ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.8 by root, Mon Nov 12 08:20:02 2007 UTC vs.
Revision 1.446 by root, Mon Mar 18 19:28:15 2019 UTC

		1	=encoding utf-8
		2
1	=head1 NAME	3	=head1 NAME
2		4
3	libev - a high performance full-featured event loop written in C	5	libev - a high performance full-featured event loop written in C
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
6		8
7	#include <ev.h>	9	#include <ev.h>
8		10
9	=head1 DESCRIPTION	11	=head2 EXAMPLE PROGRAM
		12
		13	// a single header file is required
		14	#include <ev.h>
		15
		16	#include <stdio.h> // for puts
		17
		18	// every watcher type has its own typedef'd struct
		19	// with the name ev_TYPE
		20	ev_io stdin_watcher;
		21	ev_timer timeout_watcher;
		22
		23	// all watcher callbacks have a similar signature
		24	// this callback is called when data is readable on stdin
		25	static void
		26	stdin_cb (EV_P_ ev_io *w, int revents)
		27	{
		28	puts ("stdin ready");
		29	// for one-shot events, one must manually stop the watcher
		30	// with its corresponding stop function.
		31	ev_io_stop (EV_A_ w);
		32
		33	// this causes all nested ev_run's to stop iterating
		34	ev_break (EV_A_ EVBREAK_ALL);
		35	}
		36
		37	// another callback, this time for a time-out
		38	static void
		39	timeout_cb (EV_P_ ev_timer *w, int revents)
		40	{
		41	puts ("timeout");
		42	// this causes the innermost ev_run to stop iterating
		43	ev_break (EV_A_ EVBREAK_ONE);
		44	}
		45
		46	int
		47	main (void)
		48	{
		49	// use the default event loop unless you have special needs
		50	struct ev_loop *loop = EV_DEFAULT;
		51
		52	// initialise an io watcher, then start it
		53	// this one will watch for stdin to become readable
		54	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		55	ev_io_start (loop, &stdin_watcher);
		56
		57	// initialise a timer watcher, then start it
		58	// simple non-repeating 5.5 second timeout
		59	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		60	ev_timer_start (loop, &timeout_watcher);
		61
		62	// now wait for events to arrive
		63	ev_run (loop, 0);
		64
		65	// break was called, so exit
		66	return 0;
		67	}
		68
		69	=head1 ABOUT THIS DOCUMENT
		70
		71	This document documents the libev software package.
		72
		73	The newest version of this document is also available as an html-formatted
		74	web page you might find easier to navigate when reading it for the first
		75	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		76
		77	While this document tries to be as complete as possible in documenting
		78	libev, its usage and the rationale behind its design, it is not a tutorial
		79	on event-based programming, nor will it introduce event-based programming
		80	with libev.
		81
		82	Familiarity with event based programming techniques in general is assumed
		83	throughout this document.
		84
		85	=head1 WHAT TO READ WHEN IN A HURRY
		86
		87	This manual tries to be very detailed, but unfortunately, this also makes
		88	it very long. If you just want to know the basics of libev, I suggest
		89	reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
		90	look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
		91	C<ev_timer> sections in L</WATCHER TYPES>.
		92
		93	=head1 ABOUT LIBEV
10		94
11	Libev is an event loop: you register interest in certain events (such as a	95	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	96	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	97	these event sources and provide your program with events.
14		98
15	To do this, it must take more or less complete control over your process	99	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	100	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	101	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	103	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	104	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	105	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	106	watcher.
23		107
24	=head1 FEATURES	108	=head2 FEATURES
25		109
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	110	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	111	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	112	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		113	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		114	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		115	timers (C<ev_timer>), absolute timers with customised rescheduling
		116	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	117	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	118	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		119	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		120	limited support for fork events (C<ev_fork>).
		121
		122	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	123	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	124	for example).
33		125
34	=head1 CONVENTIONS	126	=head2 CONVENTIONS
35		127
36	Libev is very configurable. In this manual the default configuration	128	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	129	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	130	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	131	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	132	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	133	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	134	this argument.
43		135
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	136	=head2 TIME REPRESENTATION
45		137
46	Libev represents time as a single floating point number, representing the	138	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	139	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	140	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	141	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	142	too. It usually aliases to the C<double> type in C. When you need to do
		143	any calculations on it, you should treat it as some floating point value.
		144
		145	Unlike the name component C<stamp> might indicate, it is also used for
		146	time differences (e.g. delays) throughout libev.
		147
		148	=head1 ERROR HANDLING
		149
		150	Libev knows three classes of errors: operating system errors, usage errors
		151	and internal errors (bugs).
		152
		153	When libev catches an operating system error it cannot handle (for example
		154	a system call indicating a condition libev cannot fix), it calls the callback
		155	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		156	abort. The default is to print a diagnostic message and to call C<abort
		157	()>.
		158
		159	When libev detects a usage error such as a negative timer interval, then
		160	it will print a diagnostic message and abort (via the C<assert> mechanism,
		161	so C<NDEBUG> will disable this checking): these are programming errors in
		162	the libev caller and need to be fixed there.
		163
		164	Libev also has a few internal error-checking C<assert>ions, and also has
		165	extensive consistency checking code. These do not trigger under normal
		166	circumstances, as they indicate either a bug in libev or worse.
		167
		168
		169	=head1 GLOBAL FUNCTIONS
		170
		171	These functions can be called anytime, even before initialising the
		172	library in any way.
51		173
52	=over 4	174	=over 4
53		175
54	=item ev_tstamp ev_time ()	176	=item ev_tstamp ev_time ()
55		177
56	Returns the current time as libev would use it.	178	Returns the current time as libev would use it. Please note that the
		179	C<ev_now> function is usually faster and also often returns the timestamp
		180	you actually want to know. Also interesting is the combination of
		181	C<ev_now_update> and C<ev_now>.
		182
		183	=item ev_sleep (ev_tstamp interval)
		184
		185	Sleep for the given interval: The current thread will be blocked
		186	until either it is interrupted or the given time interval has
		187	passed (approximately - it might return a bit earlier even if not
		188	interrupted). Returns immediately if C<< interval <= 0 >>.
		189
		190	Basically this is a sub-second-resolution C<sleep ()>.
		191
		192	The range of the C<interval> is limited - libev only guarantees to work
		193	with sleep times of up to one day (C<< interval <= 86400 >>).
57		194
58	=item int ev_version_major ()	195	=item int ev_version_major ()
59		196
60	=item int ev_version_minor ()	197	=item int ev_version_minor ()
61		198
62	You can find out the major and minor version numbers of the library	199	You can find out the major and minor ABI version numbers of the library
63	you linked against by calling the functions C<ev_version_major> and	200	you linked against by calling the functions C<ev_version_major> and
64	C<ev_version_minor>. If you want, you can compare against the global	201	C<ev_version_minor>. If you want, you can compare against the global
65	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	202	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
66	version of the library your program was compiled against.	203	version of the library your program was compiled against.
67		204
		205	These version numbers refer to the ABI version of the library, not the
		206	release version.
		207
68	Usually, its a good idea to terminate if the major versions mismatch,	208	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	209	as this indicates an incompatible change. Minor versions are usually
70	compatible to older versions, so a larger minor version alone is usually	210	compatible to older versions, so a larger minor version alone is usually
71	not a problem.	211	not a problem.
72		212
		213	Example: Make sure we haven't accidentally been linked against the wrong
		214	version (note, however, that this will not detect other ABI mismatches,
		215	such as LFS or reentrancy).
		216
		217	assert (("libev version mismatch",
		218	ev_version_major () == EV_VERSION_MAJOR
		219	&& ev_version_minor () >= EV_VERSION_MINOR));
		220
		221	=item unsigned int ev_supported_backends ()
		222
		223	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		224	value) compiled into this binary of libev (independent of their
		225	availability on the system you are running on). See C<ev_default_loop> for
		226	a description of the set values.
		227
		228	Example: make sure we have the epoll method, because yeah this is cool and
		229	a must have and can we have a torrent of it please!!!11
		230
		231	assert (("sorry, no epoll, no sex",
		232	ev_supported_backends () & EVBACKEND_EPOLL));
		233
		234	=item unsigned int ev_recommended_backends ()
		235
		236	Return the set of all backends compiled into this binary of libev and
		237	also recommended for this platform, meaning it will work for most file
		238	descriptor types. This set is often smaller than the one returned by
		239	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		240	and will not be auto-detected unless you explicitly request it (assuming
		241	you know what you are doing). This is the set of backends that libev will
		242	probe for if you specify no backends explicitly.
		243
		244	=item unsigned int ev_embeddable_backends ()
		245
		246	Returns the set of backends that are embeddable in other event loops. This
		247	value is platform-specific but can include backends not available on the
		248	current system. To find which embeddable backends might be supported on
		249	the current system, you would need to look at C<ev_embeddable_backends ()
		250	& ev_supported_backends ()>, likewise for recommended ones.
		251
		252	See the description of C<ev_embed> watchers for more info.
		253
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	254	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
74		255
75	Sets the allocation function to use (the prototype is similar to the	256	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	257	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
77	and free memory (no surprises here). If it returns zero when memory	258	used to allocate and free memory (no surprises here). If it returns zero
78	needs to be allocated, the library might abort or take some potentially	259	when memory needs to be allocated (C<size != 0>), the library might abort
79	destructive action. The default is your system realloc function.	260	or take some potentially destructive action.
		261
		262	Since some systems (at least OpenBSD and Darwin) fail to implement
		263	correct C<realloc> semantics, libev will use a wrapper around the system
		264	C<realloc> and C<free> functions by default.
80		265
81	You could override this function in high-availability programs to, say,	266	You could override this function in high-availability programs to, say,
82	free some memory if it cannot allocate memory, to use a special allocator,	267	free some memory if it cannot allocate memory, to use a special allocator,
83	or even to sleep a while and retry until some memory is available.	268	or even to sleep a while and retry until some memory is available.
84		269
		270	Example: The following is the C<realloc> function that libev itself uses
		271	which should work with C<realloc> and C<free> functions of all kinds and
		272	is probably a good basis for your own implementation.
		273
		274	static void *
		275	ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
		276	{
		277	if (size)
		278	return realloc (ptr, size);
		279
		280	free (ptr);
		281	return 0;
		282	}
		283
		284	Example: Replace the libev allocator with one that waits a bit and then
		285	retries.
		286
		287	static void *
		288	persistent_realloc (void *ptr, size_t size)
		289	{
		290	if (!size)
		291	{
		292	free (ptr);
		293	return 0;
		294	}
		295
		296	for (;;)
		297	{
		298	void *newptr = realloc (ptr, size);
		299
		300	if (newptr)
		301	return newptr;
		302
		303	sleep (60);
		304	}
		305	}
		306
		307	...
		308	ev_set_allocator (persistent_realloc);
		309
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	310	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
86		311
87	Set the callback function to call on a retryable syscall error (such	312	Set the callback function to call on a retryable system call error (such
88	as failed select, poll, epoll_wait). The message is a printable string	313	as failed select, poll, epoll_wait). The message is a printable string
89	indicating the system call or subsystem causing the problem. If this	314	indicating the system call or subsystem causing the problem. If this
90	callback is set, then libev will expect it to remedy the sitution, no	315	callback is set, then libev will expect it to remedy the situation, no
91	matter what, when it returns. That is, libev will generally retry the	316	matter what, when it returns. That is, libev will generally retry the
92	requested operation, or, if the condition doesn't go away, do bad stuff	317	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	318	(such as abort).
94		319
		320	Example: This is basically the same thing that libev does internally, too.
		321
		322	static void
		323	fatal_error (const char *msg)
		324	{
		325	perror (msg);
		326	abort ();
		327	}
		328
		329	...
		330	ev_set_syserr_cb (fatal_error);
		331
		332	=item ev_feed_signal (int signum)
		333
		334	This function can be used to "simulate" a signal receive. It is completely
		335	safe to call this function at any time, from any context, including signal
		336	handlers or random threads.
		337
		338	Its main use is to customise signal handling in your process, especially
		339	in the presence of threads. For example, you could block signals
		340	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		341	creating any loops), and in one thread, use C<sigwait> or any other
		342	mechanism to wait for signals, then "deliver" them to libev by calling
		343	C<ev_feed_signal>.
		344
95	=back	345	=back
96		346
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	347	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
98		348
99	An event loop is described by a C<struct ev_loop *>. The library knows two	349	An event loop is described by a C<struct ev_loop *> (the C<struct> is
100	types of such loops, the I<default> loop, which supports signals and child	350	I<not> optional in this case unless libev 3 compatibility is disabled, as
101	events, and dynamically created loops which do not.	351	libev 3 had an C<ev_loop> function colliding with the struct name).
102		352
103	If you use threads, a common model is to run the default event loop	353	The library knows two types of such loops, the I<default> loop, which
104	in your main thread (or in a separate thrad) and for each thread you	354	supports child process events, and dynamically created event loops which
105	create, you also create another event loop. Libev itself does no locking	355	do not.
106	whatsoever, so if you mix calls to the same event loop in different
107	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because its hideous and inefficient).
109		356
110	=over 4	357	=over 4
111		358
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	359	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		360
114	This will initialise the default event loop if it hasn't been initialised	361	This returns the "default" event loop object, which is what you should
115	yet and return it. If the default loop could not be initialised, returns	362	normally use when you just need "the event loop". Event loop objects and
116	false. If it already was initialised it simply returns it (and ignores the	363	the C<flags> parameter are described in more detail in the entry for
117	flags).	364	C<ev_loop_new>.
		365
		366	If the default loop is already initialised then this function simply
		367	returns it (and ignores the flags. If that is troubling you, check
		368	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		369	flags, which should almost always be C<0>, unless the caller is also the
		370	one calling C<ev_run> or otherwise qualifies as "the main program".
118		371
119	If you don't know what event loop to use, use the one returned from this	372	If you don't know what event loop to use, use the one returned from this
120	function.	373	function (or via the C<EV_DEFAULT> macro).
		374
		375	Note that this function is I<not> thread-safe, so if you want to use it
		376	from multiple threads, you have to employ some kind of mutex (note also
		377	that this case is unlikely, as loops cannot be shared easily between
		378	threads anyway).
