again).
513		2122
514	They can also be used to implement vastly more complex timers, such as	2123	You can tell a periodic watcher to trigger after some specific point
		2124	in time: for example, if you tell a periodic watcher to trigger "in 10
		2125	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2126	not a delay) and then reset your system clock to January of the previous
		2127	year, then it will take a year or more to trigger the event (unlike an
		2128	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2129	it, as it uses a relative timeout).
		2130
		2131	C<ev_periodic> watchers can also be used to implement vastly more complex
515	triggering an event on eahc midnight, local time.	2132	timers, such as triggering an event on each "midnight, local time", or
		2133	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		2134	those cannot react to time jumps.
		2135
		2136	As with timers, the callback is guaranteed to be invoked only when the
		2137	point in time where it is supposed to trigger has passed. If multiple
		2138	timers become ready during the same loop iteration then the ones with
		2139	earlier time-out values are invoked before ones with later time-out values
		2140	(but this is no longer true when a callback calls C<ev_run> recursively).
		2141
		2142	=head3 Watcher-Specific Functions and Data Members
516		2143
517	=over 4	2144	=over 4
518		2145
519	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2146	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
520		2147
521	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2148	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
522		2149
523	Lots of arguments, lets sort it out... There are basically three modes of	2150	Lots of arguments, let's sort it out... There are basically three modes of
524	operation, and we will explain them from simplest to complex:	2151	operation, and we will explain them from simplest to most complex:
525
526		2152
527	=over 4	2153	=over 4
528		2154
529	=item * absolute timer (interval = reschedule_cb = 0)	2155	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
530		2156
531	In this configuration the watcher triggers an event at the wallclock time	2157	In this configuration the watcher triggers an event after the wall clock
532	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2158	time C<offset> has passed. It will not repeat and will not adjust when a
533	that is, if it is to be run at January 1st 2011 then it will run when the	2159	time jump occurs, that is, if it is to be run at January 1st 2011 then it
534	system time reaches or surpasses this time.	2160	will be stopped and invoked when the system clock reaches or surpasses
		2161	this point in time.
535		2162
536	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2163	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
537		2164
538	In this mode the watcher will always be scheduled to time out at the next	2165	In this mode the watcher will always be scheduled to time out at the next
539	C<at + N * interval> time (for some integer N) and then repeat, regardless	2166	C<offset + N * interval> time (for some integer N, which can also be
540	of any time jumps.	2167	negative) and then repeat, regardless of any time jumps. The C<offset>
		2168	argument is merely an offset into the C<interval> periods.
541		2169
542	This can be used to create timers that do not drift with respect to system	2170	This can be used to create timers that do not drift with respect to the
543	time:	2171	system clock, for example, here is an C<ev_periodic> that triggers each
		2172	hour, on the hour (with respect to UTC):
544		2173
545	ev_periodic_set (&periodic, 0., 3600., 0);	2174	ev_periodic_set (&periodic, 0., 3600., 0);
546		2175
547	This doesn't mean there will always be 3600 seconds in between triggers,	2176	This doesn't mean there will always be 3600 seconds in between triggers,
548	but only that the the callback will be called when the system time shows a	2177	but only that the callback will be called when the system time shows a
549	full hour (UTC), or more correctly, when the system time is evenly divisible	2178	full hour (UTC), or more correctly, when the system time is evenly divisible
550	by 3600.	2179	by 3600.
551		2180
552	Another way to think about it (for the mathematically inclined) is that	2181	Another way to think about it (for the mathematically inclined) is that
553	C<ev_periodic> will try to run the callback in this mode at the next possible	2182	C<ev_periodic> will try to run the callback in this mode at the next possible
554	time where C<time = at (mod interval)>, regardless of any time jumps.	2183	time where C<time = offset (mod interval)>, regardless of any time jumps.
555		2184
		2185	For numerical stability it is preferable that the C<offset> value is near
		2186	C<ev_now ()> (the current time), but there is no range requirement for
		2187	this value, and in fact is often specified as zero.
		2188
		2189	Note also that there is an upper limit to how often a timer can fire (CPU
		2190	speed for example), so if C<interval> is very small then timing stability
		2191	will of course deteriorate. Libev itself tries to be exact to be about one
		2192	millisecond (if the OS supports it and the machine is fast enough).
		2193
556	=item * manual reschedule mode (reschedule_cb = callback)	2194	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
557		2195
558	In this mode the values for C<interval> and C<at> are both being	2196	In this mode the values for C<interval> and C<offset> are both being
559	ignored. Instead, each time the periodic watcher gets scheduled, the	2197	ignored. Instead, each time the periodic watcher gets scheduled, the
560	reschedule callback will be called with the watcher as first, and the	2198	reschedule callback will be called with the watcher as first, and the
561	current time as second argument.	2199	current time as second argument.
562		2200
563	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	2201	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
564	periodic watcher, ever, or make any event loop modifications>. If you need	2202	or make ANY other event loop modifications whatsoever, unless explicitly
		2203	allowed by documentation here>.
		2204
565	to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards.	2205	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2206	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2207	only event loop modification you are allowed to do).
566		2208
567	Also, I<< this callback must always return a time that is later than the
568	passed C<now> value >>. Not even C<now> itself will be ok.
569
570	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2209	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
571	ev_tstamp now)>, e.g.:	2210	*w, ev_tstamp now)>, e.g.:
572		2211
		2212	static ev_tstamp
573	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2213	my_rescheduler (ev_periodic *w, ev_tstamp now)
574	{	2214	{
575	return now + 60.;	2215	return now + 60.;
576	}	2216	}
577		2217
578	It must return the next time to trigger, based on the passed time value	2218	It must return the next time to trigger, based on the passed time value
579	(that is, the lowest time value larger than to the second argument). It	2219	(that is, the lowest time value larger than to the second argument). It
580	will usually be called just before the callback will be triggered, but	2220	will usually be called just before the callback will be triggered, but
581	might be called at other times, too.	2221	might be called at other times, too.
582		2222
		2223	NOTE: I<< This callback must always return a time that is higher than or
		2224	equal to the passed C<now> value >>.
		2225
583	This can be used to create very complex timers, such as a timer that	2226	This can be used to create very complex timers, such as a timer that
584	triggers on each midnight, local time. To do this, you would calculate the	2227	triggers on "next midnight, local time". To do this, you would calculate the
585	next midnight after C<now> and return the timestamp value for this. How you do this	2228	next midnight after C<now> and return the timestamp value for this. How
586	is, again, up to you (but it is not trivial).	2229	you do this is, again, up to you (but it is not trivial, which is the main
		2230	reason I omitted it as an example).
587		2231
588	=back	2232	=back
589		2233
590	=item ev_periodic_again (loop, ev_periodic *)	2234	=item ev_periodic_again (loop, ev_periodic *)
591		2235
592	Simply stops and restarts the periodic watcher again. This is only useful	2236	Simply stops and restarts the periodic watcher again. This is only useful
593	when you changed some parameters or the reschedule callback would return	2237	when you changed some parameters or the reschedule callback would return
594	a different time than the last time it was called (e.g. in a crond like	2238	a different time than the last time it was called (e.g. in a crond like
595	program when the crontabs have changed).	2239	program when the crontabs have changed).
596		2240
		2241	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2242
		2243	When active, returns the absolute time that the watcher is supposed
		2244	to trigger next. This is not the same as the C<offset> argument to
		2245	C<ev_periodic_set>, but indeed works even in interval and manual
		2246	rescheduling modes.
		2247
		2248	=item ev_tstamp offset [read-write]
		2249
		2250	When repeating, this contains the offset value, otherwise this is the
		2251	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2252	although libev might modify this value for better numerical stability).
		2253
		2254	Can be modified any time, but changes only take effect when the periodic
		2255	timer fires or C<ev_periodic_again> is being called.
		2256
		2257	=item ev_tstamp interval [read-write]
		2258
		2259	The current interval value. Can be modified any time, but changes only
		2260	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2261	called.
		2262
		2263	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2264
		2265	The current reschedule callback, or C<0>, if this functionality is
		2266	switched off. Can be changed any time, but changes only take effect when
		2267	the periodic timer fires or C<ev_periodic_again> is being called.
		2268
597	=back	2269	=back
598		2270
		2271	=head3 Examples
		2272
		2273	Example: Call a callback every hour, or, more precisely, whenever the
		2274	system time is divisible by 3600. The callback invocation times have
		2275	potentially a lot of jitter, but good long-term stability.
		2276
		2277	static void
		2278	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2279	{
		2280	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2281	}
		2282
		2283	ev_periodic hourly_tick;
		2284	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2285	ev_periodic_start (loop, &hourly_tick);
		2286
		2287	Example: The same as above, but use a reschedule callback to do it:
		2288
		2289	#include <math.h>
		2290
		2291	static ev_tstamp
		2292	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2293	{
		2294	return now + (3600. - fmod (now, 3600.));
		2295	}
		2296
		2297	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2298
		2299	Example: Call a callback every hour, starting now:
		2300
		2301	ev_periodic hourly_tick;
		2302	ev_periodic_init (&hourly_tick, clock_cb,
		2303	fmod (ev_now (loop), 3600.), 3600., 0);
		2304	ev_periodic_start (loop, &hourly_tick);
		2305
		2306
599	=head2 C<ev_signal> - signal me when a signal gets signalled	2307	=head2 C<ev_signal> - signal me when a signal gets signalled!
600		2308
601	Signal watchers will trigger an event when the process receives a specific	2309	Signal watchers will trigger an event when the process receives a specific
602	signal one or more times. Even though signals are very asynchronous, libev	2310	signal one or more times. Even though signals are very asynchronous, libev
603	will try it's best to deliver signals synchronously, i.e. as part of the	2311	will try its best to deliver signals synchronously, i.e. as part of the
604	normal event processing, like any other event.	2312	normal event processing, like any other event.
605		2313
		2314	If you want signals to be delivered truly asynchronously, just use
		2315	C<sigaction> as you would do without libev and forget about sharing
		2316	the signal. You can even use C<ev_async> from a signal handler to
		2317	synchronously wake up an event loop.
		2318
606	You can configure as many watchers as you like per signal. Only when the	2319	You can configure as many watchers as you like for the same signal, but
		2320	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2321	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2322	C<SIGINT> in both the default loop and another loop at the same time. At
		2323	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2324
607	first watcher gets started will libev actually register a signal watcher	2325	When the first watcher gets started will libev actually register something
608	with the kernel (thus it coexists with your own signal handlers as long	2326	with the kernel (thus it coexists with your own signal handlers as long as
609	as you don't register any with libev). Similarly, when the last signal	2327	you don't register any with libev for the same signal).
610	watcher for a signal is stopped libev will reset the signal handler to	2328
611	SIG_DFL (regardless of what it was set to before).	2329	If possible and supported, libev will install its handlers with
		2330	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2331	not be unduly interrupted. If you have a problem with system calls getting
		2332	interrupted by signals you can block all signals in an C<ev_check> watcher
		2333	and unblock them in an C<ev_prepare> watcher.
		2334
		2335	=head3 The special problem of inheritance over fork/execve/pthread_create
		2336
		2337	Both the signal mask (C<sigprocmask>) and the signal disposition
		2338	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2339	stopping it again), that is, libev might or might not block the signal,
		2340	and might or might not set or restore the installed signal handler.
		2341
		2342	While this does not matter for the signal disposition (libev never
		2343	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2344	C<execve>), this matters for the signal mask: many programs do not expect
		2345	certain signals to be blocked.
		2346
		2347	This means that before calling C<exec> (from the child) you should reset
		2348	the signal mask to whatever "default" you expect (all clear is a good
		2349	choice usually).
		2350
		2351	The simplest way to ensure that the signal mask is reset in the child is
		2352	to install a fork handler with C<pthread_atfork> that resets it. That will
		2353	catch fork calls done by libraries (such as the libc) as well.
		2354
		2355	In current versions of libev, the signal will not be blocked indefinitely
		2356	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2357	the window of opportunity for problems, it will not go away, as libev
		2358	I<has> to modify the signal mask, at least temporarily.
		2359
		2360	So I can't stress this enough: I<If you do not reset your signal mask when
		2361	you expect it to be empty, you have a race condition in your code>. This
		2362	is not a libev-specific thing, this is true for most event libraries.
		2363
		2364	=head3 The special problem of threads signal handling
		2365
		2366	POSIX threads has problematic signal handling semantics, specifically,
		2367	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2368	threads in a process block signals, which is hard to achieve.
		2369
		2370	When you want to use sigwait (or mix libev signal handling with your own
		2371	for the same signals), you can tackle this problem by globally blocking
		2372	all signals before creating any threads (or creating them with a fully set
		2373	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2374	loops. Then designate one thread as "signal receiver thread" which handles
		2375	these signals. You can pass on any signals that libev might be interested
		2376	in by calling C<ev_feed_signal>.
		2377
		2378	=head3 Watcher-Specific Functions and Data Members
612		2379
613	=over 4	2380	=over 4
614		2381
615	=item ev_signal_init (ev_signal *, callback, int signum)	2382	=item ev_signal_init (ev_signal *, callback, int signum)
616		2383
617	=item ev_signal_set (ev_signal *, int signum)	2384	=item ev_signal_set (ev_signal *, int signum)
618		2385
619	Configures the watcher to trigger on the given signal number (usually one	2386	Configures the watcher to trigger on the given signal number (usually one
620	of the C<SIGxxx> constants).	2387	of the C<SIGxxx> constants).
621		2388
		2389	=item int signum [read-only]
		2390
		2391	The signal the watcher watches out for.
		2392
622	=back	2393	=back
623		2394
		2395	=head3 Examples
		2396
		2397	Example: Try to exit cleanly on SIGINT.
		2398
		2399	static void
		2400	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2401	{
		2402	ev_break (loop, EVBREAK_ALL);
		2403	}
		2404
		2405	ev_signal signal_watcher;
		2406	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2407	ev_signal_start (loop, &signal_watcher);
		2408
		2409
624	=head2 C<ev_child> - wait for pid status changes	2410	=head2 C<ev_child> - watch out for process status changes
625		2411
626	Child watchers trigger when your process receives a SIGCHLD in response to	2412	Child watchers trigger when your process receives a SIGCHLD in response to
627	some child status changes (most typically when a child of yours dies).	2413	some child status changes (most typically when a child of yours dies or
		2414	exits). It is permissible to install a child watcher I<after> the child
		2415	has been forked (which implies it might have already exited), as long
		2416	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2417	forking and then immediately registering a watcher for the child is fine,
		2418	but forking and registering a watcher a few event loop iterations later or
		2419	in the next callback invocation is not.