		379
		380	The default loop is the only loop that can handle C<ev_child> watchers,
		381	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		382	a problem for your application you can either create a dynamic loop with
		383	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		384	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		385
		386	Example: This is the most typical usage.
		387
		388	if (!ev_default_loop (0))
		389	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		390
		391	Example: Restrict libev to the select and poll backends, and do not allow
		392	environment settings to be taken into account:
		393
		394	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		395
		396	=item struct ev_loop *ev_loop_new (unsigned int flags)
		397
		398	This will create and initialise a new event loop object. If the loop
		399	could not be initialised, returns false.
		400
		401	This function is thread-safe, and one common way to use libev with
		402	threads is indeed to create one loop per thread, and using the default
		403	loop in the "main" or "initial" thread.
121		404
122	The flags argument can be used to specify special behaviour or specific	405	The flags argument can be used to specify special behaviour or specific
123	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	406	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
124		407
125	It supports the following flags:	408	The following flags are supported:
126		409
127	=over 4	410	=over 4
128		411
129	=item EVFLAG_AUTO	412	=item C<EVFLAG_AUTO>
130		413
131	The default flags value. Use this if you have no clue (its the right	414	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	415	thing, believe me).
133		416
134	=item EVFLAG_NOENV	417	=item C<EVFLAG_NOENV>
135		418
136	If this flag bit is ored into the flag value (or the program runs setuid	419	If this flag bit is or'ed into the flag value (or the program runs setuid
137	or setgid) then libev will I<not> look at the environment variable	420	or setgid) then libev will I<not> look at the environment variable
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	421	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	422	override the flags completely if it is found in the environment. This is
140	useful to try out specific backends to test their performance, or to work	423	useful to try out specific backends to test their performance, to work
141	around bugs.	424	around bugs, or to make libev threadsafe (accessing environment variables
		425	cannot be done in a threadsafe way, but usually it works if no other
		426	thread modifies them).
142		427
		428	=item C<EVFLAG_FORKCHECK>
		429
		430	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		431	make libev check for a fork in each iteration by enabling this flag.
		432
		433	This works by calling C<getpid ()> on every iteration of the loop,
		434	and thus this might slow down your event loop if you do a lot of loop
		435	iterations and little real work, but is usually not noticeable (on my
		436	GNU/Linux system for example, C<getpid> is actually a simple 5-insn
		437	sequence without a system call and thus I<very> fast, but my GNU/Linux
		438	system also has C<pthread_atfork> which is even faster). (Update: glibc
		439	versions 2.25 apparently removed the C<getpid> optimisation again).
		440
		441	The big advantage of this flag is that you can forget about fork (and
		442	forget about forgetting to tell libev about forking, although you still
		443	have to ignore C<SIGPIPE>) when you use this flag.
		444
		445	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		446	environment variable.
		447
		448	=item C<EVFLAG_NOINOTIFY>
		449
		450	When this flag is specified, then libev will not attempt to use the
		451	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		452	testing, this flag can be useful to conserve inotify file descriptors, as
		453	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		454
		455	=item C<EVFLAG_SIGNALFD>
		456
		457	When this flag is specified, then libev will attempt to use the
		458	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		459	delivers signals synchronously, which makes it both faster and might make
		460	it possible to get the queued signal data. It can also simplify signal
		461	handling with threads, as long as you properly block signals in your
		462	threads that are not interested in handling them.
		463
		464	Signalfd will not be used by default as this changes your signal mask, and
		465	there are a lot of shoddy libraries and programs (glib's threadpool for
		466	example) that can't properly initialise their signal masks.
		467
		468	=item C<EVFLAG_NOSIGMASK>
		469
		470	When this flag is specified, then libev will avoid to modify the signal
		471	mask. Specifically, this means you have to make sure signals are unblocked
		472	when you want to receive them.
		473
		474	This behaviour is useful when you want to do your own signal handling, or
		475	want to handle signals only in specific threads and want to avoid libev
		476	unblocking the signals.
		477
		478	It's also required by POSIX in a threaded program, as libev calls
		479	C<sigprocmask>, whose behaviour is officially unspecified.
		480
		481	This flag's behaviour will become the default in future versions of libev.
		482
143	=item EVMETHOD_SELECT portable select backend	483	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		484
145	=item EVMETHOD_POLL poll backend (everywhere except windows)	485	This is your standard select(2) backend. Not I<completely> standard, as
		486	libev tries to roll its own fd_set with no limits on the number of fds,
		487	but if that fails, expect a fairly low limit on the number of fds when
		488	using this backend. It doesn't scale too well (O(highest_fd)), but its
		489	usually the fastest backend for a low number of (low-numbered :) fds.
146		490
147	=item EVMETHOD_EPOLL linux only	491	To get good performance out of this backend you need a high amount of
		492	parallelism (most of the file descriptors should be busy). If you are
		493	writing a server, you should C<accept ()> in a loop to accept as many
		494	connections as possible during one iteration. You might also want to have
		495	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		496	readiness notifications you get per iteration.
148		497
149	=item EVMETHOD_KQUEUE some bsds only	498	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		499	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		500	C<exceptfds> set on that platform).
150		501
151	=item EVMETHOD_DEVPOLL solaris 8 only	502	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
152		503
153	=item EVMETHOD_PORT solaris 10 only	504	And this is your standard poll(2) backend. It's more complicated
		505	than select, but handles sparse fds better and has no artificial
		506	limit on the number of fds you can use (except it will slow down
		507	considerably with a lot of inactive fds). It scales similarly to select,
		508	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		509	performance tips.
154		510
155	If one or more of these are ored into the flags value, then only these	511	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
156	backends will be tried (in the reverse order as given here). If one are	512	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157	specified, any backend will do.	513
		514	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		515
		516	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		517	kernels).
		518
		519	For few fds, this backend is a bit little slower than poll and select, but
		520	it scales phenomenally better. While poll and select usually scale like
		521	O(total_fds) where total_fds is the total number of fds (or the highest
		522	fd), epoll scales either O(1) or O(active_fds).
		523
		524	The epoll mechanism deserves honorable mention as the most misdesigned
		525	of the more advanced event mechanisms: mere annoyances include silently
		526	dropping file descriptors, requiring a system call per change per file
		527	descriptor (and unnecessary guessing of parameters), problems with dup,
		528	returning before the timeout value, resulting in additional iterations
		529	(and only giving 5ms accuracy while select on the same platform gives
		530	0.1ms) and so on. The biggest issue is fork races, however - if a program
		531	forks then I<both> parent and child process have to recreate the epoll
		532	set, which can take considerable time (one syscall per file descriptor)
		533	and is of course hard to detect.
		534
		535	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
		536	but of course I<doesn't>, and epoll just loves to report events for
		537	totally I<different> file descriptors (even already closed ones, so
		538	one cannot even remove them from the set) than registered in the set
		539	(especially on SMP systems). Libev tries to counter these spurious
		540	notifications by employing an additional generation counter and comparing
		541	that against the events to filter out spurious ones, recreating the set
		542	when required. Epoll also erroneously rounds down timeouts, but gives you
		543	no way to know when and by how much, so sometimes you have to busy-wait
		544	because epoll returns immediately despite a nonzero timeout. And last
		545	not least, it also refuses to work with some file descriptors which work
		546	perfectly fine with C<select> (files, many character devices...).
		547
		548	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		549	cobbled together in a hurry, no thought to design or interaction with
		550	others. Oh, the pain, will it ever stop...
		551
		552	While stopping, setting and starting an I/O watcher in the same iteration
		553	will result in some caching, there is still a system call per such
		554	incident (because the same I<file descriptor> could point to a different
		555	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		556	file descriptors might not work very well if you register events for both
		557	file descriptors.
		558
		559	Best performance from this backend is achieved by not unregistering all
		560	watchers for a file descriptor until it has been closed, if possible,
		561	i.e. keep at least one watcher active per fd at all times. Stopping and
		562	starting a watcher (without re-setting it) also usually doesn't cause
		563	extra overhead. A fork can both result in spurious notifications as well
		564	as in libev having to destroy and recreate the epoll object, which can
		565	take considerable time and thus should be avoided.
		566
		567	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		568	faster than epoll for maybe up to a hundred file descriptors, depending on
		569	the usage. So sad.
		570
		571	While nominally embeddable in other event loops, this feature is broken in
		572	all kernel versions tested so far.
		573
		574	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		575	C<EVBACKEND_POLL>.
		576
		577	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		578
		579	Kqueue deserves special mention, as at the time of this writing, it
		580	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		581	with anything but sockets and pipes, except on Darwin, where of course
		582	it's completely useless). Unlike epoll, however, whose brokenness
		583	is by design, these kqueue bugs can (and eventually will) be fixed
		584	without API changes to existing programs. For this reason it's not being
		585	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		586	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		587	system like NetBSD.
		588
		589	You still can embed kqueue into a normal poll or select backend and use it
		590	only for sockets (after having made sure that sockets work with kqueue on
		591	the target platform). See C<ev_embed> watchers for more info.
		592
		593	It scales in the same way as the epoll backend, but the interface to the
		594	kernel is more efficient (which says nothing about its actual speed, of
		595	course). While stopping, setting and starting an I/O watcher does never
		596	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		597	two event changes per incident. Support for C<fork ()> is very bad (you
		598	might have to leak fd's on fork, but it's more sane than epoll) and it
		599	drops fds silently in similarly hard-to-detect cases.
		600
		601	This backend usually performs well under most conditions.
		602
		603	While nominally embeddable in other event loops, this doesn't work
		604	everywhere, so you might need to test for this. And since it is broken
		605	almost everywhere, you should only use it when you have a lot of sockets
		606	(for which it usually works), by embedding it into another event loop
		607	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		608	also broken on OS X)) and, did I mention it, using it only for sockets.
		609
		610	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		611	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		612	C<NOTE_EOF>.
		613
		614	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		615
		616	This is not implemented yet (and might never be, unless you send me an
		617	implementation). According to reports, C</dev/poll> only supports sockets
		618	and is not embeddable, which would limit the usefulness of this backend
		619	immensely.
		620
		621	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		622
		623	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		624	it's really slow, but it still scales very well (O(active_fds)).
		625
		626	While this backend scales well, it requires one system call per active
		627	file descriptor per loop iteration. For small and medium numbers of file
		628	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		629	might perform better.
		630
		631	On the positive side, this backend actually performed fully to
		632	specification in all tests and is fully embeddable, which is a rare feat
		633	among the OS-specific backends (I vastly prefer correctness over speed
		634	hacks).
		635
		636	On the negative side, the interface is I<bizarre> - so bizarre that
		637	even sun itself gets it wrong in their code examples: The event polling
		638	function sometimes returns events to the caller even though an error
		639	occurred, but with no indication whether it has done so or not (yes, it's
		640	even documented that way) - deadly for edge-triggered interfaces where you
		641	absolutely have to know whether an event occurred or not because you have
		642	to re-arm the watcher.
		643
		644	Fortunately libev seems to be able to work around these idiocies.
		645
		646	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		647	C<EVBACKEND_POLL>.
		648
		649	=item C<EVBACKEND_ALL>
		650
		651	Try all backends (even potentially broken ones that wouldn't be tried
		652	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		653	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		654
		655	It is definitely not recommended to use this flag, use whatever
		656	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		657	at all.
		658
		659	=item C<EVBACKEND_MASK>
		660
		661	Not a backend at all, but a mask to select all backend bits from a
		662	C<flags> value, in case you want to mask out any backends from a flags
		663	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
158		664
159	=back	665	=back
160		666
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	667	If one or more of the backend flags are or'ed into the flags value,
		668	then only these backends will be tried (in the reverse order as listed
		669	here). If none are specified, all backends in C<ev_recommended_backends
		670	()> will be tried.
162		671
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	672	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		673
168	=item ev_default_destroy ()	674	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		675	if (!epoller)
		676	fatal ("no epoll found here, maybe it hides under your chair");
169		677
170	Destroys the default loop again (frees all memory and kernel state	678	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	679	used if available.
172	any way whatsoever, although you cnanot rely on this :).	680
		681	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
173		682
174	=item ev_loop_destroy (loop)	683	=item ev_loop_destroy (loop)
175		684
176	Like C<ev_default_destroy>, but destroys an event loop created by an	685	Destroys an event loop object (frees all memory and kernel state
177	earlier call to C<ev_loop_new>.	686	etc.). None of the active event watchers will be stopped in the normal
		687	sense, so e.g. C<ev_is_active> might still return true. It is your
		688	responsibility to either stop all watchers cleanly yourself I<before>
		689	calling this function, or cope with the fact afterwards (which is usually
		690	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		691	for example).
178		692
179	=item ev_default_fork ()	693	Note that certain global state, such as signal state (and installed signal
		694	handlers), will not be freed by this function, and related watchers (such
		695	as signal and child watchers) would need to be stopped manually.
180		696
181	This function reinitialises the kernel state for backends that have	697	This function is normally used on loop objects allocated by
182	one. Despite the name, you can call it anytime, but it makes most sense	698	C<ev_loop_new>, but it can also be used on the default loop returned by
183	after forking, in either the parent or child process (or both, but that	699	C<ev_default_loop>, in which case it is not thread-safe.
184	again makes little sense).
185		700
186	You I<must> call this function after forking if and only if you want to	701	Note that it is not advisable to call this function on the default loop
187	use the event library in both processes. If you just fork+exec, you don't	702	except in the rare occasion where you really need to free its resources.
188	have to call it.	703	If you need dynamically allocated loops it is better to use C<ev_loop_new>
189		704	and C<ev_loop_destroy>.
190	The function itself is quite fast and its usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in
192	quite nicely into a call to C<pthread_atfork>:
193
194	pthread_atfork (0, 0, ev_default_fork);
195		705
196	=item ev_loop_fork (loop)	706	=item ev_loop_fork (loop)
197		707
198	Like C<ev_default_fork>, but acts on an event loop created by	708	This function sets a flag that causes subsequent C<ev_run> iterations
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	709	to reinitialise the kernel state for backends that have one. Despite
200	after fork, and how you do this is entirely your own problem.	710	the name, you can call it anytime you are allowed to start or stop
		711	watchers (except inside an C<ev_prepare> callback), but it makes most
		712	sense after forking, in the child process. You I<must> call it (or use
		713	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
201		714
		715	In addition, if you want to reuse a loop (via this function or
		716	C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
		717
		718	Again, you I<have> to call it on I<any> loop that you want to re-use after
		719	a fork, I<even if you do not plan to use the loop in the parent>. This is
		720	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		721	during fork.
		722
		723	On the other hand, you only need to call this function in the child
		724	process if and only if you want to use the event loop in the child. If
		725	you just fork+exec or create a new loop in the child, you don't have to
		726	call it at all (in fact, C<epoll> is so badly broken that it makes a
		727	difference, but libev will usually detect this case on its own and do a
		728	costly reset of the backend).
		729
		730	The function itself is quite fast and it's usually not a problem to call
		731	it just in case after a fork.
		732
		733	Example: Automate calling C<ev_loop_fork> on the default loop when
		734	using pthreads.
		735
		736	static void
		737	post_fork_child (void)
		738	{
		739	ev_loop_fork (EV_DEFAULT);
		740	}
		741
		742	...
		743	pthread_atfork (0, 0, post_fork_child);
		744
		745	=item int ev_is_default_loop (loop)
		746
		747	Returns true when the given loop is, in fact, the default loop, and false
		748	otherwise.
		749
		750	=item unsigned int ev_iteration (loop)
		751
		752	Returns the current iteration count for the event loop, which is identical
		753	to the number of times libev did poll for new events. It starts at C<0>
		754	and happily wraps around with enough iterations.
		755
		756	This value can sometimes be useful as a generation counter of sorts (it
		757	"ticks" the number of loop iterations), as it roughly corresponds with
		758	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		759	prepare and check phases.
		760
202	=item unsigned int ev_method (loop)	761	=item unsigned int ev_depth (loop)
203		762
		763	Returns the number of times C<ev_run> was entered minus the number of
		764	times C<ev_run> was exited normally, in other words, the recursion depth.
		765
		766	Outside C<ev_run>, this number is zero. In a callback, this number is
		767	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		768	in which case it is higher.
		769
		770	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		771	throwing an exception etc.), doesn't count as "exit" - consider this
		772	as a hint to avoid such ungentleman-like behaviour unless it's really
		773	convenient, in which case it is fully supported.
		774
		775	=item unsigned int ev_backend (loop)
		776
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	777	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	778	use.
206		779
207	=item ev_tstamp = ev_now (loop)	780	=item ev_tstamp ev_now (loop)
208		781
209	Returns the current "event loop time", which is the time the event loop	782	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	783	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	784	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	785	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).	786	event occurring (or more correctly, libev finding out about it).
214		787
		788	=item ev_now_update (loop)
		789
		790	Establishes the current time by querying the kernel, updating the time
		791	returned by C<ev_now ()> in the progress. This is a costly operation and
		792	is usually done automatically within C<ev_run ()>.
		793
		794	This function is rarely useful, but when some event callback runs for a
		795	very long time without entering the event loop, updating libev's idea of
		796	the current time is a good idea.
		797
		798	See also L</The special problem of time updates> in the C<ev_timer> section.
		799
		800	=item ev_suspend (loop)
		801
		802	=item ev_resume (loop)
		803
		804	These two functions suspend and resume an event loop, for use when the
		805	loop is not used for a while and timeouts should not be processed.
		806
		807	A typical use case would be an interactive program such as a game: When
		808	the user presses C<^Z> to suspend the game and resumes it an hour later it
		809	would be best to handle timeouts as if no time had actually passed while
		810	the program was suspended. This can be achieved by calling C<ev_suspend>
		811	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		812	C<ev_resume> directly afterwards to resume timer processing.
		813
		814	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		815	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		816	will be rescheduled (that is, they will lose any events that would have
		817	occurred while suspended).
		818
		819	After calling C<ev_suspend> you B<must not> call I<any> function on the
		820	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		821	without a previous call to C<ev_suspend>.
		822
		823	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		824	event loop time (see C<ev_now_update>).
		825
215	=item ev_loop (loop, int flags)	826	=item bool ev_run (loop, int flags)
216		827
217	Finally, this is it, the event handler. This function usually is called	828	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	829	after you have initialised all your watchers and you want to start
219	events.	830	handling events. It will ask the operating system for any new events, call
		831	the watcher callbacks, and then repeat the whole process indefinitely: This
		832	is why event loops are called I<loops>.
220		833
221	If the flags argument is specified as 0, it will not return until either	834	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.	835	until either no event watchers are active anymore or C<ev_break> was
		836	called.
223		837
		838	The return value is false if there are no more active watchers (which
		839	usually means "all jobs done" or "deadlock"), and true in all other cases
		840	(which usually means " you should call C<ev_run> again").
		841
		842	Please note that an explicit C<ev_break> is usually better than
		843	relying on all watchers to be stopped when deciding when a program has
		844	finished (especially in interactive programs), but having a program
		845	that automatically loops as long as it has to and no longer by virtue
		846	of relying on its watchers stopping correctly, that is truly a thing of
		847	beauty.
		848
		849	This function is I<mostly> exception-safe - you can break out of a
		850	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		851	exception and so on. This does not decrement the C<ev_depth> value, nor
		852	will it clear any outstanding C<EVBREAK_ONE> breaks.
		853
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	854	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	855	those events and any already outstanding ones, but will not wait and
226	case there are no events.	856	block your process in case there are no events and will return after one
		857	iteration of the loop. This is sometimes useful to poll and handle new
		858	events while doing lengthy calculations, to keep the program responsive.
227		859
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	860	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	861	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives.	862	will block your process until at least one new event arrives (which could
		863	be an event internal to libev itself, so there is no guarantee that a
		864	user-registered callback will be called), and will return after one
		865	iteration of the loop.
231		866
232	This flags value could be used to implement alternative looping	867	This is useful if you are waiting for some external event in conjunction
233	constructs, but the C<prepare> and C<check> watchers provide a better and	868	with something not expressible using other libev watchers (i.e. "roll your
234	more generic mechanism.	869	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		870	usually a better approach for this kind of thing.
235		871
		872	Here are the gory details of what C<ev_run> does (this is for your
		873	understanding, not a guarantee that things will work exactly like this in
		874	future versions):
		875
		876	- Increment loop depth.
		877	- Reset the ev_break status.
		878	- Before the first iteration, call any pending watchers.
		879	LOOP:
		880	- If EVFLAG_FORKCHECK was used, check for a fork.
		881	- If a fork was detected (by any means), queue and call all fork watchers.
		882	- Queue and call all prepare watchers.
		883	- If ev_break was called, goto FINISH.
		884	- If we have been forked, detach and recreate the kernel state
		885	as to not disturb the other process.
		886	- Update the kernel state with all outstanding changes.
		887	- Update the "event loop time" (ev_now ()).
		888	- Calculate for how long to sleep or block, if at all
		889	(active idle watchers, EVRUN_NOWAIT or not having
		890	any active watchers at all will result in not sleeping).
		891	- Sleep if the I/O and timer collect interval say so.
		892	- Increment loop iteration counter.
		893	- Block the process, waiting for any events.
		894	- Queue all outstanding I/O (fd) events.
		895	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		896	- Queue all expired timers.
		897	- Queue all expired periodics.
		898	- Queue all idle watchers with priority higher than that of pending events.
		899	- Queue all check watchers.
		900	- Call all queued watchers in reverse order (i.e. check watchers first).
		901	Signals and child watchers are implemented as I/O watchers, and will
		902	be handled here by queueing them when their watcher gets executed.
		903	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		904	were used, or there are no active watchers, goto FINISH, otherwise
		905	continue with step LOOP.
		906	FINISH:
		907	- Reset the ev_break status iff it was EVBREAK_ONE.
		908	- Decrement the loop depth.
		909	- Return.
		910
		911	Example: Queue some jobs and then loop until no events are outstanding
		912	anymore.
		913
		914	... queue jobs here, make sure they register event watchers as long
		915	... as they still have work to do (even an idle watcher will do..)
		916	ev_run (my_loop, 0);
		917	... jobs done or somebody called break. yeah!
		918
236	=item ev_unloop (loop, how)	919	=item ev_break (loop, how)
237		920
238	Can be used to make a call to C<ev_loop> return early. The C<how> argument	921	Can be used to make a call to C<ev_run> return early (but only after it
239	must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop>	922	has processed all outstanding events). The C<how> argument must be either
240	call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop>	923	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
241	calls return.	924	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		925
		926	This "break state" will be cleared on the next call to C<ev_run>.
		927
		928	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		929	which case it will have no effect.
242		930
243	=item ev_ref (loop)	931	=item ev_ref (loop)
244		932
245	=item ev_unref (loop)	933	=item ev_unref (loop)
246		934
247	Ref/unref can be used to add or remove a refcount on the event loop: Every	935	Ref/unref can be used to add or remove a reference count on the event
248	watcher keeps one reference. If you have a long-runing watcher you never	936	loop: Every watcher keeps one reference, and as long as the reference
249	unregister that should not keep ev_loop from running, ev_unref() after	937	count is nonzero, C<ev_run> will not return on its own.
250	starting, and ev_ref() before stopping it. Libev itself uses this for	938
251	example for its internal signal pipe: It is not visible to you as a user	939	This is useful when you have a watcher that you never intend to
252	and should not keep C<ev_loop> from exiting if the work is done. It is	940	unregister, but that nevertheless should not keep C<ev_run> from
		941	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		942	before stopping it.
		943
		944	As an example, libev itself uses this for its internal signal pipe: It
		945	is not visible to the libev user and should not keep C<ev_run> from
		946	exiting if no event watchers registered by it are active. It is also an
253	also an excellent way to do this for generic recurring timers or from	947	excellent way to do this for generic recurring timers or from within
254	within third-party libraries. Just remember to unref after start and ref	948	third-party libraries. Just remember to I<unref after start> and I<ref
255	before stop.	949	before stop> (but only if the watcher wasn't active before, or was active
		950	before, respectively. Note also that libev might stop watchers itself
		951	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		952	in the callback).
		953
		954	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		955	running when nothing else is active.
		956
		957	ev_signal exitsig;
		958	ev_signal_init (&exitsig, sig_cb, SIGINT);
		959	ev_signal_start (loop, &exitsig);
		960	ev_unref (loop);
		961
		962	Example: For some weird reason, unregister the above signal handler again.
		963
		964	ev_ref (loop);
		965	ev_signal_stop (loop, &exitsig);
		966
		967	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		968
		969	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		970
		971	These advanced functions influence the time that libev will spend waiting
		972	for events. Both time intervals are by default C<0>, meaning that libev
		973	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		974	latency.
		975
		976	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		977	allows libev to delay invocation of I/O and timer/periodic callbacks
		978	to increase efficiency of loop iterations (or to increase power-saving
		979	opportunities).
		980
		981	The idea is that sometimes your program runs just fast enough to handle
		982	one (or very few) event(s) per loop iteration. While this makes the
		983	program responsive, it also wastes a lot of CPU time to poll for new
		984	events, especially with backends like C<select ()> which have a high
		985	overhead for the actual polling but can deliver many events at once.
		986
		987	By setting a higher I<io collect interval> you allow libev to spend more
		988	time collecting I/O events, so you can handle more events per iteration,
		989	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		990	C<ev_timer>) will not be affected. Setting this to a non-null value will
		991	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		992	sleep time ensures that libev will not poll for I/O events more often then
		993	once per this interval, on average (as long as the host time resolution is
		994	good enough).
		995
		996	Likewise, by setting a higher I<timeout collect interval> you allow libev
		997	to spend more time collecting timeouts, at the expense of increased
		998	latency/jitter/inexactness (the watcher callback will be called
		999	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		1000	value will not introduce any overhead in libev.
		1001
		1002	Many (busy) programs can usually benefit by setting the I/O collect
		1003	interval to a value near C<0.1> or so, which is often enough for
		1004	interactive servers (of course not for games), likewise for timeouts. It
		1005	usually doesn't make much sense to set it to a lower value than C<0.01>,
		1006	as this approaches the timing granularity of most systems. Note that if
		1007	you do transactions with the outside world and you can't increase the
		1008	parallelity, then this setting will limit your transaction rate (if you
		1009	need to poll once per transaction and the I/O collect interval is 0.01,
		1010	then you can't do more than 100 transactions per second).
		1011
		1012	Setting the I<timeout collect interval> can improve the opportunity for
		1013	saving power, as the program will "bundle" timer callback invocations that
		1014	are "near" in time together, by delaying some, thus reducing the number of
		1015	times the process sleeps and wakes up again. Another useful technique to
		1016	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		1017	they fire on, say, one-second boundaries only.
		1018
		1019	Example: we only need 0.1s timeout granularity, and we wish not to poll
		1020	more often than 100 times per second:
		1021
		1022	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		1023	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		1024
		1025	=item ev_invoke_pending (loop)
		1026
		1027	This call will simply invoke all pending watchers while resetting their
		1028	pending state. Normally, C<ev_run> does this automatically when required,
		1029	but when overriding the invoke callback this call comes handy. This
		1030	function can be invoked from a watcher - this can be useful for example
		1031	when you want to do some lengthy calculation and want to pass further
		1032	event handling to another thread (you still have to make sure only one
		1033	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		1034
		1035	=item int ev_pending_count (loop)
		1036
		1037	Returns the number of pending watchers - zero indicates that no watchers
		1038	are pending.
		1039
		1040	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1041
		1042	This overrides the invoke pending functionality of the loop: Instead of
		1043	invoking all pending watchers when there are any, C<ev_run> will call
		1044	this callback instead. This is useful, for example, when you want to
		1045	invoke the actual watchers inside another context (another thread etc.).
		1046
		1047	If you want to reset the callback, use C<ev_invoke_pending> as new
		1048	callback.
		1049
		1050	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
		1051
		1052	Sometimes you want to share the same loop between multiple threads. This
		1053	can be done relatively simply by putting mutex_lock/unlock calls around
		1054	each call to a libev function.