		2420
		2421	Only the default event loop is capable of handling signals, and therefore
		2422	you can only register child watchers in the default event loop.
		2423
		2424	Due to some design glitches inside libev, child watchers will always be
		2425	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2426	libev)
		2427
		2428	=head3 Process Interaction
		2429
		2430	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2431	initialised. This is necessary to guarantee proper behaviour even if the
		2432	first child watcher is started after the child exits. The occurrence
		2433	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2434	synchronously as part of the event loop processing. Libev always reaps all
		2435	children, even ones not watched.
		2436
		2437	=head3 Overriding the Built-In Processing
		2438
		2439	Libev offers no special support for overriding the built-in child
		2440	processing, but if your application collides with libev's default child
		2441	handler, you can override it easily by installing your own handler for
		2442	C<SIGCHLD> after initialising the default loop, and making sure the
		2443	default loop never gets destroyed. You are encouraged, however, to use an
		2444	event-based approach to child reaping and thus use libev's support for
		2445	that, so other libev users can use C<ev_child> watchers freely.
		2446
		2447	=head3 Stopping the Child Watcher
		2448
		2449	Currently, the child watcher never gets stopped, even when the
		2450	child terminates, so normally one needs to stop the watcher in the
		2451	callback. Future versions of libev might stop the watcher automatically
		2452	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2453	problem).
		2454
		2455	=head3 Watcher-Specific Functions and Data Members
628		2456
629	=over 4	2457	=over 4
630		2458
631	=item ev_child_init (ev_child *, callback, int pid)	2459	=item ev_child_init (ev_child *, callback, int pid, int trace)
632		2460
633	=item ev_child_set (ev_child *, int pid)	2461	=item ev_child_set (ev_child *, int pid, int trace)
634		2462
635	Configures the watcher to wait for status changes of process C<pid> (or	2463	Configures the watcher to wait for status changes of process C<pid> (or
636	I<any> process if C<pid> is specified as C<0>). The callback can look	2464	I<any> process if C<pid> is specified as C<0>). The callback can look
637	at the C<rstatus> member of the C<ev_child> watcher structure to see	2465	at the C<rstatus> member of the C<ev_child> watcher structure to see
638	the status word (use the macros from C<sys/wait.h> and see your systems	2466	the status word (use the macros from C<sys/wait.h> and see your systems
639	C<waitpid> documentation). The C<rpid> member contains the pid of the	2467	C<waitpid> documentation). The C<rpid> member contains the pid of the
640	process causing the status change.	2468	process causing the status change. C<trace> must be either C<0> (only
		2469	activate the watcher when the process terminates) or C<1> (additionally
		2470	activate the watcher when the process is stopped or continued).
		2471
		2472	=item int pid [read-only]
		2473
		2474	The process id this watcher watches out for, or C<0>, meaning any process id.
		2475
		2476	=item int rpid [read-write]
		2477
		2478	The process id that detected a status change.
		2479
		2480	=item int rstatus [read-write]
		2481
		2482	The process exit/trace status caused by C<rpid> (see your systems
		2483	C<waitpid> and C<sys/wait.h> documentation for details).
641		2484
642	=back	2485	=back
643		2486
		2487	=head3 Examples
		2488
		2489	Example: C<fork()> a new process and install a child handler to wait for
		2490	its completion.
		2491
		2492	ev_child cw;
		2493
		2494	static void
		2495	child_cb (EV_P_ ev_child *w, int revents)
		2496	{
		2497	ev_child_stop (EV_A_ w);
		2498	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2499	}
		2500
		2501	pid_t pid = fork ();
		2502
		2503	if (pid < 0)
		2504	// error
		2505	else if (pid == 0)
		2506	{
		2507	// the forked child executes here
		2508	exit (1);
		2509	}
		2510	else
		2511	{
		2512	ev_child_init (&cw, child_cb, pid, 0);
		2513	ev_child_start (EV_DEFAULT_ &cw);
		2514	}
		2515
		2516
		2517	=head2 C<ev_stat> - did the file attributes just change?
		2518
		2519	This watches a file system path for attribute changes. That is, it calls
		2520	C<stat> on that path in regular intervals (or when the OS says it changed)
		2521	and sees if it changed compared to the last time, invoking the callback if
		2522	it did.
		2523
		2524	The path does not need to exist: changing from "path exists" to "path does
		2525	not exist" is a status change like any other. The condition "path does not
		2526	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2527	C<st_nlink> field being zero (which is otherwise always forced to be at
		2528	least one) and all the other fields of the stat buffer having unspecified
		2529	contents.
		2530
		2531	The path I<must not> end in a slash or contain special components such as
		2532	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2533	your working directory changes, then the behaviour is undefined.
		2534
		2535	Since there is no portable change notification interface available, the
		2536	portable implementation simply calls C<stat(2)> regularly on the path
		2537	to see if it changed somehow. You can specify a recommended polling
		2538	interval for this case. If you specify a polling interval of C<0> (highly
		2539	recommended!) then a I<suitable, unspecified default> value will be used
		2540	(which you can expect to be around five seconds, although this might
		2541	change dynamically). Libev will also impose a minimum interval which is
		2542	currently around C<0.1>, but that's usually overkill.
		2543
		2544	This watcher type is not meant for massive numbers of stat watchers,
		2545	as even with OS-supported change notifications, this can be
		2546	resource-intensive.
		2547
		2548	At the time of this writing, the only OS-specific interface implemented
		2549	is the Linux inotify interface (implementing kqueue support is left as an
		2550	exercise for the reader. Note, however, that the author sees no way of
		2551	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2552
		2553	=head3 ABI Issues (Largefile Support)
		2554
		2555	Libev by default (unless the user overrides this) uses the default
		2556	compilation environment, which means that on systems with large file
		2557	support disabled by default, you get the 32 bit version of the stat
		2558	structure. When using the library from programs that change the ABI to
		2559	use 64 bit file offsets the programs will fail. In that case you have to
		2560	compile libev with the same flags to get binary compatibility. This is
		2561	obviously the case with any flags that change the ABI, but the problem is
		2562	most noticeably displayed with ev_stat and large file support.
		2563
		2564	The solution for this is to lobby your distribution maker to make large
		2565	file interfaces available by default (as e.g. FreeBSD does) and not
		2566	optional. Libev cannot simply switch on large file support because it has
		2567	to exchange stat structures with application programs compiled using the
		2568	default compilation environment.
		2569
		2570	=head3 Inotify and Kqueue
		2571
		2572	When C<inotify (7)> support has been compiled into libev and present at
		2573	runtime, it will be used to speed up change detection where possible. The
		2574	inotify descriptor will be created lazily when the first C<ev_stat>
		2575	watcher is being started.
		2576
		2577	Inotify presence does not change the semantics of C<ev_stat> watchers
		2578	except that changes might be detected earlier, and in some cases, to avoid
		2579	making regular C<stat> calls. Even in the presence of inotify support
		2580	there are many cases where libev has to resort to regular C<stat> polling,
		2581	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2582	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2583	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2584	xfs are fully working) libev usually gets away without polling.
		2585
		2586	There is no support for kqueue, as apparently it cannot be used to
		2587	implement this functionality, due to the requirement of having a file
		2588	descriptor open on the object at all times, and detecting renames, unlinks
		2589	etc. is difficult.
		2590
		2591	=head3 C<stat ()> is a synchronous operation
		2592
		2593	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2594	the process. The exception are C<ev_stat> watchers - those call C<stat
		2595	()>, which is a synchronous operation.
		2596
		2597	For local paths, this usually doesn't matter: unless the system is very
		2598	busy or the intervals between stat's are large, a stat call will be fast,
		2599	as the path data is usually in memory already (except when starting the
		2600	watcher).
		2601
		2602	For networked file systems, calling C<stat ()> can block an indefinite
		2603	time due to network issues, and even under good conditions, a stat call
		2604	often takes multiple milliseconds.
		2605
		2606	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2607	paths, although this is fully supported by libev.
		2608
		2609	=head3 The special problem of stat time resolution
		2610
		2611	The C<stat ()> system call only supports full-second resolution portably,
		2612	and even on systems where the resolution is higher, most file systems
		2613	still only support whole seconds.
		2614
		2615	That means that, if the time is the only thing that changes, you can
		2616	easily miss updates: on the first update, C<ev_stat> detects a change and
		2617	calls your callback, which does something. When there is another update
		2618	within the same second, C<ev_stat> will be unable to detect unless the
		2619	stat data does change in other ways (e.g. file size).
		2620
		2621	The solution to this is to delay acting on a change for slightly more
		2622	than a second (or till slightly after the next full second boundary), using
		2623	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2624	ev_timer_again (loop, w)>).
		2625
		2626	The C<.02> offset is added to work around small timing inconsistencies
		2627	of some operating systems (where the second counter of the current time
		2628	might be be delayed. One such system is the Linux kernel, where a call to
		2629	C<gettimeofday> might return a timestamp with a full second later than
		2630	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2631	update file times then there will be a small window where the kernel uses
		2632	the previous second to update file times but libev might already execute
		2633	the timer callback).
		2634
		2635	=head3 Watcher-Specific Functions and Data Members
		2636
		2637	=over 4
		2638
		2639	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2640
		2641	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2642
		2643	Configures the watcher to wait for status changes of the given
		2644	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2645	be detected and should normally be specified as C<0> to let libev choose
		2646	a suitable value. The memory pointed to by C<path> must point to the same
		2647	path for as long as the watcher is active.
		2648
		2649	The callback will receive an C<EV_STAT> event when a change was detected,
		2650	relative to the attributes at the time the watcher was started (or the
		2651	last change was detected).
		2652
		2653	=item ev_stat_stat (loop, ev_stat *)
		2654
		2655	Updates the stat buffer immediately with new values. If you change the
		2656	watched path in your callback, you could call this function to avoid
		2657	detecting this change (while introducing a race condition if you are not
		2658	the only one changing the path). Can also be useful simply to find out the
		2659	new values.
		2660
		2661	=item ev_statdata attr [read-only]
		2662
		2663	The most-recently detected attributes of the file. Although the type is
		2664	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2665	suitable for your system, but you can only rely on the POSIX-standardised
		2666	members to be present. If the C<st_nlink> member is C<0>, then there was
		2667	some error while C<stat>ing the file.
		2668
		2669	=item ev_statdata prev [read-only]
		2670
		2671	The previous attributes of the file. The callback gets invoked whenever
		2672	C<prev> != C<attr>, or, more precisely, one or more of these members
		2673	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2674	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2675
		2676	=item ev_tstamp interval [read-only]
		2677
		2678	The specified interval.
		2679
		2680	=item const char *path [read-only]
		2681
		2682	The file system path that is being watched.
		2683
		2684	=back
		2685
		2686	=head3 Examples
		2687
		2688	Example: Watch C</etc/passwd> for attribute changes.
		2689
		2690	static void
		2691	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2692	{
		2693	/* /etc/passwd changed in some way */
		2694	if (w->attr.st_nlink)
		2695	{
		2696	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2697	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2698	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2699	}
		2700	else
		2701	/* you shalt not abuse printf for puts */
		2702	puts ("wow, /etc/passwd is not there, expect problems. "
		2703	"if this is windows, they already arrived\n");
		2704	}
		2705
		2706	...
		2707	ev_stat passwd;
		2708
		2709	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2710	ev_stat_start (loop, &passwd);
		2711
		2712	Example: Like above, but additionally use a one-second delay so we do not
		2713	miss updates (however, frequent updates will delay processing, too, so
		2714	one might do the work both on C<ev_stat> callback invocation I<and> on
		2715	C<ev_timer> callback invocation).
		2716
		2717	static ev_stat passwd;
		2718	static ev_timer timer;
		2719
		2720	static void
		2721	timer_cb (EV_P_ ev_timer *w, int revents)
		2722	{
		2723	ev_timer_stop (EV_A_ w);
		2724
		2725	/* now it's one second after the most recent passwd change */
		2726	}
		2727
		2728	static void
		2729	stat_cb (EV_P_ ev_stat *w, int revents)
		2730	{
		2731	/* reset the one-second timer */
		2732	ev_timer_again (EV_A_ &timer);
		2733	}
		2734
		2735	...
		2736	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2737	ev_stat_start (loop, &passwd);
		2738	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2739
		2740
644	=head2 C<ev_idle> - when you've got nothing better to do	2741	=head2 C<ev_idle> - when you've got nothing better to do...
645		2742
646	Idle watchers trigger events when there are no other events are pending	2743	Idle watchers trigger events when no other events of the same or higher
647	(prepare, check and other idle watchers do not count). That is, as long	2744	priority are pending (prepare, check and other idle watchers do not count
648	as your process is busy handling sockets or timeouts (or even signals,	2745	as receiving "events").
649	imagine) it will not be triggered. But when your process is idle all idle	2746
650	watchers are being called again and again, once per event loop iteration -	2747	That is, as long as your process is busy handling sockets or timeouts
		2748	(or even signals, imagine) of the same or higher priority it will not be
		2749	triggered. But when your process is idle (or only lower-priority watchers
		2750	are pending), the idle watchers are being called once per event loop
651	until stopped, that is, or your process receives more events and becomes	2751	iteration - until stopped, that is, or your process receives more events
652	busy.	2752	and becomes busy again with higher priority stuff.
653		2753
654	The most noteworthy effect is that as long as any idle watchers are	2754	The most noteworthy effect is that as long as any idle watchers are
655	active, the process will not block when waiting for new events.	2755	active, the process will not block when waiting for new events.
656		2756
657	Apart from keeping your process non-blocking (which is a useful	2757	Apart from keeping your process non-blocking (which is a useful
658	effect on its own sometimes), idle watchers are a good place to do	2758	effect on its own sometimes), idle watchers are a good place to do
659	"pseudo-background processing", or delay processing stuff to after the	2759	"pseudo-background processing", or delay processing stuff to after the
660	event loop has handled all outstanding events.	2760	event loop has handled all outstanding events.