		1055
		1056	However, C<ev_run> can run an indefinite time, so it is not feasible
		1057	to wait for it to return. One way around this is to wake up the event
		1058	loop via C<ev_break> and C<ev_async_send>, another way is to set these
		1059	I<release> and I<acquire> callbacks on the loop.
		1060
		1061	When set, then C<release> will be called just before the thread is
		1062	suspended waiting for new events, and C<acquire> is called just
		1063	afterwards.
		1064
		1065	Ideally, C<release> will just call your mutex_unlock function, and
		1066	C<acquire> will just call the mutex_lock function again.
		1067
		1068	While event loop modifications are allowed between invocations of
		1069	C<release> and C<acquire> (that's their only purpose after all), no
		1070	modifications done will affect the event loop, i.e. adding watchers will
		1071	have no effect on the set of file descriptors being watched, or the time
		1072	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1073	to take note of any changes you made.
		1074
		1075	In theory, threads executing C<ev_run> will be async-cancel safe between
		1076	invocations of C<release> and C<acquire>.
		1077
		1078	See also the locking example in the C<THREADS> section later in this
		1079	document.
		1080
		1081	=item ev_set_userdata (loop, void *data)
		1082
		1083	=item void *ev_userdata (loop)
		1084
		1085	Set and retrieve a single C<void *> associated with a loop. When
		1086	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1087	C<0>.
		1088
		1089	These two functions can be used to associate arbitrary data with a loop,
		1090	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1091	C<acquire> callbacks described above, but of course can be (ab-)used for
		1092	any other purpose as well.
		1093
		1094	=item ev_verify (loop)
		1095
		1096	This function only does something when C<EV_VERIFY> support has been
		1097	compiled in, which is the default for non-minimal builds. It tries to go
		1098	through all internal structures and checks them for validity. If anything
		1099	is found to be inconsistent, it will print an error message to standard
		1100	error and call C<abort ()>.
		1101
		1102	This can be used to catch bugs inside libev itself: under normal
		1103	circumstances, this function will never abort as of course libev keeps its
		1104	data structures consistent.
256		1105
257	=back	1106	=back
258		1107
		1108
259	=head1 ANATOMY OF A WATCHER	1109	=head1 ANATOMY OF A WATCHER
260		1110
		1111	In the following description, uppercase C<TYPE> in names stands for the
		1112	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1113	watchers and C<ev_io_start> for I/O watchers.
		1114
261	A watcher is a structure that you create and register to record your	1115	A watcher is an opaque structure that you allocate and register to record
262	interest in some event. For instance, if you want to wait for STDIN to	1116	your interest in some event. To make a concrete example, imagine you want
263	become readable, you would create an ev_io watcher for that:	1117	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1118	for that:
264		1119
265	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1120	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
266	{	1121	{
267	ev_io_stop (w);	1122	ev_io_stop (w);
268	ev_unloop (loop, EVUNLOOP_ALL);	1123	ev_break (loop, EVBREAK_ALL);
269	}	1124	}
270		1125
271	struct ev_loop *loop = ev_default_loop (0);	1126	struct ev_loop *loop = ev_default_loop (0);
		1127
272	struct ev_io stdin_watcher;	1128	ev_io stdin_watcher;
		1129
273	ev_init (&stdin_watcher, my_cb);	1130	ev_init (&stdin_watcher, my_cb);
274	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1131	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
275	ev_io_start (loop, &stdin_watcher);	1132	ev_io_start (loop, &stdin_watcher);
		1133
276	ev_loop (loop, 0);	1134	ev_run (loop, 0);
277		1135
278	As you can see, you are responsible for allocating the memory for your	1136	As you can see, you are responsible for allocating the memory for your
279	watcher structures (and it is usually a bad idea to do this on the stack,	1137	watcher structures (and it is I<usually> a bad idea to do this on the
280	although this can sometimes be quite valid).	1138	stack).
281		1139
		1140	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1141	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1142
282	Each watcher structure must be initialised by a call to C<ev_init	1143	Each watcher structure must be initialised by a call to C<ev_init (watcher
283	(watcher *, callback)>, which expects a callback to be provided. This	1144	*, callback)>, which expects a callback to be provided. This callback is
284	callback gets invoked each time the event occurs (or, in the case of io	1145	invoked each time the event occurs (or, in the case of I/O watchers, each
285	watchers, each time the event loop detects that the file descriptor given	1146	time the event loop detects that the file descriptor given is readable
286	is readable and/or writable).	1147	and/or writable).
287		1148
288	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1149	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
289	with arguments specific to this watcher type. There is also a macro	1150	macro to configure it, with arguments specific to the watcher type. There
290	to combine initialisation and setting in one call: C<< ev_<type>_init	1151	is also a macro to combine initialisation and setting in one call: C<<
291	(watcher *, callback, ...) >>.	1152	ev_TYPE_init (watcher *, callback, ...) >>.
292		1153
293	To make the watcher actually watch out for events, you have to start it	1154	To make the watcher actually watch out for events, you have to start it
294	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1155	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
295	*) >>), and you can stop watching for events at any time by calling the	1156	*) >>), and you can stop watching for events at any time by calling the
296	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1157	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
297		1158
298	As long as your watcher is active (has been started but not stopped) you	1159	As long as your watcher is active (has been started but not stopped) you
299	must not touch the values stored in it. Most specifically you must never	1160	must not touch the values stored in it. Most specifically you must never
300	reinitialise it or call its set method.	1161	reinitialise it or call its C<ev_TYPE_set> macro.
301
302	You cna check whether an event is active by calling the C<ev_is_active
303	(watcher *)> macro. To see whether an event is outstanding (but the
304	callback for it has not been called yet) you cna use the C<ev_is_pending
305	(watcher *)> macro.
306		1162
307	Each and every callback receives the event loop pointer as first, the	1163	Each and every callback receives the event loop pointer as first, the
308	registered watcher structure as second, and a bitset of received events as	1164	registered watcher structure as second, and a bitset of received events as
309	third argument.	1165	third argument.
310		1166
311	The rceeived events usually include a single bit per event type received	1167	The received events usually include a single bit per event type received
312	(you can receive multiple events at the same time). The possible bit masks	1168	(you can receive multiple events at the same time). The possible bit masks
313	are:	1169	are:
314		1170
315	=over 4	1171	=over 4
316		1172
317	=item EV_READ	1173	=item C<EV_READ>
318		1174
319	=item EV_WRITE	1175	=item C<EV_WRITE>
320		1176
321	The file descriptor in the ev_io watcher has become readable and/or	1177	The file descriptor in the C<ev_io> watcher has become readable and/or
322	writable.	1178	writable.
323		1179
324	=item EV_TIMEOUT	1180	=item C<EV_TIMER>
325		1181
326	The ev_timer watcher has timed out.	1182	The C<ev_timer> watcher has timed out.
327		1183
328	=item EV_PERIODIC	1184	=item C<EV_PERIODIC>
329		1185
330	The ev_periodic watcher has timed out.	1186	The C<ev_periodic> watcher has timed out.
331		1187
332	=item EV_SIGNAL	1188	=item C<EV_SIGNAL>
333		1189
334	The signal specified in the ev_signal watcher has been received by a thread.	1190	The signal specified in the C<ev_signal> watcher has been received by a thread.
335		1191
336	=item EV_CHILD	1192	=item C<EV_CHILD>
337		1193
338	The pid specified in the ev_child watcher has received a status change.	1194	The pid specified in the C<ev_child> watcher has received a status change.
339		1195
		1196	=item C<EV_STAT>
		1197
		1198	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1199
340	=item EV_IDLE	1200	=item C<EV_IDLE>
341		1201
342	The ev_idle watcher has determined that you have nothing better to do.	1202	The C<ev_idle> watcher has determined that you have nothing better to do.
343		1203
344	=item EV_PREPARE	1204	=item C<EV_PREPARE>
345		1205
346	=item EV_CHECK	1206	=item C<EV_CHECK>
347		1207
348	All ev_prepare watchers are invoked just I<before> C<ev_loop> starts	1208	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
349	to gather new events, and all ev_check watchers are invoked just after	1209	gather new events, and all C<ev_check> watchers are queued (not invoked)
350	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1210	just after C<ev_run> has gathered them, but before it queues any callbacks
		1211	for any received events. That means C<ev_prepare> watchers are the last
		1212	watchers invoked before the event loop sleeps or polls for new events, and
		1213	C<ev_check> watchers will be invoked before any other watchers of the same
		1214	or lower priority within an event loop iteration.
		1215
351	received events. Callbacks of both watcher types can start and stop as	1216	Callbacks of both watcher types can start and stop as many watchers as
352	many watchers as they want, and all of them will be taken into account	1217	they want, and all of them will be taken into account (for example, a
353	(for example, a ev_prepare watcher might start an idle watcher to keep	1218	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
354	C<ev_loop> from blocking).	1219	blocking).
355		1220
		1221	=item C<EV_EMBED>
		1222
		1223	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1224
		1225	=item C<EV_FORK>
		1226
		1227	The event loop has been resumed in the child process after fork (see
		1228	C<ev_fork>).
		1229
		1230	=item C<EV_CLEANUP>
		1231
		1232	The event loop is about to be destroyed (see C<ev_cleanup>).
		1233
		1234	=item C<EV_ASYNC>
		1235
		1236	The given async watcher has been asynchronously notified (see C<ev_async>).
		1237
		1238	=item C<EV_CUSTOM>
		1239
		1240	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1241	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1242
356	=item EV_ERROR	1243	=item C<EV_ERROR>
357		1244
358	An unspecified error has occured, the watcher has been stopped. This might	1245	An unspecified error has occurred, the watcher has been stopped. This might
359	happen because the watcher could not be properly started because libev	1246	happen because the watcher could not be properly started because libev
360	ran out of memory, a file descriptor was found to be closed or any other	1247	ran out of memory, a file descriptor was found to be closed or any other
		1248	problem. Libev considers these application bugs.
		1249
361	problem. You best act on it by reporting the problem and somehow coping	1250	You best act on it by reporting the problem and somehow coping with the
362	with the watcher being stopped.	1251	watcher being stopped. Note that well-written programs should not receive
		1252	an error ever, so when your watcher receives it, this usually indicates a
		1253	bug in your program.
363		1254
364	Libev will usually signal a few "dummy" events together with an error,	1255	Libev will usually signal a few "dummy" events together with an error, for
365	for example it might indicate that a fd is readable or writable, and if	1256	example it might indicate that a fd is readable or writable, and if your
366	your callbacks is well-written it can just attempt the operation and cope	1257	callbacks is well-written it can just attempt the operation and cope with
367	with the error from read() or write(). This will not work in multithreaded	1258	the error from read() or write(). This will not work in multi-threaded
368	programs, though, so beware.	1259	programs, though, as the fd could already be closed and reused for another
		1260	thing, so beware.
369		1261
370	=back	1262	=back
371		1263
372	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1264	=head2 GENERIC WATCHER FUNCTIONS
373		1265
374	Each watcher has, by default, a member C<void *data> that you can change	1266	=over 4
375	and read at any time, libev will completely ignore it. This cna be used
376	to associate arbitrary data with your watcher. If you need more data and
377	don't want to allocate memory and store a pointer to it in that data
378	member, you can also "subclass" the watcher type and provide your own
379	data:
380		1267
381	struct my_io	1268	=item C<ev_init> (ev_TYPE *watcher, callback)
		1269
		1270	This macro initialises the generic portion of a watcher. The contents
		1271	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1272	the generic parts of the watcher are initialised, you I<need> to call
		1273	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1274	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1275	which rolls both calls into one.
		1276
		1277	You can reinitialise a watcher at any time as long as it has been stopped
		1278	(or never started) and there are no pending events outstanding.
		1279
		1280	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1281	int revents)>.
		1282
		1283	Example: Initialise an C<ev_io> watcher in two steps.
		1284
		1285	ev_io w;
		1286	ev_init (&w, my_cb);
		1287	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1288
		1289	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1290
		1291	This macro initialises the type-specific parts of a watcher. You need to
		1292	call C<ev_init> at least once before you call this macro, but you can
		1293	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1294	macro on a watcher that is active (it can be pending, however, which is a
		1295	difference to the C<ev_init> macro).
		1296
		1297	Although some watcher types do not have type-specific arguments
		1298	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1299
		1300	See C<ev_init>, above, for an example.
		1301
		1302	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1303
		1304	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1305	calls into a single call. This is the most convenient method to initialise
		1306	a watcher. The same limitations apply, of course.
		1307
		1308	Example: Initialise and set an C<ev_io> watcher in one step.
		1309
		1310	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1311
		1312	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1313
		1314	Starts (activates) the given watcher. Only active watchers will receive
		1315	events. If the watcher is already active nothing will happen.
		1316
		1317	Example: Start the C<ev_io> watcher that is being abused as example in this
		1318	whole section.
		1319
		1320	ev_io_start (EV_DEFAULT_UC, &w);
		1321
		1322	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1323
		1324	Stops the given watcher if active, and clears the pending status (whether
		1325	the watcher was active or not).
		1326
		1327	It is possible that stopped watchers are pending - for example,
		1328	non-repeating timers are being stopped when they become pending - but
		1329	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1330	pending. If you want to free or reuse the memory used by the watcher it is
		1331	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1332
		1333	=item bool ev_is_active (ev_TYPE *watcher)
		1334
		1335	Returns a true value iff the watcher is active (i.e. it has been started
		1336	and not yet been stopped). As long as a watcher is active you must not modify
		1337	it.
		1338
		1339	=item bool ev_is_pending (ev_TYPE *watcher)
		1340
		1341	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1342	events but its callback has not yet been invoked). As long as a watcher
		1343	is pending (but not active) you must not call an init function on it (but
		1344	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1345	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1346	it).
		1347
		1348	=item callback ev_cb (ev_TYPE *watcher)
		1349
		1350	Returns the callback currently set on the watcher.
		1351
		1352	=item ev_set_cb (ev_TYPE *watcher, callback)
		1353
		1354	Change the callback. You can change the callback at virtually any time
		1355	(modulo threads).
		1356
		1357	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1358
		1359	=item int ev_priority (ev_TYPE *watcher)
		1360
		1361	Set and query the priority of the watcher. The priority is a small
		1362	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1363	(default: C<-2>). Pending watchers with higher priority will be invoked
		1364	before watchers with lower priority, but priority will not keep watchers
		1365	from being executed (except for C<ev_idle> watchers).
		1366
		1367	If you need to suppress invocation when higher priority events are pending
		1368	you need to look at C<ev_idle> watchers, which provide this functionality.
		1369
		1370	You I<must not> change the priority of a watcher as long as it is active or
		1371	pending.
		1372
		1373	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1374	fine, as long as you do not mind that the priority value you query might
		1375	or might not have been clamped to the valid range.
		1376
		1377	The default priority used by watchers when no priority has been set is
		1378	always C<0>, which is supposed to not be too high and not be too low :).
		1379
		1380	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1381	priorities.
		1382
		1383	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1384
		1385	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1386	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1387	can deal with that fact, as both are simply passed through to the
		1388	callback.
		1389
		1390	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1391
		1392	If the watcher is pending, this function clears its pending status and
		1393	returns its C<revents> bitset (as if its callback was invoked). If the
		1394	watcher isn't pending it does nothing and returns C<0>.
		1395
		1396	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1397	callback to be invoked, which can be accomplished with this function.
		1398
		1399	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1400
		1401	Feeds the given event set into the event loop, as if the specified event
		1402	had happened for the specified watcher (which must be a pointer to an
		1403	initialised but not necessarily started event watcher). Obviously you must
		1404	not free the watcher as long as it has pending events.
		1405
		1406	Stopping the watcher, letting libev invoke it, or calling
		1407	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1408	not started in the first place.
		1409
		1410	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1411	functions that do not need a watcher.
		1412
		1413	=back
		1414
		1415	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
		1416	OWN COMPOSITE WATCHERS> idioms.
		1417
		1418	=head2 WATCHER STATES
		1419
		1420	There are various watcher states mentioned throughout this manual -
		1421	active, pending and so on. In this section these states and the rules to
		1422	transition between them will be described in more detail - and while these
		1423	rules might look complicated, they usually do "the right thing".
		1424
		1425	=over 4
		1426
		1427	=item initialised
		1428
		1429	Before a watcher can be registered with the event loop it has to be
		1430	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1431	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1432
		1433	In this state it is simply some block of memory that is suitable for
		1434	use in an event loop. It can be moved around, freed, reused etc. at
		1435	will - as long as you either keep the memory contents intact, or call
		1436	C<ev_TYPE_init> again.
		1437
		1438	=item started/running/active
		1439
		1440	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1441	property of the event loop, and is actively waiting for events. While in
		1442	this state it cannot be accessed (except in a few documented ways), moved,
		1443	freed or anything else - the only legal thing is to keep a pointer to it,
		1444	and call libev functions on it that are documented to work on active watchers.
		1445
		1446	=item pending
		1447
		1448	If a watcher is active and libev determines that an event it is interested
		1449	in has occurred (such as a timer expiring), it will become pending. It will
		1450	stay in this pending state until either it is stopped or its callback is
		1451	about to be invoked, so it is not normally pending inside the watcher
		1452	callback.
		1453
		1454	The watcher might or might not be active while it is pending (for example,
		1455	an expired non-repeating timer can be pending but no longer active). If it
		1456	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1457	but it is still property of the event loop at this time, so cannot be
		1458	moved, freed or reused. And if it is active the rules described in the
		1459	previous item still apply.
		1460
		1461	It is also possible to feed an event on a watcher that is not active (e.g.
		1462	via C<ev_feed_event>), in which case it becomes pending without being
		1463	active.
		1464
		1465	=item stopped
		1466
		1467	A watcher can be stopped implicitly by libev (in which case it might still
		1468	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1469	latter will clear any pending state the watcher might be in, regardless
		1470	of whether it was active or not, so stopping a watcher explicitly before
		1471	freeing it is often a good idea.
		1472
		1473	While stopped (and not pending) the watcher is essentially in the
		1474	initialised state, that is, it can be reused, moved, modified in any way
		1475	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1476	it again).
		1477
		1478	=back
		1479
		1480	=head2 WATCHER PRIORITY MODELS
		1481
		1482	Many event loops support I<watcher priorities>, which are usually small
		1483	integers that influence the ordering of event callback invocation
		1484	between watchers in some way, all else being equal.
		1485
		1486	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1487	description for the more technical details such as the actual priority
		1488	range.
		1489
		1490	There are two common ways how these these priorities are being interpreted
		1491	by event loops:
		1492
		1493	In the more common lock-out model, higher priorities "lock out" invocation
		1494	of lower priority watchers, which means as long as higher priority
		1495	watchers receive events, lower priority watchers are not being invoked.
		1496
		1497	The less common only-for-ordering model uses priorities solely to order
		1498	callback invocation within a single event loop iteration: Higher priority
		1499	watchers are invoked before lower priority ones, but they all get invoked
		1500	before polling for new events.
		1501
		1502	Libev uses the second (only-for-ordering) model for all its watchers
		1503	except for idle watchers (which use the lock-out model).
		1504
		1505	The rationale behind this is that implementing the lock-out model for
		1506	watchers is not well supported by most kernel interfaces, and most event
		1507	libraries will just poll for the same events again and again as long as
		1508	their callbacks have not been executed, which is very inefficient in the
		1509	common case of one high-priority watcher locking out a mass of lower
		1510	priority ones.
		1511
		1512	Static (ordering) priorities are most useful when you have two or more
		1513	watchers handling the same resource: a typical usage example is having an
		1514	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1515	timeouts. Under load, data might be received while the program handles
		1516	other jobs, but since timers normally get invoked first, the timeout
		1517	handler will be executed before checking for data. In that case, giving
		1518	the timer a lower priority than the I/O watcher ensures that I/O will be
		1519	handled first even under adverse conditions (which is usually, but not
		1520	always, what you want).
		1521
		1522	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1523	will only be executed when no same or higher priority watchers have
		1524	received events, they can be used to implement the "lock-out" model when
		1525	required.
		1526
		1527	For example, to emulate how many other event libraries handle priorities,
		1528	you can associate an C<ev_idle> watcher to each such watcher, and in
		1529	the normal watcher callback, you just start the idle watcher. The real
		1530	processing is done in the idle watcher callback. This causes libev to
		1531	continuously poll and process kernel event data for the watcher, but when
		1532	the lock-out case is known to be rare (which in turn is rare :), this is
		1533	workable.
		1534
		1535	Usually, however, the lock-out model implemented that way will perform
		1536	miserably under the type of load it was designed to handle. In that case,
		1537	it might be preferable to stop the real watcher before starting the
		1538	idle watcher, so the kernel will not have to process the event in case
		1539	the actual processing will be delayed for considerable time.
		1540
		1541	Here is an example of an I/O watcher that should run at a strictly lower
		1542	priority than the default, and which should only process data when no
		1543	other events are pending:
		1544
		1545	ev_idle idle; // actual processing watcher
		1546	ev_io io; // actual event watcher
		1547
		1548	static void
		1549	io_cb (EV_P_ ev_io *w, int revents)
382	{	1550	{
383	struct ev_io io;	1551	// stop the I/O watcher, we received the event, but
384	int otherfd;	1552	// are not yet ready to handle it.
385	void *somedata;	1553	ev_io_stop (EV_A_ w);
386	struct whatever *mostinteresting;	1554
		1555	// start the idle watcher to handle the actual event.
		1556	// it will not be executed as long as other watchers
		1557	// with the default priority are receiving events.
		1558	ev_idle_start (EV_A_ &idle);
387	}	1559	}
388		1560
389	And since your callback will be called with a pointer to the watcher, you	1561	static void
390	can cast it back to your own type:	1562	idle_cb (EV_P_ ev_idle *w, int revents)
391
392	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
393	{	1563	{
394	struct my_io *w = (struct my_io *)w_;	1564	// actual processing
395	...	1565	read (STDIN_FILENO, ...);
		1566
		1567	// have to start the I/O watcher again, as
		1568	// we have handled the event
		1569	ev_io_start (EV_P_ &io);
396	}	1570	}
397		1571
398	More interesting and less C-conformant ways of catsing your callback type	1572	// initialisation
399	have been omitted....	1573	ev_idle_init (&idle, idle_cb);
		1574	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1575	ev_io_start (EV_DEFAULT_ &io);
		1576
		1577	In the "real" world, it might also be beneficial to start a timer, so that
		1578	low-priority connections can not be locked out forever under load. This
		1579	enables your program to keep a lower latency for important connections
		1580	during short periods of high load, while not completely locking out less
		1581	important ones.
400		1582
401		1583
402	=head1 WATCHER TYPES	1584	=head1 WATCHER TYPES
403		1585
404	This section describes each watcher in detail, but will not repeat	1586	This section describes each watcher in detail, but will not repeat
405	information given in the last section.	1587	information given in the last section. Any initialisation/set macros,
		1588	functions and members specific to the watcher type are explained.
406		1589
		1590	Members are additionally marked with either I<[read-only]>, meaning that,
		1591	while the watcher is active, you can look at the member and expect some
		1592	sensible content, but you must not modify it (you can modify it while the
		1593	watcher is stopped to your hearts content), or I<[read-write]>, which
		1594	means you can expect it to have some sensible content while the watcher
		1595	is active, but you can also modify it. Modifying it may not do something
		1596	sensible or take immediate effect (or do anything at all), but libev will
		1597	not crash or malfunction in any way.
		1598
		1599
407	=head2 struct ev_io - is my file descriptor readable or writable	1600	=head2 C<ev_io> - is this file descriptor readable or writable?
408		1601
409	I/O watchers check whether a file descriptor is readable or writable	1602	I/O watchers check whether a file descriptor is readable or writable
410	in each iteration of the event loop (This behaviour is called	1603	in each iteration of the event loop, or, more precisely, when reading
411	level-triggering because you keep receiving events as long as the	1604	would not block the process and writing would at least be able to write
412	condition persists. Remember you cna stop the watcher if you don't want to	1605	some data. This behaviour is called level-triggering because you keep
413	act on the event and neither want to receive future events).	1606	receiving events as long as the condition persists. Remember you can stop
		1607	the watcher if you don't want to act on the event and neither want to
		1608	receive future events.
414		1609
415	In general you can register as many read and/or write event watchers oer	1610	In general you can register as many read and/or write event watchers per
416	fd as you want (as long as you don't confuse yourself). Setting all file	1611	fd as you want (as long as you don't confuse yourself). Setting all file
417	descriptors to non-blocking mode is also usually a good idea (but not	1612	descriptors to non-blocking mode is also usually a good idea (but not
418	required if you know what you are doing).	1613	required if you know what you are doing).
419		1614
420	You have to be careful with dup'ed file descriptors, though. Some backends	1615	Another thing you have to watch out for is that it is quite easy to
421	(the linux epoll backend is a notable example) cannot handle dup'ed file	1616	receive "spurious" readiness notifications, that is, your callback might
422	descriptors correctly if you register interest in two or more fds pointing	1617	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
423	to the same file/socket etc. description.	1618	because there is no data. It is very easy to get into this situation even
		1619	with a relatively standard program structure. Thus it is best to always
		1620	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
		1621	preferable to a program hanging until some data arrives.
424		1622
425	If you must do this, then force the use of a known-to-be-good backend	1623	If you cannot run the fd in non-blocking mode (for example you should
426	(at the time of this writing, this includes only EVMETHOD_SELECT and	1624	not play around with an Xlib connection), then you have to separately
427	EVMETHOD_POLL).	1625	re-test whether a file descriptor is really ready with a known-to-be good
		1626	interface such as poll (fortunately in the case of Xlib, it already does
		1627	this on its own, so its quite safe to use). Some people additionally
		1628	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1629	indefinitely.
		1630
		1631	But really, best use non-blocking mode.
		1632
		1633	=head3 The special problem of disappearing file descriptors
		1634
		1635	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1636	descriptor (either due to calling C<close> explicitly or any other means,
		1637	such as C<dup2>). The reason is that you register interest in some file
		1638	descriptor, but when it goes away, the operating system will silently drop
		1639	this interest. If another file descriptor with the same number then is
		1640	registered with libev, there is no efficient way to see that this is, in
		1641	fact, a different file descriptor.
		1642
		1643	To avoid having to explicitly tell libev about such cases, libev follows
		1644	the following policy: Each time C<ev_io_set> is being called, libev
		1645	will assume that this is potentially a new file descriptor, otherwise
		1646	it is assumed that the file descriptor stays the same. That means that
		1647	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1648	descriptor even if the file descriptor number itself did not change.
		1649
		1650	This is how one would do it normally anyway, the important point is that
		1651	the libev application should not optimise around libev but should leave
		1652	optimisations to libev.
		1653
		1654	=head3 The special problem of dup'ed file descriptors
		1655
		1656	Some backends (e.g. epoll), cannot register events for file descriptors,
		1657	but only events for the underlying file descriptions. That means when you
		1658	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1659	events for them, only one file descriptor might actually receive events.
		1660
		1661	There is no workaround possible except not registering events
		1662	for potentially C<dup ()>'ed file descriptors, or to resort to
		1663	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1664
		1665	=head3 The special problem of files
		1666
		1667	Many people try to use C<select> (or libev) on file descriptors
		1668	representing files, and expect it to become ready when their program
		1669	doesn't block on disk accesses (which can take a long time on their own).
		1670
		1671	However, this cannot ever work in the "expected" way - you get a readiness
		1672	notification as soon as the kernel knows whether and how much data is
		1673	there, and in the case of open files, that's always the case, so you
		1674	always get a readiness notification instantly, and your read (or possibly
		1675	write) will still block on the disk I/O.
		1676
		1677	Another way to view it is that in the case of sockets, pipes, character
		1678	devices and so on, there is another party (the sender) that delivers data
		1679	on its own, but in the case of files, there is no such thing: the disk
		1680	will not send data on its own, simply because it doesn't know what you
		1681	wish to read - you would first have to request some data.
		1682
		1683	Since files are typically not-so-well supported by advanced notification
		1684	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1685	to files, even though you should not use it. The reason for this is
		1686	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1687	usually a tty, often a pipe, but also sometimes files or special devices
		1688	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1689	F</dev/urandom>), and even though the file might better be served with
		1690	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1691	it "just works" instead of freezing.