661		2761
		2762	=head3 Watcher-Specific Functions and Data Members
		2763
662	=over 4	2764	=over 4
663		2765
664	=item ev_idle_init (ev_signal *, callback)	2766	=item ev_idle_init (ev_idle *, callback)
665		2767
666	Initialises and configures the idle watcher - it has no parameters of any	2768	Initialises and configures the idle watcher - it has no parameters of any
667	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2769	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
668	believe me.	2770	believe me.
669		2771
670	=back	2772	=back
671		2773
		2774	=head3 Examples
		2775
		2776	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2777	callback, free it. Also, use no error checking, as usual.
		2778
		2779	static void
		2780	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2781	{
		2782	free (w);
		2783	// now do something you wanted to do when the program has
		2784	// no longer anything immediate to do.
		2785	}
		2786
		2787	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2788	ev_idle_init (idle_watcher, idle_cb);
		2789	ev_idle_start (loop, idle_watcher);
		2790
		2791
672	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2792	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
673		2793
674	Prepare and check watchers are usually (but not always) used in tandem:	2794	Prepare and check watchers are usually (but not always) used in pairs:
675	Prepare watchers get invoked before the process blocks and check watchers	2795	prepare watchers get invoked before the process blocks and check watchers
676	afterwards.	2796	afterwards.
677		2797
		2798	You I<must not> call C<ev_run> or similar functions that enter
		2799	the current event loop from either C<ev_prepare> or C<ev_check>
		2800	watchers. Other loops than the current one are fine, however. The
		2801	rationale behind this is that you do not need to check for recursion in
		2802	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2803	C<ev_check> so if you have one watcher of each kind they will always be
		2804	called in pairs bracketing the blocking call.
		2805
678	Their main purpose is to integrate other event mechanisms into libev. This	2806	Their main purpose is to integrate other event mechanisms into libev and
679	could be used, for example, to track variable changes, implement your own	2807	their use is somewhat advanced. They could be used, for example, to track
680	watchers, integrate net-snmp or a coroutine library and lots more.	2808	variable changes, implement your own watchers, integrate net-snmp or a
		2809	coroutine library and lots more. They are also occasionally useful if
		2810	you cache some data and want to flush it before blocking (for example,
		2811	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2812	watcher).
681		2813
682	This is done by examining in each prepare call which file descriptors need	2814	This is done by examining in each prepare call which file descriptors
683	to be watched by the other library, registering C<ev_io> watchers for	2815	need to be watched by the other library, registering C<ev_io> watchers
684	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2816	for them and starting an C<ev_timer> watcher for any timeouts (many
685	provide just this functionality). Then, in the check watcher you check for	2817	libraries provide exactly this functionality). Then, in the check watcher,
686	any events that occured (by checking the pending status of all watchers	2818	you check for any events that occurred (by checking the pending status
687	and stopping them) and call back into the library. The I/O and timer	2819	of all watchers and stopping them) and call back into the library. The
688	callbacks will never actually be called (but must be valid neverthelles,	2820	I/O and timer callbacks will never actually be called (but must be valid
689	because you never know, you know?).	2821	nevertheless, because you never know, you know?).
690		2822
691	As another example, the Perl Coro module uses these hooks to integrate	2823	As another example, the Perl Coro module uses these hooks to integrate
692	coroutines into libev programs, by yielding to other active coroutines	2824	coroutines into libev programs, by yielding to other active coroutines
693	during each prepare and only letting the process block if no coroutines	2825	during each prepare and only letting the process block if no coroutines
694	are ready to run (its actually more complicated, it only runs coroutines	2826	are ready to run (it's actually more complicated: it only runs coroutines
695	with priority higher than the event loop and one lower priority once,	2827	with priority higher than or equal to the event loop and one coroutine
696	using idle watchers to keep the event loop from blocking if lower-priority	2828	of lower priority, but only once, using idle watchers to keep the event
697	coroutines exist, thus mapping low-priority coroutines to idle/background	2829	loop from blocking if lower-priority coroutines are active, thus mapping
698	tasks).	2830	low-priority coroutines to idle/background tasks).
		2831
		2832	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2833	priority, to ensure that they are being run before any other watchers
		2834	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2835
		2836	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2837	activate ("feed") events into libev. While libev fully supports this, they
		2838	might get executed before other C<ev_check> watchers did their job. As
		2839	C<ev_check> watchers are often used to embed other (non-libev) event
		2840	loops those other event loops might be in an unusable state until their
		2841	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2842	others).
		2843
		2844	=head3 Watcher-Specific Functions and Data Members
699		2845
700	=over 4	2846	=over 4
701		2847
702	=item ev_prepare_init (ev_prepare *, callback)	2848	=item ev_prepare_init (ev_prepare *, callback)
703		2849
704	=item ev_check_init (ev_check *, callback)	2850	=item ev_check_init (ev_check *, callback)
705		2851
706	Initialises and configures the prepare or check watcher - they have no	2852	Initialises and configures the prepare or check watcher - they have no
707	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2853	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
708	macros, but using them is utterly, utterly and completely pointless.	2854	macros, but using them is utterly, utterly, utterly and completely
		2855	pointless.
709		2856
710	=back	2857	=back
711		2858
		2859	=head3 Examples
		2860
		2861	There are a number of principal ways to embed other event loops or modules
		2862	into libev. Here are some ideas on how to include libadns into libev
		2863	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2864	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2865	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2866	Glib event loop).
		2867
		2868	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2869	and in a check watcher, destroy them and call into libadns. What follows
		2870	is pseudo-code only of course. This requires you to either use a low
		2871	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2872	the callbacks for the IO/timeout watchers might not have been called yet.
		2873
		2874	static ev_io iow [nfd];
		2875	static ev_timer tw;
		2876
		2877	static void
		2878	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2879	{
		2880	}
		2881
		2882	// create io watchers for each fd and a timer before blocking
		2883	static void
		2884	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2885	{
		2886	int timeout = 3600000;
		2887	struct pollfd fds [nfd];
		2888	// actual code will need to loop here and realloc etc.
		2889	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2890
		2891	/* the callback is illegal, but won't be called as we stop during check */
		2892	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2893	ev_timer_start (loop, &tw);
		2894
		2895	// create one ev_io per pollfd
		2896	for (int i = 0; i < nfd; ++i)
		2897	{
		2898	ev_io_init (iow + i, io_cb, fds [i].fd,
		2899	((fds [i].events & POLLIN ? EV_READ : 0)
		2900	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2901
		2902	fds [i].revents = 0;
		2903	ev_io_start (loop, iow + i);
		2904	}
		2905	}
		2906
		2907	// stop all watchers after blocking
		2908	static void
		2909	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2910	{
		2911	ev_timer_stop (loop, &tw);
		2912
		2913	for (int i = 0; i < nfd; ++i)
		2914	{
		2915	// set the relevant poll flags
		2916	// could also call adns_processreadable etc. here
		2917	struct pollfd *fd = fds + i;
		2918	int revents = ev_clear_pending (iow + i);
		2919	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2920	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2921
		2922	// now stop the watcher
		2923	ev_io_stop (loop, iow + i);
		2924	}
		2925
		2926	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2927	}
		2928
		2929	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2930	in the prepare watcher and would dispose of the check watcher.
		2931
		2932	Method 3: If the module to be embedded supports explicit event
		2933	notification (libadns does), you can also make use of the actual watcher
		2934	callbacks, and only destroy/create the watchers in the prepare watcher.
		2935
		2936	static void
		2937	timer_cb (EV_P_ ev_timer *w, int revents)
		2938	{
		2939	adns_state ads = (adns_state)w->data;
		2940	update_now (EV_A);
		2941
		2942	adns_processtimeouts (ads, &tv_now);
		2943	}
		2944
		2945	static void
		2946	io_cb (EV_P_ ev_io *w, int revents)
		2947	{
		2948	adns_state ads = (adns_state)w->data;
		2949	update_now (EV_A);
		2950
		2951	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2952	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2953	}
		2954
		2955	// do not ever call adns_afterpoll
		2956
		2957	Method 4: Do not use a prepare or check watcher because the module you
		2958	want to embed is not flexible enough to support it. Instead, you can
		2959	override their poll function. The drawback with this solution is that the
		2960	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2961	this approach, effectively embedding EV as a client into the horrible
		2962	libglib event loop.
		2963
		2964	static gint
		2965	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2966	{
		2967	int got_events = 0;
		2968
		2969	for (n = 0; n < nfds; ++n)
		2970	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2971
		2972	if (timeout >= 0)
		2973	// create/start timer
		2974
		2975	// poll
		2976	ev_run (EV_A_ 0);
		2977
		2978	// stop timer again
		2979	if (timeout >= 0)
		2980	ev_timer_stop (EV_A_ &to);
		2981
		2982	// stop io watchers again - their callbacks should have set
		2983	for (n = 0; n < nfds; ++n)
		2984	ev_io_stop (EV_A_ iow [n]);
		2985
		2986	return got_events;
		2987	}
		2988
		2989
		2990	=head2 C<ev_embed> - when one backend isn't enough...
		2991
		2992	This is a rather advanced watcher type that lets you embed one event loop
		2993	into another (currently only C<ev_io> events are supported in the embedded
		2994	loop, other types of watchers might be handled in a delayed or incorrect
		2995	fashion and must not be used).
		2996
		2997	There are primarily two reasons you would want that: work around bugs and
		2998	prioritise I/O.
		2999
		3000	As an example for a bug workaround, the kqueue backend might only support
		3001	sockets on some platform, so it is unusable as generic backend, but you
		3002	still want to make use of it because you have many sockets and it scales
		3003	so nicely. In this case, you would create a kqueue-based loop and embed
		3004	it into your default loop (which might use e.g. poll). Overall operation
		3005	will be a bit slower because first libev has to call C<poll> and then
		3006	C<kevent>, but at least you can use both mechanisms for what they are
		3007	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		3008
		3009	As for prioritising I/O: under rare circumstances you have the case where
		3010	some fds have to be watched and handled very quickly (with low latency),
		3011	and even priorities and idle watchers might have too much overhead. In
		3012	this case you would put all the high priority stuff in one loop and all
		3013	the rest in a second one, and embed the second one in the first.
		3014
		3015	As long as the watcher is active, the callback will be invoked every
		3016	time there might be events pending in the embedded loop. The callback
		3017	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		3018	sweep and invoke their callbacks (the callback doesn't need to invoke the
		3019	C<ev_embed_sweep> function directly, it could also start an idle watcher
		3020	to give the embedded loop strictly lower priority for example).
		3021
		3022	You can also set the callback to C<0>, in which case the embed watcher
		3023	will automatically execute the embedded loop sweep whenever necessary.
		3024
		3025	Fork detection will be handled transparently while the C<ev_embed> watcher
		3026	is active, i.e., the embedded loop will automatically be forked when the
		3027	embedding loop forks. In other cases, the user is responsible for calling
		3028	C<ev_loop_fork> on the embedded loop.
		3029
		3030	Unfortunately, not all backends are embeddable: only the ones returned by
		3031	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		3032	portable one.
		3033
		3034	So when you want to use this feature you will always have to be prepared
		3035	that you cannot get an embeddable loop. The recommended way to get around
		3036	this is to have a separate variables for your embeddable loop, try to
		3037	create it, and if that fails, use the normal loop for everything.
		3038
		3039	=head3 C<ev_embed> and fork
		3040
		3041	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3042	automatically be applied to the embedded loop as well, so no special
		3043	fork handling is required in that case. When the watcher is not running,
		3044	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3045	as applicable.
		3046
		3047	=head3 Watcher-Specific Functions and Data Members
		3048
		3049	=over 4
		3050
		3051	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		3052
		3053	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		3054
		3055	Configures the watcher to embed the given loop, which must be
		3056	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		3057	invoked automatically, otherwise it is the responsibility of the callback
		3058	to invoke it (it will continue to be called until the sweep has been done,
		3059	if you do not want that, you need to temporarily stop the embed watcher).
		3060
		3061	=item ev_embed_sweep (loop, ev_embed *)
		3062
		3063	Make a single, non-blocking sweep over the embedded loop. This works
		3064	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3065	appropriate way for embedded loops.
		3066
		3067	=item struct ev_loop *other [read-only]
		3068
		3069	The embedded event loop.
		3070
		3071	=back
		3072
		3073	=head3 Examples
		3074
		3075	Example: Try to get an embeddable event loop and embed it into the default
		3076	event loop. If that is not possible, use the default loop. The default
		3077	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3078	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3079	used).
		3080
		3081	struct ev_loop *loop_hi = ev_default_init (0);
		3082	struct ev_loop *loop_lo = 0;
		3083	ev_embed embed;
		3084
		3085	// see if there is a chance of getting one that works
		3086	// (remember that a flags value of 0 means autodetection)
		3087	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3088	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3089	: 0;
		3090
		3091	// if we got one, then embed it, otherwise default to loop_hi
		3092	if (loop_lo)
		3093	{
		3094	ev_embed_init (&embed, 0, loop_lo);
		3095	ev_embed_start (loop_hi, &embed);
		3096	}
		3097	else
		3098	loop_lo = loop_hi;
		3099
		3100	Example: Check if kqueue is available but not recommended and create
		3101	a kqueue backend for use with sockets (which usually work with any
		3102	kqueue implementation). Store the kqueue/socket-only event loop in
		3103	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3104
		3105	struct ev_loop *loop = ev_default_init (0);
		3106	struct ev_loop *loop_socket = 0;
		3107	ev_embed embed;
		3108
		3109	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3110	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3111	{
		3112	ev_embed_init (&embed, 0, loop_socket);
		3113	ev_embed_start (loop, &embed);
		3114	}
		3115
		3116	if (!loop_socket)
		3117	loop_socket = loop;
		3118
		3119	// now use loop_socket for all sockets, and loop for everything else
		3120
		3121
		3122	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3123
		3124	Fork watchers are called when a C<fork ()> was detected (usually because
		3125	whoever is a good citizen cared to tell libev about it by calling
		3126	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		3127	event loop blocks next and before C<ev_check> watchers are being called,
		3128	and only in the child after the fork. If whoever good citizen calling
		3129	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		3130	handlers will be invoked, too, of course.