		1692
		1693	So avoid file descriptors pointing to files when you know it (e.g. use
		1694	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1695	when you rarely read from a file instead of from a socket, and want to
		1696	reuse the same code path.
		1697
		1698	=head3 The special problem of fork
		1699
		1700	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1701	useless behaviour. Libev fully supports fork, but needs to be told about
		1702	it in the child if you want to continue to use it in the child.
		1703
		1704	To support fork in your child processes, you have to call C<ev_loop_fork
		1705	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
		1706	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1707
		1708	=head3 The special problem of SIGPIPE
		1709
		1710	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1711	when writing to a pipe whose other end has been closed, your program gets
		1712	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1713	this is sensible behaviour, for daemons, this is usually undesirable.
		1714
		1715	So when you encounter spurious, unexplained daemon exits, make sure you
		1716	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1717	somewhere, as that would have given you a big clue).
		1718
		1719	=head3 The special problem of accept()ing when you can't
		1720
		1721	Many implementations of the POSIX C<accept> function (for example,
		1722	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1723	connection from the pending queue in all error cases.
		1724
		1725	For example, larger servers often run out of file descriptors (because
		1726	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1727	rejecting the connection, leading to libev signalling readiness on
		1728	the next iteration again (the connection still exists after all), and
		1729	typically causing the program to loop at 100% CPU usage.
		1730
		1731	Unfortunately, the set of errors that cause this issue differs between
		1732	operating systems, there is usually little the app can do to remedy the
		1733	situation, and no known thread-safe method of removing the connection to
		1734	cope with overload is known (to me).
		1735
		1736	One of the easiest ways to handle this situation is to just ignore it
		1737	- when the program encounters an overload, it will just loop until the
		1738	situation is over. While this is a form of busy waiting, no OS offers an
		1739	event-based way to handle this situation, so it's the best one can do.
		1740
		1741	A better way to handle the situation is to log any errors other than
		1742	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1743	messages, and continue as usual, which at least gives the user an idea of
		1744	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1745	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1746	usage.
		1747
		1748	If your program is single-threaded, then you could also keep a dummy file
		1749	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1750	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1751	close that fd, and create a new dummy fd. This will gracefully refuse
		1752	clients under typical overload conditions.
		1753
		1754	The last way to handle it is to simply log the error and C<exit>, as
		1755	is often done with C<malloc> failures, but this results in an easy
		1756	opportunity for a DoS attack.
		1757
		1758	=head3 Watcher-Specific Functions
428		1759
429	=over 4	1760	=over 4
430		1761
431	=item ev_io_init (ev_io *, callback, int fd, int events)	1762	=item ev_io_init (ev_io *, callback, int fd, int events)
432		1763
433	=item ev_io_set (ev_io *, int fd, int events)	1764	=item ev_io_set (ev_io *, int fd, int events)
434		1765
435	Configures an ev_io watcher. The fd is the file descriptor to rceeive	1766	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
436	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1767	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
437	EV_WRITE> to receive the given events.	1768	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1769
		1770	=item int fd [read-only]
		1771
		1772	The file descriptor being watched.
		1773
		1774	=item int events [read-only]
		1775
		1776	The events being watched.
438		1777
439	=back	1778	=back
440		1779
		1780	=head3 Examples
		1781
		1782	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1783	readable, but only once. Since it is likely line-buffered, you could
		1784	attempt to read a whole line in the callback.
		1785
		1786	static void
		1787	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1788	{
		1789	ev_io_stop (loop, w);
		1790	.. read from stdin here (or from w->fd) and handle any I/O errors
		1791	}
		1792
		1793	...
		1794	struct ev_loop *loop = ev_default_init (0);
		1795	ev_io stdin_readable;
		1796	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1797	ev_io_start (loop, &stdin_readable);
		1798	ev_run (loop, 0);
		1799
		1800
441	=head2 struct ev_timer - relative and optionally recurring timeouts	1801	=head2 C<ev_timer> - relative and optionally repeating timeouts
442		1802
443	Timer watchers are simple relative timers that generate an event after a	1803	Timer watchers are simple relative timers that generate an event after a
444	given time, and optionally repeating in regular intervals after that.	1804	given time, and optionally repeating in regular intervals after that.
445		1805
446	The timers are based on real time, that is, if you register an event that	1806	The timers are based on real time, that is, if you register an event that
447	times out after an hour and youreset your system clock to last years	1807	times out after an hour and you reset your system clock to January last
448	time, it will still time out after (roughly) and hour. "Roughly" because	1808	year, it will still time out after (roughly) one hour. "Roughly" because
449	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1809	detecting time jumps is hard, and some inaccuracies are unavoidable (the
450	monotonic clock option helps a lot here).	1810	monotonic clock option helps a lot here).
451		1811
		1812	The callback is guaranteed to be invoked only I<after> its timeout has
		1813	passed (not I<at>, so on systems with very low-resolution clocks this
		1814	might introduce a small delay, see "the special problem of being too
		1815	early", below). If multiple timers become ready during the same loop
		1816	iteration then the ones with earlier time-out values are invoked before
		1817	ones of the same priority with later time-out values (but this is no
		1818	longer true when a callback calls C<ev_run> recursively).
		1819
		1820	=head3 Be smart about timeouts
		1821
		1822	Many real-world problems involve some kind of timeout, usually for error
		1823	recovery. A typical example is an HTTP request - if the other side hangs,
		1824	you want to raise some error after a while.
		1825
		1826	What follows are some ways to handle this problem, from obvious and
		1827	inefficient to smart and efficient.
		1828
		1829	In the following, a 60 second activity timeout is assumed - a timeout that
		1830	gets reset to 60 seconds each time there is activity (e.g. each time some
		1831	data or other life sign was received).
		1832
452	=over 4	1833	=over 4
453		1834
		1835	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1836
		1837	This is the most obvious, but not the most simple way: In the beginning,
		1838	start the watcher:
		1839
		1840	ev_timer_init (timer, callback, 60., 0.);
		1841	ev_timer_start (loop, timer);
		1842
		1843	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1844	and start it again:
		1845
		1846	ev_timer_stop (loop, timer);
		1847	ev_timer_set (timer, 60., 0.);
		1848	ev_timer_start (loop, timer);
		1849
		1850	This is relatively simple to implement, but means that each time there is
		1851	some activity, libev will first have to remove the timer from its internal
		1852	data structure and then add it again. Libev tries to be fast, but it's
		1853	still not a constant-time operation.
		1854
		1855	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1856
		1857	This is the easiest way, and involves using C<ev_timer_again> instead of
		1858	C<ev_timer_start>.
		1859
		1860	To implement this, configure an C<ev_timer> with a C<repeat> value
		1861	of C<60> and then call C<ev_timer_again> at start and each time you
		1862	successfully read or write some data. If you go into an idle state where
		1863	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1864	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1865
		1866	That means you can ignore both the C<ev_timer_start> function and the
		1867	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1868	member and C<ev_timer_again>.
		1869
		1870	At start:
		1871
		1872	ev_init (timer, callback);
		1873	timer->repeat = 60.;
		1874	ev_timer_again (loop, timer);
		1875
		1876	Each time there is some activity:
		1877
		1878	ev_timer_again (loop, timer);
		1879
		1880	It is even possible to change the time-out on the fly, regardless of
		1881	whether the watcher is active or not:
		1882
		1883	timer->repeat = 30.;
		1884	ev_timer_again (loop, timer);
		1885
		1886	This is slightly more efficient then stopping/starting the timer each time
		1887	you want to modify its timeout value, as libev does not have to completely
		1888	remove and re-insert the timer from/into its internal data structure.
		1889
		1890	It is, however, even simpler than the "obvious" way to do it.
		1891
		1892	=item 3. Let the timer time out, but then re-arm it as required.
		1893
		1894	This method is more tricky, but usually most efficient: Most timeouts are
		1895	relatively long compared to the intervals between other activity - in
		1896	our example, within 60 seconds, there are usually many I/O events with
		1897	associated activity resets.
		1898
		1899	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1900	but remember the time of last activity, and check for a real timeout only
		1901	within the callback:
		1902
		1903	ev_tstamp timeout = 60.;
		1904	ev_tstamp last_activity; // time of last activity
		1905	ev_timer timer;
		1906
		1907	static void
		1908	callback (EV_P_ ev_timer *w, int revents)
		1909	{
		1910	// calculate when the timeout would happen
		1911	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
		1912
		1913	// if negative, it means we the timeout already occurred
		1914	if (after < 0.)
		1915	{
		1916	// timeout occurred, take action
		1917	}
		1918	else
		1919	{
		1920	// callback was invoked, but there was some recent
		1921	// activity. simply restart the timer to time out
		1922	// after "after" seconds, which is the earliest time
		1923	// the timeout can occur.
		1924	ev_timer_set (w, after, 0.);
		1925	ev_timer_start (EV_A_ w);
		1926	}
		1927	}
		1928
		1929	To summarise the callback: first calculate in how many seconds the
		1930	timeout will occur (by calculating the absolute time when it would occur,
		1931	C<last_activity + timeout>, and subtracting the current time, C<ev_now
		1932	(EV_A)> from that).
		1933
		1934	If this value is negative, then we are already past the timeout, i.e. we
		1935	timed out, and need to do whatever is needed in this case.
		1936
		1937	Otherwise, we now the earliest time at which the timeout would trigger,
		1938	and simply start the timer with this timeout value.
		1939
		1940	In other words, each time the callback is invoked it will check whether
		1941	the timeout occurred. If not, it will simply reschedule itself to check
		1942	again at the earliest time it could time out. Rinse. Repeat.
		1943
		1944	This scheme causes more callback invocations (about one every 60 seconds
		1945	minus half the average time between activity), but virtually no calls to
		1946	libev to change the timeout.
		1947
		1948	To start the machinery, simply initialise the watcher and set
		1949	C<last_activity> to the current time (meaning there was some activity just
		1950	now), then call the callback, which will "do the right thing" and start
		1951	the timer:
		1952
		1953	last_activity = ev_now (EV_A);
		1954	ev_init (&timer, callback);
		1955	callback (EV_A_ &timer, 0);
		1956
		1957	When there is some activity, simply store the current time in
		1958	C<last_activity>, no libev calls at all:
		1959
		1960	if (activity detected)
		1961	last_activity = ev_now (EV_A);
		1962
		1963	When your timeout value changes, then the timeout can be changed by simply
		1964	providing a new value, stopping the timer and calling the callback, which
		1965	will again do the right thing (for example, time out immediately :).
		1966
		1967	timeout = new_value;
		1968	ev_timer_stop (EV_A_ &timer);
		1969	callback (EV_A_ &timer, 0);
		1970
		1971	This technique is slightly more complex, but in most cases where the
		1972	time-out is unlikely to be triggered, much more efficient.
		1973
		1974	=item 4. Wee, just use a double-linked list for your timeouts.
		1975
		1976	If there is not one request, but many thousands (millions...), all
		1977	employing some kind of timeout with the same timeout value, then one can
		1978	do even better:
		1979
		1980	When starting the timeout, calculate the timeout value and put the timeout
		1981	at the I<end> of the list.
		1982
		1983	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1984	the list is expected to fire (for example, using the technique #3).
		1985
		1986	When there is some activity, remove the timer from the list, recalculate
		1987	the timeout, append it to the end of the list again, and make sure to
		1988	update the C<ev_timer> if it was taken from the beginning of the list.
		1989
		1990	This way, one can manage an unlimited number of timeouts in O(1) time for
		1991	starting, stopping and updating the timers, at the expense of a major
		1992	complication, and having to use a constant timeout. The constant timeout
		1993	ensures that the list stays sorted.
		1994
		1995	=back
		1996
		1997	So which method the best?
		1998
		1999	Method #2 is a simple no-brain-required solution that is adequate in most
		2000	situations. Method #3 requires a bit more thinking, but handles many cases
		2001	better, and isn't very complicated either. In most case, choosing either
		2002	one is fine, with #3 being better in typical situations.
		2003
		2004	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		2005	rather complicated, but extremely efficient, something that really pays
		2006	off after the first million or so of active timers, i.e. it's usually
		2007	overkill :)
		2008
		2009	=head3 The special problem of being too early
		2010
		2011	If you ask a timer to call your callback after three seconds, then
		2012	you expect it to be invoked after three seconds - but of course, this
		2013	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		2014	guaranteed to any precision by libev - imagine somebody suspending the
		2015	process with a STOP signal for a few hours for example.
		2016
		2017	So, libev tries to invoke your callback as soon as possible I<after> the
		2018	delay has occurred, but cannot guarantee this.
		2019
		2020	A less obvious failure mode is calling your callback too early: many event
		2021	loops compare timestamps with a "elapsed delay >= requested delay", but
		2022	this can cause your callback to be invoked much earlier than you would
		2023	expect.
		2024
		2025	To see why, imagine a system with a clock that only offers full second
		2026	resolution (think windows if you can't come up with a broken enough OS
		2027	yourself). If you schedule a one-second timer at the time 500.9, then the
		2028	event loop will schedule your timeout to elapse at a system time of 500
		2029	(500.9 truncated to the resolution) + 1, or 501.
		2030
		2031	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2032	501" and invoke the callback 0.1s after it was started, even though a
		2033	one-second delay was requested - this is being "too early", despite best
		2034	intentions.
		2035
		2036	This is the reason why libev will never invoke the callback if the elapsed
		2037	delay equals the requested delay, but only when the elapsed delay is
		2038	larger than the requested delay. In the example above, libev would only invoke
		2039	the callback at system time 502, or 1.1s after the timer was started.
		2040
		2041	So, while libev cannot guarantee that your callback will be invoked
		2042	exactly when requested, it I<can> and I<does> guarantee that the requested
		2043	delay has actually elapsed, or in other words, it always errs on the "too
		2044	late" side of things.
		2045
		2046	=head3 The special problem of time updates
		2047
		2048	Establishing the current time is a costly operation (it usually takes
		2049	at least one system call): EV therefore updates its idea of the current
		2050	time only before and after C<ev_run> collects new events, which causes a
		2051	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		2052	lots of events in one iteration.
		2053
		2054	The relative timeouts are calculated relative to the C<ev_now ()>
		2055	time. This is usually the right thing as this timestamp refers to the time
		2056	of the event triggering whatever timeout you are modifying/starting. If
		2057	you suspect event processing to be delayed and you I<need> to base the
		2058	timeout on the current time, use something like the following to adjust
		2059	for it:
		2060
		2061	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
		2062
		2063	If the event loop is suspended for a long time, you can also force an
		2064	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		2065	()>, although that will push the event time of all outstanding events
		2066	further into the future.
		2067
		2068	=head3 The special problem of unsynchronised clocks
		2069
		2070	Modern systems have a variety of clocks - libev itself uses the normal
		2071	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2072	jumps).
		2073
		2074	Neither of these clocks is synchronised with each other or any other clock
		2075	on the system, so C<ev_time ()> might return a considerably different time
		2076	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2077	a call to C<gettimeofday> might return a second count that is one higher
		2078	than a directly following call to C<time>.
		2079
		2080	The moral of this is to only compare libev-related timestamps with
		2081	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2082	a second or so.
		2083
		2084	One more problem arises due to this lack of synchronisation: if libev uses
		2085	the system monotonic clock and you compare timestamps from C<ev_time>
		2086	or C<ev_now> from when you started your timer and when your callback is
		2087	invoked, you will find that sometimes the callback is a bit "early".
		2088
		2089	This is because C<ev_timer>s work in real time, not wall clock time, so
		2090	libev makes sure your callback is not invoked before the delay happened,
		2091	I<measured according to the real time>, not the system clock.
		2092
		2093	If your timeouts are based on a physical timescale (e.g. "time out this
		2094	connection after 100 seconds") then this shouldn't bother you as it is
		2095	exactly the right behaviour.
		2096
		2097	If you want to compare wall clock/system timestamps to your timers, then
		2098	you need to use C<ev_periodic>s, as these are based on the wall clock
		2099	time, where your comparisons will always generate correct results.
		2100
		2101	=head3 The special problems of suspended animation
		2102
		2103	When you leave the server world it is quite customary to hit machines that
		2104	can suspend/hibernate - what happens to the clocks during such a suspend?
		2105
		2106	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		2107	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2108	to run until the system is suspended, but they will not advance while the
		2109	system is suspended. That means, on resume, it will be as if the program
		2110	was frozen for a few seconds, but the suspend time will not be counted
		2111	towards C<ev_timer> when a monotonic clock source is used. The real time
		2112	clock advanced as expected, but if it is used as sole clocksource, then a
		2113	long suspend would be detected as a time jump by libev, and timers would
		2114	be adjusted accordingly.
		2115
		2116	I would not be surprised to see different behaviour in different between
		2117	operating systems, OS versions or even different hardware.
		2118
		2119	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2120	time jump in the monotonic clocks and the realtime clock. If the program
		2121	is suspended for a very long time, and monotonic clock sources are in use,
		2122	then you can expect C<ev_timer>s to expire as the full suspension time
		2123	will be counted towards the timers. When no monotonic clock source is in
		2124	use, then libev will again assume a timejump and adjust accordingly.
		2125
		2126	It might be beneficial for this latter case to call C<ev_suspend>
		2127	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2128	deterministic behaviour in this case (you can do nothing against
		2129	C<SIGSTOP>).
		2130
		2131	=head3 Watcher-Specific Functions and Data Members
		2132
		2133	=over 4
		2134
454	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2135	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
455		2136
456	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2137	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
457		2138
458	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2139	Configure the timer to trigger after C<after> seconds (fractional and
459	C<0.>, then it will automatically be stopped. If it is positive, then the	2140	negative values are supported). If C<repeat> is C<0.>, then it will
		2141	automatically be stopped once the timeout is reached. If it is positive,
460	timer will automatically be configured to trigger again C<repeat> seconds	2142	then the timer will automatically be configured to trigger again C<repeat>
461	later, again, and again, until stopped manually.	2143	seconds later, again, and again, until stopped manually.
462		2144
463	The timer itself will do a best-effort at avoiding drift, that is, if you	2145	The timer itself will do a best-effort at avoiding drift, that is, if
464	configure a timer to trigger every 10 seconds, then it will trigger at	2146	you configure a timer to trigger every 10 seconds, then it will normally
465	exactly 10 second intervals. If, however, your program cannot keep up with	2147	trigger at exactly 10 second intervals. If, however, your program cannot
466	the timer (ecause it takes longer than those 10 seconds to do stuff) the	2148	keep up with the timer (because it takes longer than those 10 seconds to
467	timer will not fire more than once per event loop iteration.	2149	do stuff) the timer will not fire more than once per event loop iteration.
468		2150
469	=item ev_timer_again (loop)	2151	=item ev_timer_again (loop, ev_timer *)
470		2152
471	This will act as if the timer timed out and restart it again if it is	2153	This will act as if the timer timed out, and restarts it again if it is
472	repeating. The exact semantics are:	2154	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2155	timeout to the C<repeat> value and calling C<ev_timer_start>.
473		2156
		2157	The exact semantics are as in the following rules, all of which will be
		2158	applied to the watcher:
		2159
		2160	=over 4
		2161
		2162	=item If the timer is pending, the pending status is always cleared.
		2163
474	If the timer is started but nonrepeating, stop it.	2164	=item If the timer is started but non-repeating, stop it (as if it timed
		2165	out, without invoking it).
475		2166
476	If the timer is repeating, either start it if necessary (with the repeat	2167	=item If the timer is repeating, make the C<repeat> value the new timeout
477	value), or reset the running timer to the repeat value.	2168	and start the timer, if necessary.
478
479	This sounds a bit complicated, but here is a useful and typical
480	example: Imagine you have a tcp connection and you want a so-called idle
481	timeout, that is, you want to be called when there have been, say, 60
482	seconds of inactivity on the socket. The easiest way to do this is to
483	configure an ev_timer with after=repeat=60 and calling ev_timer_again each
484	time you successfully read or write some data. If you go into an idle
485	state where you do not expect data to travel on the socket, you can stop
486	the timer, and again will automatically restart it if need be.
487		2169
488	=back	2170	=back
489		2171
		2172	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
		2173	usage example.
		2174
		2175	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
		2176
		2177	Returns the remaining time until a timer fires. If the timer is active,
		2178	then this time is relative to the current event loop time, otherwise it's
		2179	the timeout value currently configured.
		2180
		2181	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2182	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2183	will return C<4>. When the timer expires and is restarted, it will return
		2184	roughly C<7> (likely slightly less as callback invocation takes some time,
		2185	too), and so on.
		2186
		2187	=item ev_tstamp repeat [read-write]
		2188
		2189	The current C<repeat> value. Will be used each time the watcher times out
		2190	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2191	which is also when any modifications are taken into account.
		2192
		2193	=back
		2194
		2195	=head3 Examples
		2196
		2197	Example: Create a timer that fires after 60 seconds.
		2198
		2199	static void
		2200	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2201	{
		2202	.. one minute over, w is actually stopped right here
		2203	}
		2204
		2205	ev_timer mytimer;
		2206	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2207	ev_timer_start (loop, &mytimer);
		2208
		2209	Example: Create a timeout timer that times out after 10 seconds of
		2210	inactivity.
		2211
		2212	static void
		2213	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2214	{
		2215	.. ten seconds without any activity
		2216	}
		2217
		2218	ev_timer mytimer;
		2219	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2220	ev_timer_again (&mytimer); /* start timer */
		2221	ev_run (loop, 0);
		2222
		2223	// and in some piece of code that gets executed on any "activity":
		2224	// reset the timeout to start ticking again at 10 seconds
		2225	ev_timer_again (&mytimer);
		2226
		2227
490	=head2 ev_periodic - to cron or not to cron it	2228	=head2 C<ev_periodic> - to cron or not to cron?
491		2229
492	Periodic watchers are also timers of a kind, but they are very versatile	2230	Periodic watchers are also timers of a kind, but they are very versatile
493	(and unfortunately a bit complex).	2231	(and unfortunately a bit complex).
494		2232
495	Unlike ev_timer's, they are not based on real time (or relative time)	2233	Unlike C<ev_timer>, periodic watchers are not based on real time (or
496	but on wallclock time (absolute time). You can tell a periodic watcher	2234	relative time, the physical time that passes) but on wall clock time
497	to trigger "at" some specific point in time. For example, if you tell a	2235	(absolute time, the thing you can read on your calendar or clock). The
498	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2236	difference is that wall clock time can run faster or slower than real
499	+ 10.>) and then reset your system clock to the last year, then it will	2237	time, and time jumps are not uncommon (e.g. when you adjust your
500	take a year to trigger the event (unlike an ev_timer, which would trigger	2238	wrist-watch).
501	roughly 10 seconds later and of course not if you reset your system time
502	again).
503		2239
504	They can also be used to implement vastly more complex timers, such as	2240	You can tell a periodic watcher to trigger after some specific point
		2241	in time: for example, if you tell a periodic watcher to trigger "in 10
		2242	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2243	not a delay) and then reset your system clock to January of the previous
		2244	year, then it will take a year or more to trigger the event (unlike an
		2245	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2246	it, as it uses a relative timeout).
		2247
		2248	C<ev_periodic> watchers can also be used to implement vastly more complex
505	triggering an event on eahc midnight, local time.	2249	timers, such as triggering an event on each "midnight, local time", or
		2250	other complicated rules. This cannot easily be done with C<ev_timer>
		2251	watchers, as those cannot react to time jumps.
		2252
		2253	As with timers, the callback is guaranteed to be invoked only when the
		2254	point in time where it is supposed to trigger has passed. If multiple
		2255	timers become ready during the same loop iteration then the ones with
		2256	earlier time-out values are invoked before ones with later time-out values
		2257	(but this is no longer true when a callback calls C<ev_run> recursively).
		2258
		2259	=head3 Watcher-Specific Functions and Data Members
506		2260
507	=over 4	2261	=over 4
508		2262
509	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2263	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
510		2264
511	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2265	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
512		2266
513	Lots of arguments, lets sort it out... There are basically three modes of	2267	Lots of arguments, let's sort it out... There are basically three modes of
514	operation, and we will explain them from simplest to complex:	2268	operation, and we will explain them from simplest to most complex:
515
516		2269
517	=over 4	2270	=over 4
518		2271
519	=item * absolute timer (interval = reschedule_cb = 0)	2272	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
520		2273
521	In this configuration the watcher triggers an event at the wallclock time	2274	In this configuration the watcher triggers an event after the wall clock
522	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2275	time C<offset> has passed. It will not repeat and will not adjust when a
523	that is, if it is to be run at January 1st 2011 then it will run when the	2276	time jump occurs, that is, if it is to be run at January 1st 2011 then it
524	system time reaches or surpasses this time.	2277	will be stopped and invoked when the system clock reaches or surpasses
		2278	this point in time.
525		2279
526	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2280	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
527		2281
528	In this mode the watcher will always be scheduled to time out at the next	2282	In this mode the watcher will always be scheduled to time out at the next
529	C<at + N * interval> time (for some integer N) and then repeat, regardless	2283	C<offset + N * interval> time (for some integer N, which can also be
530	of any time jumps.	2284	negative) and then repeat, regardless of any time jumps. The C<offset>
		2285	argument is merely an offset into the C<interval> periods.
531		2286
532	This can be used to create timers that do not drift with respect to system	2287	This can be used to create timers that do not drift with respect to the
533	time:	2288	system clock, for example, here is an C<ev_periodic> that triggers each
		2289	hour, on the hour (with respect to UTC):
534		2290
535	ev_periodic_set (&periodic, 0., 3600., 0);	2291	ev_periodic_set (&periodic, 0., 3600., 0);
536		2292
537	This doesn't mean there will always be 3600 seconds in between triggers,	2293	This doesn't mean there will always be 3600 seconds in between triggers,
538	but only that the the callback will be called when the system time shows a	2294	but only that the callback will be called when the system time shows a
539	full hour (UTC), or more correct, when the system time is evenly divisible	2295	full hour (UTC), or more correctly, when the system time is evenly divisible
540	by 3600.	2296	by 3600.
541		2297
542	Another way to think about it (for the mathematically inclined) is that	2298	Another way to think about it (for the mathematically inclined) is that
543	ev_periodic will try to run the callback in this mode at the next possible	2299	C<ev_periodic> will try to run the callback in this mode at the next possible
544	time where C<time = at (mod interval)>, regardless of any time jumps.	2300	time where C<time = offset (mod interval)>, regardless of any time jumps.
545		2301
		2302	The C<interval> I<MUST> be positive, and for numerical stability, the
		2303	interval value should be higher than C<1/8192> (which is around 100
		2304	microseconds) and C<offset> should be higher than C<0> and should have
		2305	at most a similar magnitude as the current time (say, within a factor of
		2306	ten). Typical values for offset are, in fact, C<0> or something between
		2307	C<0> and C<interval>, which is also the recommended range.
		2308
		2309	Note also that there is an upper limit to how often a timer can fire (CPU
		2310	speed for example), so if C<interval> is very small then timing stability
		2311	will of course deteriorate. Libev itself tries to be exact to be about one
		2312	millisecond (if the OS supports it and the machine is fast enough).
		2313
546	=item * manual reschedule mode (reschedule_cb = callback)	2314	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
547		2315
548	In this mode the values for C<interval> and C<at> are both being	2316	In this mode the values for C<interval> and C<offset> are both being
549	ignored. Instead, each time the periodic watcher gets scheduled, the	2317	ignored. Instead, each time the periodic watcher gets scheduled, the
550	reschedule callback will be called with the watcher as first, and the	2318	reschedule callback will be called with the watcher as first, and the
551	current time as second argument.	2319	current time as second argument.
552		2320
553	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	2321	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
554	periodic watcher, ever, or make any event loop modificstions>. If you need	2322	or make ANY other event loop modifications whatsoever, unless explicitly
555	to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards.	2323	allowed by documentation here>.
556		2324
		2325	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2326	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2327	only event loop modification you are allowed to do).