		3131
		3132	=head3 The special problem of life after fork - how is it possible?
		3133
		3134	Most uses of C<fork()> consist of forking, then some simple calls to set
		3135	up/change the process environment, followed by a call to C<exec()>. This
		3136	sequence should be handled by libev without any problems.
		3137
		3138	This changes when the application actually wants to do event handling
		3139	in the child, or both parent in child, in effect "continuing" after the
		3140	fork.
		3141
		3142	The default mode of operation (for libev, with application help to detect
		3143	forks) is to duplicate all the state in the child, as would be expected
		3144	when I<either> the parent I<or> the child process continues.
		3145
		3146	When both processes want to continue using libev, then this is usually the
		3147	wrong result. In that case, usually one process (typically the parent) is
		3148	supposed to continue with all watchers in place as before, while the other
		3149	process typically wants to start fresh, i.e. without any active watchers.
		3150
		3151	The cleanest and most efficient way to achieve that with libev is to
		3152	simply create a new event loop, which of course will be "empty", and
		3153	use that for new watchers. This has the advantage of not touching more
		3154	memory than necessary, and thus avoiding the copy-on-write, and the
		3155	disadvantage of having to use multiple event loops (which do not support
		3156	signal watchers).
		3157
		3158	When this is not possible, or you want to use the default loop for
		3159	other reasons, then in the process that wants to start "fresh", call
		3160	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3161	Destroying the default loop will "orphan" (not stop) all registered
		3162	watchers, so you have to be careful not to execute code that modifies
		3163	those watchers. Note also that in that case, you have to re-register any
		3164	signal watchers.
		3165
		3166	=head3 Watcher-Specific Functions and Data Members
		3167
		3168	=over 4
		3169
		3170	=item ev_fork_init (ev_fork *, callback)
		3171
		3172	Initialises and configures the fork watcher - it has no parameters of any
		3173	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3174	really.
		3175
		3176	=back
		3177
		3178
		3179	=head2 C<ev_cleanup> - even the best things end
		3180
		3181	Cleanup watchers are called just before the event loop is being destroyed
		3182	by a call to C<ev_loop_destroy>.
		3183
		3184	While there is no guarantee that the event loop gets destroyed, cleanup
		3185	watchers provide a convenient method to install cleanup hooks for your
		3186	program, worker threads and so on - you just to make sure to destroy the
		3187	loop when you want them to be invoked.
		3188
		3189	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3190	all other watchers, they do not keep a reference to the event loop (which
		3191	makes a lot of sense if you think about it). Like all other watchers, you
		3192	can call libev functions in the callback, except C<ev_cleanup_start>.
		3193
		3194	=head3 Watcher-Specific Functions and Data Members
		3195
		3196	=over 4
		3197
		3198	=item ev_cleanup_init (ev_cleanup *, callback)
		3199
		3200	Initialises and configures the cleanup watcher - it has no parameters of
		3201	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3202	pointless, I assure you.
		3203
		3204	=back
		3205
		3206	Example: Register an atexit handler to destroy the default loop, so any
		3207	cleanup functions are called.
		3208
		3209	static void
		3210	program_exits (void)
		3211	{
		3212	ev_loop_destroy (EV_DEFAULT_UC);
		3213	}
		3214
		3215	...
		3216	atexit (program_exits);
		3217
		3218
		3219	=head2 C<ev_async> - how to wake up an event loop
		3220
		3221	In general, you cannot use an C<ev_run> from multiple threads or other
		3222	asynchronous sources such as signal handlers (as opposed to multiple event
		3223	loops - those are of course safe to use in different threads).
		3224
		3225	Sometimes, however, you need to wake up an event loop you do not control,
		3226	for example because it belongs to another thread. This is what C<ev_async>
		3227	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3228	it by calling C<ev_async_send>, which is thread- and signal safe.
		3229
		3230	This functionality is very similar to C<ev_signal> watchers, as signals,
		3231	too, are asynchronous in nature, and signals, too, will be compressed
		3232	(i.e. the number of callback invocations may be less than the number of
		3233	C<ev_async_sent> calls). In fact, you could use signal watchers as a kind
		3234	of "global async watchers" by using a watcher on an otherwise unused
		3235	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3236	even without knowing which loop owns the signal.
		3237
		3238	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3239	just the default loop.
		3240
		3241	=head3 Queueing
		3242
		3243	C<ev_async> does not support queueing of data in any way. The reason
		3244	is that the author does not know of a simple (or any) algorithm for a
		3245	multiple-writer-single-reader queue that works in all cases and doesn't
		3246	need elaborate support such as pthreads or unportable memory access
		3247	semantics.
		3248
		3249	That means that if you want to queue data, you have to provide your own
		3250	queue. But at least I can tell you how to implement locking around your
		3251	queue:
		3252
		3253	=over 4
		3254
		3255	=item queueing from a signal handler context
		3256
		3257	To implement race-free queueing, you simply add to the queue in the signal
		3258	handler but you block the signal handler in the watcher callback. Here is
		3259	an example that does that for some fictitious SIGUSR1 handler:
		3260
		3261	static ev_async mysig;
		3262
		3263	static void
		3264	sigusr1_handler (void)
		3265	{
		3266	sometype data;
		3267
		3268	// no locking etc.
		3269	queue_put (data);
		3270	ev_async_send (EV_DEFAULT_ &mysig);
		3271	}
		3272
		3273	static void
		3274	mysig_cb (EV_P_ ev_async *w, int revents)
		3275	{
		3276	sometype data;
		3277	sigset_t block, prev;
		3278
		3279	sigemptyset (&block);
		3280	sigaddset (&block, SIGUSR1);
		3281	sigprocmask (SIG_BLOCK, &block, &prev);
		3282
		3283	while (queue_get (&data))
		3284	process (data);
		3285
		3286	if (sigismember (&prev, SIGUSR1)
		3287	sigprocmask (SIG_UNBLOCK, &block, 0);
		3288	}
		3289
		3290	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3291	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3292	either...).
		3293
		3294	=item queueing from a thread context
		3295
		3296	The strategy for threads is different, as you cannot (easily) block
		3297	threads but you can easily preempt them, so to queue safely you need to
		3298	employ a traditional mutex lock, such as in this pthread example:
		3299
		3300	static ev_async mysig;
		3301	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3302
		3303	static void
		3304	otherthread (void)
		3305	{
		3306	// only need to lock the actual queueing operation
		3307	pthread_mutex_lock (&mymutex);
		3308	queue_put (data);
		3309	pthread_mutex_unlock (&mymutex);
		3310
		3311	ev_async_send (EV_DEFAULT_ &mysig);
		3312	}
		3313
		3314	static void
		3315	mysig_cb (EV_P_ ev_async *w, int revents)
		3316	{
		3317	pthread_mutex_lock (&mymutex);
		3318
		3319	while (queue_get (&data))
		3320	process (data);
		3321
		3322	pthread_mutex_unlock (&mymutex);
		3323	}
		3324
		3325	=back
		3326
		3327
		3328	=head3 Watcher-Specific Functions and Data Members
		3329
		3330	=over 4
		3331
		3332	=item ev_async_init (ev_async *, callback)
		3333
		3334	Initialises and configures the async watcher - it has no parameters of any
		3335	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3336	trust me.
		3337
		3338	=item ev_async_send (loop, ev_async *)
		3339
		3340	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3341	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3342	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3343	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3344	section below on what exactly this means).
		3345
		3346	Note that, as with other watchers in libev, multiple events might get
		3347	compressed into a single callback invocation (another way to look at this
		3348	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3349	reset when the event loop detects that).
		3350
		3351	This call incurs the overhead of a system call only once per event loop
		3352	iteration, so while the overhead might be noticeable, it doesn't apply to
		3353	repeated calls to C<ev_async_send> for the same event loop.
		3354
		3355	=item bool = ev_async_pending (ev_async *)
		3356
		3357	Returns a non-zero value when C<ev_async_send> has been called on the
		3358	watcher but the event has not yet been processed (or even noted) by the
		3359	event loop.
		3360
		3361	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3362	the loop iterates next and checks for the watcher to have become active,
		3363	it will reset the flag again. C<ev_async_pending> can be used to very
		3364	quickly check whether invoking the loop might be a good idea.
		3365
		3366	Not that this does I<not> check whether the watcher itself is pending,
		3367	only whether it has been requested to make this watcher pending: there
		3368	is a time window between the event loop checking and resetting the async
		3369	notification, and the callback being invoked.
		3370
		3371	=back
		3372
		3373
712	=head1 OTHER FUNCTIONS	3374	=head1 OTHER FUNCTIONS
713		3375
714	There are some other functions of possible interest. Described. Here. Now.	3376	There are some other functions of possible interest. Described. Here. Now.
715		3377
716	=over 4	3378	=over 4
717		3379
718	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3380	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
719		3381
720	This function combines a simple timer and an I/O watcher, calls your	3382	This function combines a simple timer and an I/O watcher, calls your
721	callback on whichever event happens first and automatically stop both	3383	callback on whichever event happens first and automatically stops both
722	watchers. This is useful if you want to wait for a single event on an fd	3384	watchers. This is useful if you want to wait for a single event on an fd
723	or timeout without havign to allocate/configure/start/stop/free one or	3385	or timeout without having to allocate/configure/start/stop/free one or
724	more watchers yourself.	3386	more watchers yourself.
725		3387
726	If C<fd> is less than 0, then no I/O watcher will be started and events	3388	If C<fd> is less than 0, then no I/O watcher will be started and the
727	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3389	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
728	C<events> set will be craeted and started.	3390	the given C<fd> and C<events> set will be created and started.
729		3391
730	If C<timeout> is less than 0, then no timeout watcher will be	3392	If C<timeout> is less than 0, then no timeout watcher will be
731	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3393	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
732	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3394	repeat = 0) will be started. C<0> is a valid timeout.
733	dubious value.
734		3395
735	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3396	The callback has the type C<void (*cb)(int revents, void *arg)> and is
736	passed an events set like normal event callbacks (with a combination of	3397	passed an C<revents> set like normal event callbacks (a combination of
737	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3398	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
738	value passed to C<ev_once>:	3399	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3400	a timeout and an io event at the same time - you probably should give io
		3401	events precedence.
739		3402
		3403	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3404
740	static void stdin_ready (int revents, void *arg)	3405	static void stdin_ready (int revents, void *arg)
741	{	3406	{
742	if (revents & EV_TIMEOUT)
743	/* doh, nothing entered */;
744	else if (revents & EV_READ)	3407	if (revents & EV_READ)
745	/* stdin might have data for us, joy! */;	3408	/* stdin might have data for us, joy! */;
		3409	else if (revents & EV_TIMER)
		3410	/* doh, nothing entered */;
746	}	3411	}
747		3412
748	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3413	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
749
750	=item ev_feed_event (loop, watcher, int events)
751
752	Feeds the given event set into the event loop, as if the specified event
753	had happened for the specified watcher (which must be a pointer to an
754	initialised but not necessarily started event watcher).
755		3414
756	=item ev_feed_fd_event (loop, int fd, int revents)	3415	=item ev_feed_fd_event (loop, int fd, int revents)
757		3416
758	Feed an event on the given fd, as if a file descriptor backend detected	3417	Feed an event on the given fd, as if a file descriptor backend detected
759	the given events it.	3418	the given events it.
760		3419
761	=item ev_feed_signal_event (loop, int signum)	3420	=item ev_feed_signal_event (loop, int signum)
762		3421
763	Feed an event as if the given signal occured (loop must be the default loop!).	3422	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3423	which is async-safe.
764		3424
765	=back	3425	=back
766		3426
		3427
		3428	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3429
		3430	This section explains some common idioms that are not immediately
		3431	obvious. Note that examples are sprinkled over the whole manual, and this
		3432	section only contains stuff that wouldn't fit anywhere else.
		3433
		3434	=over 4
		3435
		3436	=item Model/nested event loop invocations and exit conditions.
		3437
		3438	Often (especially in GUI toolkits) there are places where you have
		3439	I<modal> interaction, which is most easily implemented by recursively
		3440	invoking C<ev_run>.
		3441
		3442	This brings the problem of exiting - a callback might want to finish the
		3443	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3444	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3445	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3446	other combination: In these cases, C<ev_break> will not work alone.
		3447
		3448	The solution is to maintain "break this loop" variable for each C<ev_run>
		3449	invocation, and use a loop around C<ev_run> until the condition is
		3450	triggered, using C<EVRUN_ONCE>:
		3451
		3452	// main loop
		3453	int exit_main_loop = 0;
		3454
		3455	while (!exit_main_loop)
		3456	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3457
		3458	// in a model watcher
		3459	int exit_nested_loop = 0;
		3460
		3461	while (!exit_nested_loop)
		3462	ev_run (EV_A_ EVRUN_ONCE);
		3463
		3464	To exit from any of these loops, just set the corresponding exit variable:
		3465
		3466	// exit modal loop
		3467	exit_nested_loop = 1;
		3468
		3469	// exit main program, after modal loop is finished
		3470	exit_main_loop = 1;
		3471
		3472	// exit both
		3473	exit_main_loop = exit_nested_loop = 1;
		3474
		3475	=back
		3476
		3477
		3478	=head1 LIBEVENT EMULATION
		3479
		3480	Libev offers a compatibility emulation layer for libevent. It cannot
		3481	emulate the internals of libevent, so here are some usage hints:
		3482
		3483	=over 4
		3484
		3485	=item * Only the libevent-1.4.1-beta API is being emulated.
		3486
		3487	This was the newest libevent version available when libev was implemented,
		3488	and is still mostly unchanged in 2010.
		3489
		3490	=item * Use it by including <event.h>, as usual.
		3491
		3492	=item * The following members are fully supported: ev_base, ev_callback,
		3493	ev_arg, ev_fd, ev_res, ev_events.
		3494
		3495	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		3496	maintained by libev, it does not work exactly the same way as in libevent (consider
		3497	it a private API).
		3498
		3499	=item * Priorities are not currently supported. Initialising priorities
		3500	will fail and all watchers will have the same priority, even though there
		3501	is an ev_pri field.
		3502
		3503	=item * In libevent, the last base created gets the signals, in libev, the
		3504	base that registered the signal gets the signals.
		3505
		3506	=item * Other members are not supported.
		3507
		3508	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		3509	to use the libev header file and library.
		3510
		3511	=back
		3512
		3513	=head1 C++ SUPPORT
		3514
		3515	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3516	you to use some convenience methods to start/stop watchers and also change
		3517	the callback model to a model using method callbacks on objects.
		3518
		3519	To use it,
		3520
		3521	#include <ev++.h>
		3522
		3523	This automatically includes F<ev.h> and puts all of its definitions (many
		3524	of them macros) into the global namespace. All C++ specific things are
		3525	put into the C<ev> namespace. It should support all the same embedding
		3526	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3527
		3528	Care has been taken to keep the overhead low. The only data member the C++
		3529	classes add (compared to plain C-style watchers) is the event loop pointer
		3530	that the watcher is associated with (or no additional members at all if
		3531	you disable C<EV_MULTIPLICITY> when embedding libev).
		3532
		3533	Currently, functions, static and non-static member functions and classes
		3534	with C<operator ()> can be used as callbacks. Other types should be easy
		3535	to add as long as they only need one additional pointer for context. If
		3536	you need support for other types of functors please contact the author
		3537	(preferably after implementing it).
		3538
		3539	Here is a list of things available in the C<ev> namespace:
		3540
		3541	=over 4
		3542
		3543	=item C<ev::READ>, C<ev::WRITE> etc.
		3544
		3545	These are just enum values with the same values as the C<EV_READ> etc.
		3546	macros from F<ev.h>.
		3547
		3548	=item C<ev::tstamp>, C<ev::now>
		3549
		3550	Aliases to the same types/functions as with the C<ev_> prefix.
		3551
		3552	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3553
		3554	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3555	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3556	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3557	defines by many implementations.
		3558
		3559	All of those classes have these methods:
		3560
		3561	=over 4
		3562
		3563	=item ev::TYPE::TYPE ()
		3564
		3565	=item ev::TYPE::TYPE (loop)
		3566
		3567	=item ev::TYPE::~TYPE
		3568
		3569	The constructor (optionally) takes an event loop to associate the watcher
		3570	with. If it is omitted, it will use C<EV_DEFAULT>.
		3571
		3572	The constructor calls C<ev_init> for you, which means you have to call the
		3573	C<set> method before starting it.
		3574
		3575	It will not set a callback, however: You have to call the templated C<set>
		3576	method to set a callback before you can start the watcher.
		3577
		3578	(The reason why you have to use a method is a limitation in C++ which does
		3579	not allow explicit template arguments for constructors).
		3580
		3581	The destructor automatically stops the watcher if it is active.
		3582
		3583	=item w->set<class, &class::method> (object *)
		3584
		3585	This method sets the callback method to call. The method has to have a
		3586	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3587	first argument and the C<revents> as second. The object must be given as
		3588	parameter and is stored in the C<data> member of the watcher.
		3589
		3590	This method synthesizes efficient thunking code to call your method from
		3591	the C callback that libev requires. If your compiler can inline your
		3592	callback (i.e. it is visible to it at the place of the C<set> call and
		3593	your compiler is good :), then the method will be fully inlined into the
		3594	thunking function, making it as fast as a direct C callback.
		3595
		3596	Example: simple class declaration and watcher initialisation
		3597
		3598	struct myclass
		3599	{
		3600	void io_cb (ev::io &w, int revents) { }
		3601	}
		3602
		3603	myclass obj;
		3604	ev::io iow;
		3605	iow.set <myclass, &myclass::io_cb> (&obj);
		3606
		3607	=item w->set (object *)
		3608
		3609	This is a variation of a method callback - leaving out the method to call
		3610	will default the method to C<operator ()>, which makes it possible to use
		3611	functor objects without having to manually specify the C<operator ()> all
		3612	the time. Incidentally, you can then also leave out the template argument
		3613	list.
		3614
		3615	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3616	int revents)>.
		3617
		3618	See the method-C<set> above for more details.
		3619
		3620	Example: use a functor object as callback.
		3621
		3622	struct myfunctor
		3623	{
		3624	void operator() (ev::io &w, int revents)
		3625	{
		3626	...
		3627	}
		3628	}
		3629
		3630	myfunctor f;
		3631
		3632	ev::io w;
		3633	w.set (&f);
		3634
		3635	=item w->set<function> (void *data = 0)
		3636
		3637	Also sets a callback, but uses a static method or plain function as
		3638	callback. The optional C<data> argument will be stored in the watcher's
		3639	C<data> member and is free for you to use.
		3640
		3641	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3642
		3643	See the method-C<set> above for more details.
		3644
		3645	Example: Use a plain function as callback.
		3646
		3647	static void io_cb (ev::io &w, int revents) { }
		3648	iow.set <io_cb> ();
		3649
		3650	=item w->set (loop)
		3651
		3652	Associates a different C<struct ev_loop> with this watcher. You can only
		3653	do this when the watcher is inactive (and not pending either).
		3654
		3655	=item w->set ([arguments])
		3656
		3657	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
		3658	method or a suitable start method must be called at least once. Unlike the
		3659	C counterpart, an active watcher gets automatically stopped and restarted
		3660	when reconfiguring it with this method.
		3661
		3662	=item w->start ()
		3663
		3664	Starts the watcher. Note that there is no C<loop> argument, as the
		3665	constructor already stores the event loop.
		3666
		3667	=item w->start ([arguments])
		3668
		3669	Instead of calling C<set> and C<start> methods separately, it is often
		3670	convenient to wrap them in one call. Uses the same type of arguments as
		3671	the configure C<set> method of the watcher.
		3672
		3673	=item w->stop ()
		3674
		3675	Stops the watcher if it is active. Again, no C<loop> argument.
		3676
		3677	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3678
		3679	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3680	C<ev_TYPE_again> function.
		3681
		3682	=item w->sweep () (C<ev::embed> only)
		3683
		3684	Invokes C<ev_embed_sweep>.
		3685
		3686	=item w->update () (C<ev::stat> only)
		3687
		3688	Invokes C<ev_stat_stat>.
		3689
		3690	=back
		3691
		3692	=back
		3693
		3694	Example: Define a class with two I/O and idle watchers, start the I/O
		3695	watchers in the constructor.
		3696
		3697	class myclass
		3698	{
		3699	ev::io io ; void io_cb (ev::io &w, int revents);
		3700	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
		3701	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3702
		3703	myclass (int fd)
		3704	{
		3705	io .set <myclass, &myclass::io_cb > (this);
		3706	io2 .set <myclass, &myclass::io2_cb > (this);
		3707	idle.set <myclass, &myclass::idle_cb> (this);
		3708
		3709	io.set (fd, ev::WRITE); // configure the watcher
		3710	io.start (); // start it whenever convenient
		3711
		3712	io2.start (fd, ev::READ); // set + start in one call
		3713	}
		3714	};
		3715
		3716
		3717	=head1 OTHER LANGUAGE BINDINGS
		3718
		3719	Libev does not offer other language bindings itself, but bindings for a
		3720	number of languages exist in the form of third-party packages. If you know
		3721	any interesting language binding in addition to the ones listed here, drop
		3722	me a note.
		3723
		3724	=over 4
		3725
		3726	=item Perl
		3727
		3728	The EV module implements the full libev API and is actually used to test
		3729	libev. EV is developed together with libev. Apart from the EV core module,
		3730	there are additional modules that implement libev-compatible interfaces
		3731	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3732	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3733	and C<EV::Glib>).
		3734
		3735	It can be found and installed via CPAN, its homepage is at
		3736	L<http://software.schmorp.de/pkg/EV>.
		3737
		3738	=item Python
		3739
		3740	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3741	seems to be quite complete and well-documented.
		3742
		3743	=item Ruby
		3744
		3745	Tony Arcieri has written a ruby extension that offers access to a subset
		3746	of the libev API and adds file handle abstractions, asynchronous DNS and
		3747	more on top of it. It can be found via gem servers. Its homepage is at
		3748	L<http://rev.rubyforge.org/>.
		3749
		3750	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3751	makes rev work even on mingw.
		3752
		3753	=item Haskell
		3754
		3755	A haskell binding to libev is available at
		3756	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3757
		3758	=item D
		3759
		3760	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3761	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3762
		3763	=item Ocaml
		3764
		3765	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3766	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3767
		3768	=item Lua
		3769
		3770	Brian Maher has written a partial interface to libev for lua (at the
		3771	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3772	L<http://github.com/brimworks/lua-ev>.
		3773
		3774	=back
		3775
		3776
		3777	=head1 MACRO MAGIC
		3778
		3779	Libev can be compiled with a variety of options, the most fundamental
		3780	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3781	functions and callbacks have an initial C<struct ev_loop *> argument.
		3782
		3783	To make it easier to write programs that cope with either variant, the
		3784	following macros are defined:
		3785
		3786	=over 4
		3787
		3788	=item C<EV_A>, C<EV_A_>
		3789
		3790	This provides the loop I<argument> for functions, if one is required ("ev
		3791	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3792	C<EV_A_> is used when other arguments are following. Example:
		3793
		3794	ev_unref (EV_A);
		3795	ev_timer_add (EV_A_ watcher);
		3796	ev_run (EV_A_ 0);
		3797
		3798	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3799	which is often provided by the following macro.
		3800
		3801	=item C<EV_P>, C<EV_P_>
		3802
		3803	This provides the loop I<parameter> for functions, if one is required ("ev
		3804	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3805	C<EV_P_> is used when other parameters are following. Example:
		3806
		3807	// this is how ev_unref is being declared
		3808	static void ev_unref (EV_P);
		3809
		3810	// this is how you can declare your typical callback
		3811	static void cb (EV_P_ ev_timer *w, int revents)
		3812
		3813	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3814	suitable for use with C<EV_A>.
		3815
		3816	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3817
		3818	Similar to the other two macros, this gives you the value of the default
		3819	loop, if multiple loops are supported ("ev loop default").
		3820
		3821	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3822
		3823	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3824	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3825	is undefined when the default loop has not been initialised by a previous
		3826	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3827
		3828	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3829	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3830
		3831	=back
		3832
		3833	Example: Declare and initialise a check watcher, utilising the above
		3834	macros so it will work regardless of whether multiple loops are supported
		3835	or not.
		3836
		3837	static void
		3838	check_cb (EV_P_ ev_timer *w, int revents)
		3839	{
		3840	ev_check_stop (EV_A_ w);
		3841	}
		3842
		3843	ev_check check;
		3844	ev_check_init (&check, check_cb);
		3845	ev_check_start (EV_DEFAULT_ &check);
		3846	ev_run (EV_DEFAULT_ 0);
		3847
		3848	=head1 EMBEDDING
		3849
		3850	Libev can (and often is) directly embedded into host
		3851	applications. Examples of applications that embed it include the Deliantra
		3852	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3853	and rxvt-unicode.
		3854
		3855	The goal is to enable you to just copy the necessary files into your
		3856	source directory without having to change even a single line in them, so
		3857	you can easily upgrade by simply copying (or having a checked-out copy of
		3858	libev somewhere in your source tree).
		3859
		3860	=head2 FILESETS
		3861
		3862	Depending on what features you need you need to include one or more sets of files
		3863	in your application.
		3864
		3865	=head3 CORE EVENT LOOP
		3866
		3867	To include only the libev core (all the C<ev_*> functions), with manual
		3868	configuration (no autoconf):
		3869
		3870	#define EV_STANDALONE 1
		3871	#include "ev.c"
		3872
		3873	This will automatically include F<ev.h>, too, and should be done in a
		3874	single C source file only to provide the function implementations. To use
		3875	it, do the same for F<ev.h> in all files wishing to use this API (best
		3876	done by writing a wrapper around F<ev.h> that you can include instead and
		3877	where you can put other configuration options):
		3878
		3879	#define EV_STANDALONE 1
		3880	#include "ev.h"
		3881
		3882	Both header files and implementation files can be compiled with a C++
		3883	compiler (at least, that's a stated goal, and breakage will be treated
		3884	as a bug).
		3885
		3886	You need the following files in your source tree, or in a directory
		3887	in your include path (e.g. in libev/ when using -Ilibev):
		3888
		3889	ev.h
		3890	ev.c
		3891	ev_vars.h
		3892	ev_wrap.h
		3893
		3894	ev_win32.c required on win32 platforms only
		3895
		3896	ev_select.c only when select backend is enabled (which is enabled by default)
		3897	ev_poll.c only when poll backend is enabled (disabled by default)
		3898	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3899	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3900	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3901
		3902	F<ev.c> includes the backend files directly when enabled, so you only need
		3903	to compile this single file.
		3904
		3905	=head3 LIBEVENT COMPATIBILITY API
		3906
		3907	To include the libevent compatibility API, also include:
		3908
		3909	#include "event.c"
		3910
		3911	in the file including F<ev.c>, and:
		3912
		3913	#include "event.h"
		3914
		3915	in the files that want to use the libevent API. This also includes F<ev.h>.
		3916
		3917	You need the following additional files for this:
		3918
		3919	event.h
		3920	event.c
		3921
		3922	=head3 AUTOCONF SUPPORT
		3923
		3924	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3925	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3926	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3927	include F<config.h> and configure itself accordingly.
		3928
		3929	For this of course you need the m4 file:
		3930
		3931	libev.m4
		3932
		3933	=head2 PREPROCESSOR SYMBOLS/MACROS
		3934
		3935	Libev can be configured via a variety of preprocessor symbols you have to
		3936	define before including (or compiling) any of its files. The default in
		3937	the absence of autoconf is documented for every option.
		3938
		3939	Symbols marked with "(h)" do not change the ABI, and can have different
		3940	values when compiling libev vs. including F<ev.h>, so it is permissible
		3941	to redefine them before including F<ev.h> without breaking compatibility
		3942	to a compiled library. All other symbols change the ABI, which means all
		3943	users of libev and the libev code itself must be compiled with compatible
		3944	settings.
		3945
		3946	=over 4
		3947
		3948	=item EV_COMPAT3 (h)
		3949
		3950	Backwards compatibility is a major concern for libev. This is why this
		3951	release of libev comes with wrappers for the functions and symbols that
		3952	have been renamed between libev version 3 and 4.
		3953
		3954	You can disable these wrappers (to test compatibility with future
		3955	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		3956	sources. This has the additional advantage that you can drop the C<struct>
		3957	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		3958	typedef in that case.
		3959
		3960	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		3961	and in some even more future version the compatibility code will be
		3962	removed completely.
		3963
		3964	=item EV_STANDALONE (h)
		3965
		3966	Must always be C<1> if you do not use autoconf configuration, which
		3967	keeps libev from including F<config.h>, and it also defines dummy
		3968	implementations for some libevent functions (such as logging, which is not
		3969	supported). It will also not define any of the structs usually found in
		3970	F<event.h> that are not directly supported by the libev core alone.
		3971
		3972	In standalone mode, libev will still try to automatically deduce the
		3973	configuration, but has to be more conservative.
		3974
		3975	=item EV_USE_MONOTONIC
		3976
		3977	If defined to be C<1>, libev will try to detect the availability of the
		3978	monotonic clock option at both compile time and runtime. Otherwise no
		3979	use of the monotonic clock option will be attempted. If you enable this,
		3980	you usually have to link against librt or something similar. Enabling it
		3981	when the functionality isn't available is safe, though, although you have
		3982	to make sure you link against any libraries where the C<clock_gettime>
		3983	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3984
		3985	=item EV_USE_REALTIME
		3986
		3987	If defined to be C<1>, libev will try to detect the availability of the
		3988	real-time clock option at compile time (and assume its availability
		3989	at runtime if successful). Otherwise no use of the real-time clock
		3990	option will be attempted. This effectively replaces C<gettimeofday>
		3991	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3992	correctness. See the note about libraries in the description of
		3993	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3994	C<EV_USE_CLOCK_SYSCALL>.
		3995
		3996	=item EV_USE_CLOCK_SYSCALL
		3997
		3998	If defined to be C<1>, libev will try to use a direct syscall instead
		3999	of calling the system-provided C<clock_gettime> function. This option
		4000	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4001	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4002	programs needlessly. Using a direct syscall is slightly slower (in
		4003	theory), because no optimised vdso implementation can be used, but avoids
		4004	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4005	higher, as it simplifies linking (no need for C<-lrt>).
		4006
		4007	=item EV_USE_NANOSLEEP
		4008
		4009	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4010	and will use it for delays. Otherwise it will use C<select ()>.
		4011
		4012	=item EV_USE_EVENTFD
		4013
		4014	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4015	available and will probe for kernel support at runtime. This will improve
		4016	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4017	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4018	2.7 or newer, otherwise disabled.
		4019
		4020	=item EV_USE_SELECT
		4021
		4022	If undefined or defined to be C<1>, libev will compile in support for the
		4023	C<select>(2) backend. No attempt at auto-detection will be done: if no
		4024	other method takes over, select will be it. Otherwise the select backend
		4025	will not be compiled in.
		4026
		4027	=item EV_SELECT_USE_FD_SET
		4028
		4029	If defined to C<1>, then the select backend will use the system C<fd_set>
		4030	structure. This is useful if libev doesn't compile due to a missing
		4031	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		4032	on exotic systems. This usually limits the range of file descriptors to
		4033	some low limit such as 1024 or might have other limitations (winsocket
		4034	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		4035	configures the maximum size of the C<fd_set>.
		4036
		4037	=item EV_SELECT_IS_WINSOCKET
		4038
		4039	When defined to C<1>, the select backend will assume that
		4040	select/socket/connect etc. don't understand file descriptors but
		4041	wants osf handles on win32 (this is the case when the select to
		4042	be used is the winsock select). This means that it will call
		4043	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		4044	it is assumed that all these functions actually work on fds, even
		4045	on win32. Should not be defined on non-win32 platforms.
		4046
		4047	=item EV_FD_TO_WIN32_HANDLE(fd)
		4048
		4049	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4050	file descriptors to socket handles. When not defining this symbol (the
		4051	default), then libev will call C<_get_osfhandle>, which is usually
		4052	correct. In some cases, programs use their own file descriptor management,
		4053	in which case they can provide this function to map fds to socket handles.
		4054
		4055	=item EV_WIN32_HANDLE_TO_FD(handle)
		4056
		4057	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4058	using the standard C<_open_osfhandle> function. For programs implementing
		4059	their own fd to handle mapping, overwriting this function makes it easier
		4060	to do so. This can be done by defining this macro to an appropriate value.
		4061
		4062	=item EV_WIN32_CLOSE_FD(fd)
		4063
		4064	If programs implement their own fd to handle mapping on win32, then this
		4065	macro can be used to override the C<close> function, useful to unregister
		4066	file descriptors again. Note that the replacement function has to close
		4067	the underlying OS handle.
		4068
		4069	=item EV_USE_POLL
		4070
		4071	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		4072	backend. Otherwise it will be enabled on non-win32 platforms. It
		4073	takes precedence over select.
		4074
		4075	=item EV_USE_EPOLL
		4076
		4077	If defined to be C<1>, libev will compile in support for the Linux
		4078	C<epoll>(7) backend. Its availability will be detected at runtime,
		4079	otherwise another method will be used as fallback. This is the preferred
		4080	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4081	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4082
		4083	=item EV_USE_KQUEUE
		4084
		4085	If defined to be C<1>, libev will compile in support for the BSD style
		4086	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		4087	otherwise another method will be used as fallback. This is the preferred
		4088	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		4089	supports some types of fds correctly (the only platform we found that
		4090	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		4091	not be used unless explicitly requested. The best way to use it is to find
		4092	out whether kqueue supports your type of fd properly and use an embedded
		4093	kqueue loop.
		4094
		4095	=item EV_USE_PORT
		4096
		4097	If defined to be C<1>, libev will compile in support for the Solaris
		4098	10 port style backend. Its availability will be detected at runtime,
		4099	otherwise another method will be used as fallback. This is the preferred
		4100	backend for Solaris 10 systems.
		4101
		4102	=item EV_USE_DEVPOLL
		4103
		4104	Reserved for future expansion, works like the USE symbols above.
		4105
		4106	=item EV_USE_INOTIFY
		4107
		4108	If defined to be C<1>, libev will compile in support for the Linux inotify
		4109	interface to speed up C<ev_stat> watchers. Its actual availability will
		4110	be detected at runtime. If undefined, it will be enabled if the headers
		4111	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4112
		4113	=item EV_ATOMIC_T
		4114
		4115	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4116	access is atomic with respect to other threads or signal contexts. No such
		4117	type is easily found in the C language, so you can provide your own type
		4118	that you know is safe for your purposes. It is used both for signal handler "locking"
		4119	as well as for signal and thread safety in C<ev_async> watchers.
		4120
		4121	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4122	(from F<signal.h>), which is usually good enough on most platforms.
		4123
		4124	=item EV_H (h)
		4125
		4126	The name of the F<ev.h> header file used to include it. The default if
		4127	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4128	used to virtually rename the F<ev.h> header file in case of conflicts.
		4129
		4130	=item EV_CONFIG_H (h)
		4131
		4132	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4133	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4134	C<EV_H>, above.
		4135
		4136	=item EV_EVENT_H (h)
		4137
		4138	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4139	of how the F<event.h> header can be found, the default is C<"event.h">.
		4140
		4141	=item EV_PROTOTYPES (h)
		4142
		4143	If defined to be C<0>, then F<ev.h> will not define any function
		4144	prototypes, but still define all the structs and other symbols. This is
		4145	occasionally useful if you want to provide your own wrapper functions
		4146	around libev functions.
		4147
		4148	=item EV_MULTIPLICITY
		4149
		4150	If undefined or defined to C<1>, then all event-loop-specific functions
		4151	will have the C<struct ev_loop *> as first argument, and you can create
		4152	additional independent event loops. Otherwise there will be no support
		4153	for multiple event loops and there is no first event loop pointer
		4154	argument. Instead, all functions act on the single default loop.
		4155
		4156	=item EV_MINPRI
		4157
		4158	=item EV_MAXPRI
		4159
		4160	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4161	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4162	provide for more priorities by overriding those symbols (usually defined
		4163	to be C<-2> and C<2>, respectively).
		4164
		4165	When doing priority-based operations, libev usually has to linearly search
		4166	all the priorities, so having many of them (hundreds) uses a lot of space
		4167	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4168	fine.
		4169
		4170	If your embedding application does not need any priorities, defining these
		4171	both to C<0> will save some memory and CPU.
		4172
		4173	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4174	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4175	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4176
		4177	If undefined or defined to be C<1> (and the platform supports it), then
		4178	the respective watcher type is supported. If defined to be C<0>, then it
		4179	is not. Disabling watcher types mainly saves code size.
		4180
		4181	=item EV_FEATURES
		4182
		4183	If you need to shave off some kilobytes of code at the expense of some
		4184	speed (but with the full API), you can define this symbol to request
		4185	certain subsets of functionality. The default is to enable all features
		4186	that can be enabled on the platform.
		4187
		4188	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4189	with some broad features you want) and then selectively re-enable
		4190	additional parts you want, for example if you want everything minimal,
		4191	but multiple event loop support, async and child watchers and the poll
		4192	backend, use this:
		4193
		4194	#define EV_FEATURES 0
		4195	#define EV_MULTIPLICITY 1
		4196	#define EV_USE_POLL 1
		4197	#define EV_CHILD_ENABLE 1
		4198	#define EV_ASYNC_ENABLE 1
		4199
		4200	The actual value is a bitset, it can be a combination of the following
		4201	values:
		4202
		4203	=over 4
		4204
		4205	=item C<1> - faster/larger code
		4206
		4207	Use larger code to speed up some operations.
		4208
		4209	Currently this is used to override some inlining decisions (enlarging the
		4210	code size by roughly 30% on amd64).
		4211
		4212	When optimising for size, use of compiler flags such as C<-Os> with
		4213	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4214	assertions.
		4215
		4216	=item C<2> - faster/larger data structures
		4217
		4218	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4219	hash table sizes and so on. This will usually further increase code size
		4220	and can additionally have an effect on the size of data structures at
		4221	runtime.
		4222
		4223	=item C<4> - full API configuration
		4224
		4225	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4226	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4227
		4228	=item C<8> - full API
		4229
		4230	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4231	details on which parts of the API are still available without this
		4232	feature, and do not complain if this subset changes over time.
		4233
		4234	=item C<16> - enable all optional watcher types
		4235
		4236	Enables all optional watcher types. If you want to selectively enable
		4237	only some watcher types other than I/O and timers (e.g. prepare,
		4238	embed, async, child...) you can enable them manually by defining
		4239	C<EV_watchertype_ENABLE> to C<1> instead.
		4240
		4241	=item C<32> - enable all backends
		4242
		4243	This enables all backends - without this feature, you need to enable at
		4244	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4245
		4246	=item C<64> - enable OS-specific "helper" APIs
		4247
		4248	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4249	default.
		4250
		4251	=back
		4252
		4253	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4254	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4255	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4256	watchers, timers and monotonic clock support.
		4257
		4258	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4259	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4260	your program might be left out as well - a binary starting a timer and an
		4261	I/O watcher then might come out at only 5Kb.
		4262
		4263	=item EV_AVOID_STDIO
		4264
		4265	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4266	functions (printf, scanf, perror etc.). This will increase the code size
		4267	somewhat, but if your program doesn't otherwise depend on stdio and your
		4268	libc allows it, this avoids linking in the stdio library which is quite
		4269	big.
		4270
		4271	Note that error messages might become less precise when this option is
		4272	enabled.
		4273
		4274	=item EV_NSIG
		4275
		4276	The highest supported signal number, +1 (or, the number of
		4277	signals): Normally, libev tries to deduce the maximum number of signals
		4278	automatically, but sometimes this fails, in which case it can be
		4279	specified. Also, using a lower number than detected (C<32> should be
		4280	good for about any system in existence) can save some memory, as libev
		4281	statically allocates some 12-24 bytes per signal number.
		4282
		4283	=item EV_PID_HASHSIZE
		4284
		4285	C<ev_child> watchers use a small hash table to distribute workload by
		4286	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4287	usually more than enough. If you need to manage thousands of children you
		4288	might want to increase this value (I<must> be a power of two).
		4289
		4290	=item EV_INOTIFY_HASHSIZE
		4291
		4292	C<ev_stat> watchers use a small hash table to distribute workload by
		4293	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4294	disabled), usually more than enough. If you need to manage thousands of
		4295	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4296	power of two).
		4297
		4298	=item EV_USE_4HEAP
		4299
		4300	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4301	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4302	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4303	faster performance with many (thousands) of watchers.
		4304
		4305	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4306	will be C<0>.
		4307
		4308	=item EV_HEAP_CACHE_AT
		4309
		4310	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4311	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4312	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4313	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4314	but avoids random read accesses on heap changes. This improves performance
		4315	noticeably with many (hundreds) of watchers.
		4316
		4317	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4318	will be C<0>.
		4319
		4320	=item EV_VERIFY
		4321
		4322	Controls how much internal verification (see C<ev_verify ()>) will
		4323	be done: If set to C<0>, no internal verification code will be compiled
		4324	in. If set to C<1>, then verification code will be compiled in, but not
		4325	called. If set to C<2>, then the internal verification code will be
		4326	called once per loop, which can slow down libev. If set to C<3>, then the
		4327	verification code will be called very frequently, which will slow down
		4328	libev considerably.
		4329
		4330	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4331	will be C<0>.
		4332
		4333	=item EV_COMMON
		4334
		4335	By default, all watchers have a C<void *data> member. By redefining
		4336	this macro to something else you can include more and other types of
		4337	members. You have to define it each time you include one of the files,
		4338	though, and it must be identical each time.
		4339
		4340	For example, the perl EV module uses something like this:
		4341
		4342	#define EV_COMMON \
		4343	SV *self; /* contains this struct */ \
		4344	SV *cb_sv, *fh /* note no trailing ";" */
		4345
		4346	=item EV_CB_DECLARE (type)
		4347
		4348	=item EV_CB_INVOKE (watcher, revents)
		4349
		4350	=item ev_set_cb (ev, cb)
		4351
		4352	Can be used to change the callback member declaration in each watcher,
		4353	and the way callbacks are invoked and set. Must expand to a struct member
		4354	definition and a statement, respectively. See the F<ev.h> header file for
		4355	their default definitions. One possible use for overriding these is to
		4356	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4357	method calls instead of plain function calls in C++.
		4358
		4359	=back
		4360
		4361	=head2 EXPORTED API SYMBOLS
		4362
		4363	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4364	exported symbols, you can use the provided F<Symbol.*> files which list
		4365	all public symbols, one per line:
		4366
		4367	Symbols.ev for libev proper
		4368	Symbols.event for the libevent emulation
		4369
		4370	This can also be used to rename all public symbols to avoid clashes with
		4371	multiple versions of libev linked together (which is obviously bad in
		4372	itself, but sometimes it is inconvenient to avoid this).
		4373
		4374	A sed command like this will create wrapper C<#define>'s that you need to
		4375	include before including F<ev.h>:
		4376
		4377	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4378
		4379	This would create a file F<wrap.h> which essentially looks like this:
		4380
		4381	#define ev_backend myprefix_ev_backend
		4382	#define ev_check_start myprefix_ev_check_start
		4383	#define ev_check_stop myprefix_ev_check_stop
		4384	...
		4385
		4386	=head2 EXAMPLES
		4387
		4388	For a real-world example of a program the includes libev
		4389	verbatim, you can have a look at the EV perl module
		4390	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4391	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4392	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4393	will be compiled. It is pretty complex because it provides its own header
		4394	file.
		4395
		4396	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4397	that everybody includes and which overrides some configure choices:
		4398
		4399	#define EV_FEATURES 8
		4400	#define EV_USE_SELECT 1
		4401	#define EV_PREPARE_ENABLE 1
		4402	#define EV_IDLE_ENABLE 1
		4403	#define EV_SIGNAL_ENABLE 1
		4404	#define EV_CHILD_ENABLE 1
		4405	#define EV_USE_STDEXCEPT 0
		4406	#define EV_CONFIG_H <config.h>
		4407
		4408	#include "ev++.h"
		4409
		4410	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4411
		4412	#include "ev_cpp.h"
		4413	#include "ev.c"
		4414
		4415	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4416
		4417	=head2 THREADS AND COROUTINES
		4418
		4419	=head3 THREADS
		4420
		4421	All libev functions are reentrant and thread-safe unless explicitly
		4422	documented otherwise, but libev implements no locking itself. This means
		4423	that you can use as many loops as you want in parallel, as long as there
		4424	are no concurrent calls into any libev function with the same loop
		4425	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4426	of course): libev guarantees that different event loops share no data
		4427	structures that need any locking.
		4428
		4429	Or to put it differently: calls with different loop parameters can be done
		4430	concurrently from multiple threads, calls with the same loop parameter
		4431	must be done serially (but can be done from different threads, as long as
		4432	only one thread ever is inside a call at any point in time, e.g. by using
		4433	a mutex per loop).
		4434
		4435	Specifically to support threads (and signal handlers), libev implements
		4436	so-called C<ev_async> watchers, which allow some limited form of
		4437	concurrency on the same event loop, namely waking it up "from the
		4438	outside".
		4439
		4440	If you want to know which design (one loop, locking, or multiple loops
		4441	without or something else still) is best for your problem, then I cannot
		4442	help you, but here is some generic advice:
		4443
		4444	=over 4
		4445
		4446	=item * most applications have a main thread: use the default libev loop
		4447	in that thread, or create a separate thread running only the default loop.
		4448
		4449	This helps integrating other libraries or software modules that use libev
		4450	themselves and don't care/know about threading.
		4451
		4452	=item * one loop per thread is usually a good model.
		4453
		4454	Doing this is almost never wrong, sometimes a better-performance model
		4455	exists, but it is always a good start.
		4456
		4457	=item * other models exist, such as the leader/follower pattern, where one
		4458	loop is handed through multiple threads in a kind of round-robin fashion.
		4459
		4460	Choosing a model is hard - look around, learn, know that usually you can do
		4461	better than you currently do :-)
		4462
		4463	=item * often you need to talk to some other thread which blocks in the
		4464	event loop.
		4465
		4466	C<ev_async> watchers can be used to wake them up from other threads safely
		4467	(or from signal contexts...).
		4468
		4469	An example use would be to communicate signals or other events that only
		4470	work in the default loop by registering the signal watcher with the
		4471	default loop and triggering an C<ev_async> watcher from the default loop
		4472	watcher callback into the event loop interested in the signal.
		4473
		4474	=back
		4475
		4476	=head4 THREAD LOCKING EXAMPLE
		4477
		4478	Here is a fictitious example of how to run an event loop in a different
		4479	thread than where callbacks are being invoked and watchers are
		4480	created/added/removed.
		4481
		4482	For a real-world example, see the C<EV::Loop::Async> perl module,
		4483	which uses exactly this technique (which is suited for many high-level
		4484	languages).
		4485
		4486	The example uses a pthread mutex to protect the loop data, a condition
		4487	variable to wait for callback invocations, an async watcher to notify the
		4488	event loop thread and an unspecified mechanism to wake up the main thread.
		4489
		4490	First, you need to associate some data with the event loop:
		4491
		4492	typedef struct {
		4493	mutex_t lock; /* global loop lock */
		4494	ev_async async_w;
		4495	thread_t tid;
		4496	cond_t invoke_cv;
		4497	} userdata;
		4498
		4499	void prepare_loop (EV_P)
		4500	{
		4501	// for simplicity, we use a static userdata struct.
		4502	static userdata u;
		4503
		4504	ev_async_init (&u->async_w, async_cb);
		4505	ev_async_start (EV_A_ &u->async_w);
		4506
		4507	pthread_mutex_init (&u->lock, 0);
		4508	pthread_cond_init (&u->invoke_cv, 0);
		4509
		4510	// now associate this with the loop
		4511	ev_set_userdata (EV_A_ u);
		4512	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4513	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4514
		4515	// then create the thread running ev_loop
		4516	pthread_create (&u->tid, 0, l_run, EV_A);
		4517	}
		4518
		4519	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4520	solely to wake up the event loop so it takes notice of any new watchers
		4521	that might have been added:
		4522
		4523	static void
		4524	async_cb (EV_P_ ev_async *w, int revents)
		4525	{
		4526	// just used for the side effects
		4527	}
		4528
		4529	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4530	protecting the loop data, respectively.
		4531
		4532	static void
		4533	l_release (EV_P)
		4534	{
		4535	userdata *u = ev_userdata (EV_A);
		4536	pthread_mutex_unlock (&u->lock);
		4537	}
		4538
		4539	static void
		4540	l_acquire (EV_P)
		4541	{
		4542	userdata *u = ev_userdata (EV_A);
		4543	pthread_mutex_lock (&u->lock);
		4544	}
		4545
		4546	The event loop thread first acquires the mutex, and then jumps straight
		4547	into C<ev_run>:
		4548
		4549	void *
		4550	l_run (void *thr_arg)
		4551	{
		4552	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4553
		4554	l_acquire (EV_A);
		4555	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4556	ev_run (EV_A_ 0);
		4557	l_release (EV_A);
		4558
		4559	return 0;
		4560	}
		4561
		4562	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4563	signal the main thread via some unspecified mechanism (signals? pipe
		4564	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4565	have been called (in a while loop because a) spurious wakeups are possible
		4566	and b) skipping inter-thread-communication when there are no pending
		4567	watchers is very beneficial):
		4568
		4569	static void
		4570	l_invoke (EV_P)
		4571	{
		4572	userdata *u = ev_userdata (EV_A);
		4573
		4574	while (ev_pending_count (EV_A))
		4575	{
		4576	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4577	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4578	}
		4579	}
		4580
		4581	Now, whenever the main thread gets told to invoke pending watchers, it
		4582	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4583	thread to continue:
		4584
		4585	static void
		4586	real_invoke_pending (EV_P)
		4587	{
		4588	userdata *u = ev_userdata (EV_A);
		4589
		4590	pthread_mutex_lock (&u->lock);
		4591	ev_invoke_pending (EV_A);
		4592	pthread_cond_signal (&u->invoke_cv);
		4593	pthread_mutex_unlock (&u->lock);
		4594	}
		4595
		4596	Whenever you want to start/stop a watcher or do other modifications to an
		4597	event loop, you will now have to lock:
		4598
		4599	ev_timer timeout_watcher;
		4600	userdata *u = ev_userdata (EV_A);
		4601
		4602	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4603
		4604	pthread_mutex_lock (&u->lock);
		4605	ev_timer_start (EV_A_ &timeout_watcher);
		4606	ev_async_send (EV_A_ &u->async_w);
		4607	pthread_mutex_unlock (&u->lock);
		4608
		4609	Note that sending the C<ev_async> watcher is required because otherwise
		4610	an event loop currently blocking in the kernel will have no knowledge
		4611	about the newly added timer. By waking up the loop it will pick up any new
		4612	watchers in the next event loop iteration.
		4613
		4614	=head3 COROUTINES
		4615
		4616	Libev is very accommodating to coroutines ("cooperative threads"):
		4617	libev fully supports nesting calls to its functions from different
		4618	coroutines (e.g. you can call C<ev_run> on the same loop from two
		4619	different coroutines, and switch freely between both coroutines running
		4620	the loop, as long as you don't confuse yourself). The only exception is
		4621	that you must not do this from C<ev_periodic> reschedule callbacks.
		4622
		4623	Care has been taken to ensure that libev does not keep local state inside
		4624	C<ev_run>, and other calls do not usually allow for coroutine switches as
		4625	they do not call any callbacks.
		4626
		4627	=head2 COMPILER WARNINGS
		4628
		4629	Depending on your compiler and compiler settings, you might get no or a
		4630	lot of warnings when compiling libev code. Some people are apparently
		4631	scared by this.
		4632
		4633	However, these are unavoidable for many reasons. For one, each compiler
		4634	has different warnings, and each user has different tastes regarding
		4635	warning options. "Warn-free" code therefore cannot be a goal except when
		4636	targeting a specific compiler and compiler-version.
		4637
		4638	Another reason is that some compiler warnings require elaborate
		4639	workarounds, or other changes to the code that make it less clear and less
		4640	maintainable.
		4641
		4642	And of course, some compiler warnings are just plain stupid, or simply
		4643	wrong (because they don't actually warn about the condition their message
		4644	seems to warn about). For example, certain older gcc versions had some
		4645	warnings that resulted in an extreme number of false positives. These have
		4646	been fixed, but some people still insist on making code warn-free with
		4647	such buggy versions.
		4648
		4649	While libev is written to generate as few warnings as possible,
		4650	"warn-free" code is not a goal, and it is recommended not to build libev
		4651	with any compiler warnings enabled unless you are prepared to cope with
		4652	them (e.g. by ignoring them). Remember that warnings are just that:
		4653	warnings, not errors, or proof of bugs.
		4654
		4655
		4656	=head2 VALGRIND
		4657
		4658	Valgrind has a special section here because it is a popular tool that is
		4659	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4660
		4661	If you think you found a bug (memory leak, uninitialised data access etc.)
		4662	in libev, then check twice: If valgrind reports something like:
		4663
		4664	==2274== definitely lost: 0 bytes in 0 blocks.
		4665	==2274== possibly lost: 0 bytes in 0 blocks.
		4666	==2274== still reachable: 256 bytes in 1 blocks.
		4667
		4668	Then there is no memory leak, just as memory accounted to global variables
		4669	is not a memleak - the memory is still being referenced, and didn't leak.
		4670
		4671	Similarly, under some circumstances, valgrind might report kernel bugs
		4672	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4673	although an acceptable workaround has been found here), or it might be
		4674	confused.
		4675
		4676	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4677	make it into some kind of religion.
		4678
		4679	If you are unsure about something, feel free to contact the mailing list
		4680	with the full valgrind report and an explanation on why you think this
		4681	is a bug in libev (best check the archives, too :). However, don't be
		4682	annoyed when you get a brisk "this is no bug" answer and take the chance
		4683	of learning how to interpret valgrind properly.
		4684
		4685	If you need, for some reason, empty reports from valgrind for your project
		4686	I suggest using suppression lists.
		4687
		4688
		4689	=head1 PORTABILITY NOTES
		4690
		4691	=head2 GNU/LINUX 32 BIT LIMITATIONS
		4692
		4693	GNU/Linux is the only common platform that supports 64 bit file/large file
		4694	interfaces but I<disables> them by default.
		4695
		4696	That means that libev compiled in the default environment doesn't support
		4697	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		4698
		4699	Unfortunately, many programs try to work around this GNU/Linux issue
		4700	by enabling the large file API, which makes them incompatible with the
		4701	standard libev compiled for their system.
		4702
		4703	Likewise, libev cannot enable the large file API itself as this would
		4704	suddenly make it incompatible to the default compile time environment,
		4705	i.e. all programs not using special compile switches.
		4706
		4707	=head2 OS/X AND DARWIN BUGS
		4708
		4709	The whole thing is a bug if you ask me - basically any system interface
		4710	you touch is broken, whether it is locales, poll, kqueue or even the
		4711	OpenGL drivers.
		4712
		4713	=head3 C<kqueue> is buggy
		4714
		4715	The kqueue syscall is broken in all known versions - most versions support
		4716	only sockets, many support pipes.
		4717
		4718	Libev tries to work around this by not using C<kqueue> by default on this
		4719	rotten platform, but of course you can still ask for it when creating a
		4720	loop - embedding a socket-only kqueue loop into a select-based one is
		4721	probably going to work well.
		4722
		4723	=head3 C<poll> is buggy
		4724
		4725	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4726	implementation by something calling C<kqueue> internally around the 10.5.6
		4727	release, so now C<kqueue> I<and> C<poll> are broken.
		4728
		4729	Libev tries to work around this by not using C<poll> by default on
		4730	this rotten platform, but of course you can still ask for it when creating
		4731	a loop.
		4732
		4733	=head3 C<select> is buggy
		4734
		4735	All that's left is C<select>, and of course Apple found a way to fuck this
		4736	one up as well: On OS/X, C<select> actively limits the number of file
		4737	descriptors you can pass in to 1024 - your program suddenly crashes when
		4738	you use more.
		4739
		4740	There is an undocumented "workaround" for this - defining
		4741	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4742	work on OS/X.
		4743
		4744	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4745
		4746	=head3 C<errno> reentrancy
		4747
		4748	The default compile environment on Solaris is unfortunately so
		4749	thread-unsafe that you can't even use components/libraries compiled
		4750	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		4751	defined by default. A valid, if stupid, implementation choice.
		4752
		4753	If you want to use libev in threaded environments you have to make sure
		4754	it's compiled with C<_REENTRANT> defined.
		4755
		4756	=head3 Event port backend
		4757
		4758	The scalable event interface for Solaris is called "event
		4759	ports". Unfortunately, this mechanism is very buggy in all major
		4760	releases. If you run into high CPU usage, your program freezes or you get
		4761	a large number of spurious wakeups, make sure you have all the relevant
		4762	and latest kernel patches applied. No, I don't know which ones, but there
		4763	are multiple ones to apply, and afterwards, event ports actually work
		4764	great.
		4765
		4766	If you can't get it to work, you can try running the program by setting
		4767	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4768	C<select> backends.
		4769
		4770	=head2 AIX POLL BUG
		4771
		4772	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4773	this by trying to avoid the poll backend altogether (i.e. it's not even
		4774	compiled in), which normally isn't a big problem as C<select> works fine
		4775	with large bitsets on AIX, and AIX is dead anyway.
		4776
		4777	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4778
		4779	=head3 General issues
		4780
		4781	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4782	requires, and its I/O model is fundamentally incompatible with the POSIX
		4783	model. Libev still offers limited functionality on this platform in
		4784	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4785	descriptors. This only applies when using Win32 natively, not when using
		4786	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4787	as every compielr comes with a slightly differently broken/incompatible
		4788	environment.
		4789
		4790	Lifting these limitations would basically require the full
		4791	re-implementation of the I/O system. If you are into this kind of thing,
		4792	then note that glib does exactly that for you in a very portable way (note
		4793	also that glib is the slowest event library known to man).
		4794
		4795	There is no supported compilation method available on windows except
		4796	embedding it into other applications.
		4797
		4798	Sensible signal handling is officially unsupported by Microsoft - libev
		4799	tries its best, but under most conditions, signals will simply not work.
		4800
		4801	Not a libev limitation but worth mentioning: windows apparently doesn't
		4802	accept large writes: instead of resulting in a partial write, windows will
		4803	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4804	so make sure you only write small amounts into your sockets (less than a
		4805	megabyte seems safe, but this apparently depends on the amount of memory
		4806	available).
		4807
		4808	Due to the many, low, and arbitrary limits on the win32 platform and
		4809	the abysmal performance of winsockets, using a large number of sockets
		4810	is not recommended (and not reasonable). If your program needs to use
		4811	more than a hundred or so sockets, then likely it needs to use a totally
		4812	different implementation for windows, as libev offers the POSIX readiness
		4813	notification model, which cannot be implemented efficiently on windows
		4814	(due to Microsoft monopoly games).
		4815
		4816	A typical way to use libev under windows is to embed it (see the embedding
		4817	section for details) and use the following F<evwrap.h> header file instead
		4818	of F<ev.h>:
		4819
		4820	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4821	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4822
		4823	#include "ev.h"
		4824
		4825	And compile the following F<evwrap.c> file into your project (make sure
		4826	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4827
		4828	#include "evwrap.h"
		4829	#include "ev.c"
		4830
		4831	=head3 The winsocket C<select> function
		4832
		4833	The winsocket C<select> function doesn't follow POSIX in that it
		4834	requires socket I<handles> and not socket I<file descriptors> (it is
		4835	also extremely buggy). This makes select very inefficient, and also
		4836	requires a mapping from file descriptors to socket handles (the Microsoft
		4837	C runtime provides the function C<_open_osfhandle> for this). See the
		4838	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4839	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4840
		4841	The configuration for a "naked" win32 using the Microsoft runtime
		4842	libraries and raw winsocket select is:
		4843
		4844	#define EV_USE_SELECT 1
		4845	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4846
		4847	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4848	complexity in the O(n²) range when using win32.
		4849
		4850	=head3 Limited number of file descriptors
		4851
		4852	Windows has numerous arbitrary (and low) limits on things.
		4853
		4854	Early versions of winsocket's select only supported waiting for a maximum
		4855	of C<64> handles (probably owning to the fact that all windows kernels
		4856	can only wait for C<64> things at the same time internally; Microsoft
		4857	recommends spawning a chain of threads and wait for 63 handles and the
		4858	previous thread in each. Sounds great!).
		4859
		4860	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4861	to some high number (e.g. C<2048>) before compiling the winsocket select
		4862	call (which might be in libev or elsewhere, for example, perl and many
		4863	other interpreters do their own select emulation on windows).
		4864
		4865	Another limit is the number of file descriptors in the Microsoft runtime
		4866	libraries, which by default is C<64> (there must be a hidden I<64>
		4867	fetish or something like this inside Microsoft). You can increase this
		4868	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4869	(another arbitrary limit), but is broken in many versions of the Microsoft
		4870	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4871	(depending on windows version and/or the phase of the moon). To get more,
		4872	you need to wrap all I/O functions and provide your own fd management, but
		4873	the cost of calling select (O(n²)) will likely make this unworkable.
		4874
		4875	=head2 PORTABILITY REQUIREMENTS
		4876
		4877	In addition to a working ISO-C implementation and of course the
		4878	backend-specific APIs, libev relies on a few additional extensions:
		4879
		4880	=over 4
		4881
		4882	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4883	calling conventions regardless of C<ev_watcher_type *>.
		4884
		4885	Libev assumes not only that all watcher pointers have the same internal
		4886	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4887	assumes that the same (machine) code can be used to call any watcher
		4888	callback: The watcher callbacks have different type signatures, but libev
		4889	calls them using an C<ev_watcher *> internally.
		4890
		4891	=item pointer accesses must be thread-atomic
		4892
		4893	Accessing a pointer value must be atomic, it must both be readable and
		4894	writable in one piece - this is the case on all current architectures.
		4895
		4896	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4897
		4898	The type C<sig_atomic_t volatile> (or whatever is defined as
		4899	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4900	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4901	believed to be sufficiently portable.
		4902
		4903	=item C<sigprocmask> must work in a threaded environment
		4904
		4905	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4906	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4907	pthread implementations will either allow C<sigprocmask> in the "main
		4908	thread" or will block signals process-wide, both behaviours would
		4909	be compatible with libev. Interaction between C<sigprocmask> and
		4910	C<pthread_sigmask> could complicate things, however.
		4911
		4912	The most portable way to handle signals is to block signals in all threads
		4913	except the initial one, and run the default loop in the initial thread as
		4914	well.
		4915
		4916	=item C<long> must be large enough for common memory allocation sizes
		4917
		4918	To improve portability and simplify its API, libev uses C<long> internally
		4919	instead of C<size_t> when allocating its data structures. On non-POSIX
		4920	systems (Microsoft...) this might be unexpectedly low, but is still at
		4921	least 31 bits everywhere, which is enough for hundreds of millions of
		4922	watchers.
		4923
		4924	=item C<double> must hold a time value in seconds with enough accuracy
		4925
		4926	The type C<double> is used to represent timestamps. It is required to
		4927	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		4928	good enough for at least into the year 4000 with millisecond accuracy
		4929	(the design goal for libev). This requirement is overfulfilled by
		4930	implementations using IEEE 754, which is basically all existing ones. With
		4931	IEEE 754 doubles, you get microsecond accuracy until at least 2200.
		4932
		4933	=back
		4934
		4935	If you know of other additional requirements drop me a note.
		4936
		4937
		4938	=head1 ALGORITHMIC COMPLEXITIES
		4939
		4940	In this section the complexities of (many of) the algorithms used inside
		4941	libev will be documented. For complexity discussions about backends see
		4942	the documentation for C<ev_default_init>.
		4943
		4944	All of the following are about amortised time: If an array needs to be
		4945	extended, libev needs to realloc and move the whole array, but this
		4946	happens asymptotically rarer with higher number of elements, so O(1) might
		4947	mean that libev does a lengthy realloc operation in rare cases, but on
		4948	average it is much faster and asymptotically approaches constant time.
		4949
		4950	=over 4
		4951
		4952	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4953
		4954	This means that, when you have a watcher that triggers in one hour and
		4955	there are 100 watchers that would trigger before that, then inserting will
		4956	have to skip roughly seven (C<ld 100>) of these watchers.
		4957
		4958	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4959
		4960	That means that changing a timer costs less than removing/adding them,
		4961	as only the relative motion in the event queue has to be paid for.
		4962
		4963	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4964
		4965	These just add the watcher into an array or at the head of a list.
		4966
		4967	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4968
		4969	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4970
		4971	These watchers are stored in lists, so they need to be walked to find the
		4972	correct watcher to remove. The lists are usually short (you don't usually
		4973	have many watchers waiting for the same fd or signal: one is typical, two
		4974	is rare).
		4975
		4976	=item Finding the next timer in each loop iteration: O(1)
		4977
		4978	By virtue of using a binary or 4-heap, the next timer is always found at a
		4979	fixed position in the storage array.
		4980
		4981	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4982
		4983	A change means an I/O watcher gets started or stopped, which requires
		4984	libev to recalculate its status (and possibly tell the kernel, depending
		4985	on backend and whether C<ev_io_set> was used).
		4986
		4987	=item Activating one watcher (putting it into the pending state): O(1)
		4988
		4989	=item Priority handling: O(number_of_priorities)
		4990
		4991	Priorities are implemented by allocating some space for each
		4992	priority. When doing priority-based operations, libev usually has to
		4993	linearly search all the priorities, but starting/stopping and activating
		4994	watchers becomes O(1) with respect to priority handling.
		4995
		4996	=item Sending an ev_async: O(1)
		4997
		4998	=item Processing ev_async_send: O(number_of_async_watchers)
		4999
		5000	=item Processing signals: O(max_signal_number)
		5001
		5002	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5003	calls in the current loop iteration. Checking for async and signal events
		5004	involves iterating over all running async watchers or all signal numbers.
		5005
		5006	=back
		5007
		5008
		5009	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5010
		5011	The major version 4 introduced some incompatible changes to the API.
		5012
		5013	At the moment, the C<ev.h> header file provides compatibility definitions
		5014	for all changes, so most programs should still compile. The compatibility
		5015	layer might be removed in later versions of libev, so better update to the
		5016	new API early than late.
		5017
		5018	=over 4
		5019
		5020	=item C<EV_COMPAT3> backwards compatibility mechanism
		5021
		5022	The backward compatibility mechanism can be controlled by
		5023	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>
		5024	section.
		5025
		5026	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5027
		5028	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5029
		5030	ev_loop_destroy (EV_DEFAULT_UC);
		5031	ev_loop_fork (EV_DEFAULT);
		5032
		5033	=item function/symbol renames
		5034
		5035	A number of functions and symbols have been renamed:
		5036
		5037	ev_loop => ev_run
		5038	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5039	EVLOOP_ONESHOT => EVRUN_ONCE
		5040
		5041	ev_unloop => ev_break
		5042	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5043	EVUNLOOP_ONE => EVBREAK_ONE
		5044	EVUNLOOP_ALL => EVBREAK_ALL
		5045
		5046	EV_TIMEOUT => EV_TIMER
		5047
		5048	ev_loop_count => ev_iteration
		5049	ev_loop_depth => ev_depth
		5050	ev_loop_verify => ev_verify
		5051
		5052	Most functions working on C<struct ev_loop> objects don't have an
		5053	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5054	associated constants have been renamed to not collide with the C<struct
		5055	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5056	as all other watcher types. Note that C<ev_loop_fork> is still called
		5057	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5058	typedef.
		5059
		5060	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5061
		5062	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5063	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5064	and work, but the library code will of course be larger.
		5065
		5066	=back
		5067
		5068
		5069	=head1 GLOSSARY
		5070
		5071	=over 4
		5072
		5073	=item active
		5074
		5075	A watcher is active as long as it has been started and not yet stopped.
		5076	See L<WATCHER STATES> for details.
		5077
		5078	=item application
		5079
		5080	In this document, an application is whatever is using libev.
		5081
		5082	=item backend
		5083
		5084	The part of the code dealing with the operating system interfaces.
		5085
		5086	=item callback
		5087
		5088	The address of a function that is called when some event has been
		5089	detected. Callbacks are being passed the event loop, the watcher that
		5090	received the event, and the actual event bitset.
		5091
		5092	=item callback/watcher invocation
		5093
		5094	The act of calling the callback associated with a watcher.
		5095
		5096	=item event
		5097
		5098	A change of state of some external event, such as data now being available
		5099	for reading on a file descriptor, time having passed or simply not having
		5100	any other events happening anymore.
		5101
		5102	In libev, events are represented as single bits (such as C<EV_READ> or
		5103	C<EV_TIMER>).
		5104
		5105	=item event library
		5106
		5107	A software package implementing an event model and loop.
		5108
		5109	=item event loop
		5110
		5111	An entity that handles and processes external events and converts them
		5112	into callback invocations.
		5113
		5114	=item event model
		5115
		5116	The model used to describe how an event loop handles and processes
		5117	watchers and events.
		5118
		5119	=item pending
		5120
		5121	A watcher is pending as soon as the corresponding event has been
		5122	detected. See L<WATCHER STATES> for details.
		5123
		5124	=item real time
		5125
		5126	The physical time that is observed. It is apparently strictly monotonic :)
		5127
		5128	=item wall-clock time
		5129
		5130	The time and date as shown on clocks. Unlike real time, it can actually
		5131	be wrong and jump forwards and backwards, e.g. when the you adjust your
		5132	clock.
		5133
		5134	=item watcher
		5135
		5136	A data structure that describes interest in certain events. Watchers need
		5137	to be started (attached to an event loop) before they can receive events.
		5138
		5139	=back
		5140
767	=head1 AUTHOR	5141	=head1 AUTHOR
768		5142
769	Marc Lehmann <libev@schmorp.de>.	5143	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5144	Magnusson and Emanuele Giaquinta.
770		5145

Diff Legend

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