		2328
557	Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2329	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
558	ev_tstamp now)>, e.g.:	2330	*w, ev_tstamp now)>, e.g.:
559		2331
		2332	static ev_tstamp
560	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2333	my_rescheduler (ev_periodic *w, ev_tstamp now)
561	{	2334	{
562	return now + 60.;	2335	return now + 60.;
563	}	2336	}
564		2337
565	It must return the next time to trigger, based on the passed time value	2338	It must return the next time to trigger, based on the passed time value
566	(that is, the lowest time value larger than to the second argument). It	2339	(that is, the lowest time value larger than to the second argument). It
567	will usually be called just before the callback will be triggered, but	2340	will usually be called just before the callback will be triggered, but
568	might be called at other times, too.	2341	might be called at other times, too.
569		2342
		2343	NOTE: I<< This callback must always return a time that is higher than or
		2344	equal to the passed C<now> value >>.
		2345
570	This can be used to create very complex timers, such as a timer that	2346	This can be used to create very complex timers, such as a timer that
571	triggers on each midnight, local time. To do this, you would calculate the	2347	triggers on "next midnight, local time". To do this, you would calculate
572	next midnight after C<now> and return the timestamp value for this. How you do this	2348	the next midnight after C<now> and return the timestamp value for
573	is, again, up to you (but it is not trivial).	2349	this. Here is a (completely untested, no error checking) example on how to
		2350	do this:
		2351
		2352	#include <time.h>
		2353
		2354	static ev_tstamp
		2355	my_rescheduler (ev_periodic *w, ev_tstamp now)
		2356	{
		2357	time_t tnow = (time_t)now;
		2358	struct tm tm;
		2359	localtime_r (&tnow, &tm);
		2360
		2361	tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
		2362	++tm.tm_mday; // midnight next day
		2363
		2364	return mktime (&tm);
		2365	}
		2366
		2367	Note: this code might run into trouble on days that have more then two
		2368	midnights (beginning and end).
574		2369
575	=back	2370	=back
576		2371
577	=item ev_periodic_again (loop, ev_periodic *)	2372	=item ev_periodic_again (loop, ev_periodic *)
578		2373
579	Simply stops and restarts the periodic watcher again. This is only useful	2374	Simply stops and restarts the periodic watcher again. This is only useful
580	when you changed some parameters or the reschedule callback would return	2375	when you changed some parameters or the reschedule callback would return
581	a different time than the last time it was called (e.g. in a crond like	2376	a different time than the last time it was called (e.g. in a crond like
582	program when the crontabs have changed).	2377	program when the crontabs have changed).
583		2378
		2379	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2380
		2381	When active, returns the absolute time that the watcher is supposed
		2382	to trigger next. This is not the same as the C<offset> argument to
		2383	C<ev_periodic_set>, but indeed works even in interval and manual
		2384	rescheduling modes.
		2385
		2386	=item ev_tstamp offset [read-write]
		2387
		2388	When repeating, this contains the offset value, otherwise this is the
		2389	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2390	although libev might modify this value for better numerical stability).
		2391
		2392	Can be modified any time, but changes only take effect when the periodic
		2393	timer fires or C<ev_periodic_again> is being called.
		2394
		2395	=item ev_tstamp interval [read-write]
		2396
		2397	The current interval value. Can be modified any time, but changes only
		2398	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2399	called.
		2400
		2401	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2402
		2403	The current reschedule callback, or C<0>, if this functionality is
		2404	switched off. Can be changed any time, but changes only take effect when
		2405	the periodic timer fires or C<ev_periodic_again> is being called.
		2406
584	=back	2407	=back
585		2408
		2409	=head3 Examples
		2410
		2411	Example: Call a callback every hour, or, more precisely, whenever the
		2412	system time is divisible by 3600. The callback invocation times have
		2413	potentially a lot of jitter, but good long-term stability.
		2414
		2415	static void
		2416	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2417	{
		2418	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2419	}
		2420
		2421	ev_periodic hourly_tick;
		2422	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2423	ev_periodic_start (loop, &hourly_tick);
		2424
		2425	Example: The same as above, but use a reschedule callback to do it:
		2426
		2427	#include <math.h>
		2428
		2429	static ev_tstamp
		2430	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2431	{
		2432	return now + (3600. - fmod (now, 3600.));
		2433	}
		2434
		2435	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2436
		2437	Example: Call a callback every hour, starting now:
		2438
		2439	ev_periodic hourly_tick;
		2440	ev_periodic_init (&hourly_tick, clock_cb,
		2441	fmod (ev_now (loop), 3600.), 3600., 0);
		2442	ev_periodic_start (loop, &hourly_tick);
		2443
		2444
586	=head2 ev_signal - signal me when a signal gets signalled	2445	=head2 C<ev_signal> - signal me when a signal gets signalled!
587		2446
588	Signal watchers will trigger an event when the process receives a specific	2447	Signal watchers will trigger an event when the process receives a specific
589	signal one or more times. Even though signals are very asynchronous, libev	2448	signal one or more times. Even though signals are very asynchronous, libev
590	will try its best to deliver signals synchronously, i.e. as part of the	2449	will try its best to deliver signals synchronously, i.e. as part of the
591	normal event processing, like any other event.	2450	normal event processing, like any other event.
592		2451
		2452	If you want signals to be delivered truly asynchronously, just use
		2453	C<sigaction> as you would do without libev and forget about sharing
		2454	the signal. You can even use C<ev_async> from a signal handler to
		2455	synchronously wake up an event loop.
		2456
593	You cna configure as many watchers as you like per signal. Only when the	2457	You can configure as many watchers as you like for the same signal, but
594	first watcher gets started will libev actually register a signal watcher	2458	only within the same loop, i.e. you can watch for C<SIGINT> in your
595	with the kernel (thus it coexists with your own signal handlers as long	2459	default loop and for C<SIGIO> in another loop, but you cannot watch for
596	as you don't register any with libev). Similarly, when the last signal	2460	C<SIGINT> in both the default loop and another loop at the same time. At
597	watcher for a signal is stopped libev will reset the signal handler to	2461	the moment, C<SIGCHLD> is permanently tied to the default loop.
598	SIG_DFL (regardless of what it was set to before).	2462
		2463	Only after the first watcher for a signal is started will libev actually
		2464	register something with the kernel. It thus coexists with your own signal
		2465	handlers as long as you don't register any with libev for the same signal.
		2466
		2467	If possible and supported, libev will install its handlers with
		2468	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2469	not be unduly interrupted. If you have a problem with system calls getting
		2470	interrupted by signals you can block all signals in an C<ev_check> watcher
		2471	and unblock them in an C<ev_prepare> watcher.
		2472
		2473	=head3 The special problem of inheritance over fork/execve/pthread_create
		2474
		2475	Both the signal mask (C<sigprocmask>) and the signal disposition
		2476	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2477	stopping it again), that is, libev might or might not block the signal,
		2478	and might or might not set or restore the installed signal handler (but
		2479	see C<EVFLAG_NOSIGMASK>).
		2480
		2481	While this does not matter for the signal disposition (libev never
		2482	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2483	C<execve>), this matters for the signal mask: many programs do not expect
		2484	certain signals to be blocked.
		2485
		2486	This means that before calling C<exec> (from the child) you should reset
		2487	the signal mask to whatever "default" you expect (all clear is a good
		2488	choice usually).
		2489
		2490	The simplest way to ensure that the signal mask is reset in the child is
		2491	to install a fork handler with C<pthread_atfork> that resets it. That will
		2492	catch fork calls done by libraries (such as the libc) as well.
		2493
		2494	In current versions of libev, the signal will not be blocked indefinitely
		2495	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2496	the window of opportunity for problems, it will not go away, as libev
		2497	I<has> to modify the signal mask, at least temporarily.
		2498
		2499	So I can't stress this enough: I<If you do not reset your signal mask when
		2500	you expect it to be empty, you have a race condition in your code>. This
		2501	is not a libev-specific thing, this is true for most event libraries.
		2502
		2503	=head3 The special problem of threads signal handling
		2504
		2505	POSIX threads has problematic signal handling semantics, specifically,
		2506	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2507	threads in a process block signals, which is hard to achieve.
		2508
		2509	When you want to use sigwait (or mix libev signal handling with your own
		2510	for the same signals), you can tackle this problem by globally blocking
		2511	all signals before creating any threads (or creating them with a fully set
		2512	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2513	loops. Then designate one thread as "signal receiver thread" which handles
		2514	these signals. You can pass on any signals that libev might be interested
		2515	in by calling C<ev_feed_signal>.
		2516
		2517	=head3 Watcher-Specific Functions and Data Members
599		2518
600	=over 4	2519	=over 4
601		2520
602	=item ev_signal_init (ev_signal *, callback, int signum)	2521	=item ev_signal_init (ev_signal *, callback, int signum)
603		2522
604	=item ev_signal_set (ev_signal *, int signum)	2523	=item ev_signal_set (ev_signal *, int signum)
605		2524
606	Configures the watcher to trigger on the given signal number (usually one	2525	Configures the watcher to trigger on the given signal number (usually one
607	of the C<SIGxxx> constants).	2526	of the C<SIGxxx> constants).
608		2527
		2528	=item int signum [read-only]
		2529
		2530	The signal the watcher watches out for.
		2531
609	=back	2532	=back
610		2533
		2534	=head3 Examples
		2535
		2536	Example: Try to exit cleanly on SIGINT.
		2537
		2538	static void
		2539	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2540	{
		2541	ev_break (loop, EVBREAK_ALL);
		2542	}
		2543
		2544	ev_signal signal_watcher;
		2545	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2546	ev_signal_start (loop, &signal_watcher);
		2547
		2548
611	=head2 ev_child - wait for pid status changes	2549	=head2 C<ev_child> - watch out for process status changes
612		2550
613	Child watchers trigger when your process receives a SIGCHLD in response to	2551	Child watchers trigger when your process receives a SIGCHLD in response to
614	some child status changes (most typically when a child of yours dies).	2552	some child status changes (most typically when a child of yours dies or
		2553	exits). It is permissible to install a child watcher I<after> the child
		2554	has been forked (which implies it might have already exited), as long
		2555	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2556	forking and then immediately registering a watcher for the child is fine,
		2557	but forking and registering a watcher a few event loop iterations later or
		2558	in the next callback invocation is not.
		2559
		2560	Only the default event loop is capable of handling signals, and therefore
		2561	you can only register child watchers in the default event loop.
		2562
		2563	Due to some design glitches inside libev, child watchers will always be
		2564	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2565	libev)
		2566
		2567	=head3 Process Interaction
		2568
		2569	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2570	initialised. This is necessary to guarantee proper behaviour even if the
		2571	first child watcher is started after the child exits. The occurrence
		2572	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2573	synchronously as part of the event loop processing. Libev always reaps all
		2574	children, even ones not watched.
		2575
		2576	=head3 Overriding the Built-In Processing
		2577
		2578	Libev offers no special support for overriding the built-in child
		2579	processing, but if your application collides with libev's default child
		2580	handler, you can override it easily by installing your own handler for
		2581	C<SIGCHLD> after initialising the default loop, and making sure the
		2582	default loop never gets destroyed. You are encouraged, however, to use an
		2583	event-based approach to child reaping and thus use libev's support for
		2584	that, so other libev users can use C<ev_child> watchers freely.
		2585
		2586	=head3 Stopping the Child Watcher
		2587
		2588	Currently, the child watcher never gets stopped, even when the
		2589	child terminates, so normally one needs to stop the watcher in the
		2590	callback. Future versions of libev might stop the watcher automatically
		2591	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2592	problem).
		2593
		2594	=head3 Watcher-Specific Functions and Data Members
615		2595
616	=over 4	2596	=over 4
617		2597
618	=item ev_child_init (ev_child *, callback, int pid)	2598	=item ev_child_init (ev_child *, callback, int pid, int trace)
619		2599
620	=item ev_child_set (ev_child *, int pid)	2600	=item ev_child_set (ev_child *, int pid, int trace)
621		2601
622	Configures the watcher to wait for status changes of process C<pid> (or	2602	Configures the watcher to wait for status changes of process C<pid> (or
623	I<any> process if C<pid> is specified as C<0>). The callback can look	2603	I<any> process if C<pid> is specified as C<0>). The callback can look
624	at the C<rstatus> member of the C<ev_child> watcher structure to see	2604	at the C<rstatus> member of the C<ev_child> watcher structure to see
625	the status word (use the macros from C<sys/wait.h>). The C<rpid> member	2605	the status word (use the macros from C<sys/wait.h> and see your systems
626	contains the pid of the process causing the status change.	2606	C<waitpid> documentation). The C<rpid> member contains the pid of the
		2607	process causing the status change. C<trace> must be either C<0> (only
		2608	activate the watcher when the process terminates) or C<1> (additionally
		2609	activate the watcher when the process is stopped or continued).
		2610
		2611	=item int pid [read-only]
		2612
		2613	The process id this watcher watches out for, or C<0>, meaning any process id.
		2614
		2615	=item int rpid [read-write]
		2616
		2617	The process id that detected a status change.
		2618
		2619	=item int rstatus [read-write]
		2620
		2621	The process exit/trace status caused by C<rpid> (see your systems
		2622	C<waitpid> and C<sys/wait.h> documentation for details).
627		2623
628	=back	2624	=back
629		2625
		2626	=head3 Examples
		2627
		2628	Example: C<fork()> a new process and install a child handler to wait for
		2629	its completion.
		2630
		2631	ev_child cw;
		2632
		2633	static void
		2634	child_cb (EV_P_ ev_child *w, int revents)
		2635	{
		2636	ev_child_stop (EV_A_ w);
		2637	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2638	}
		2639
		2640	pid_t pid = fork ();
		2641
		2642	if (pid < 0)
		2643	// error
		2644	else if (pid == 0)
		2645	{
		2646	// the forked child executes here
		2647	exit (1);
		2648	}
		2649	else
		2650	{
		2651	ev_child_init (&cw, child_cb, pid, 0);
		2652	ev_child_start (EV_DEFAULT_ &cw);
		2653	}
		2654
		2655
		2656	=head2 C<ev_stat> - did the file attributes just change?
		2657
		2658	This watches a file system path for attribute changes. That is, it calls
		2659	C<stat> on that path in regular intervals (or when the OS says it changed)
		2660	and sees if it changed compared to the last time, invoking the callback
		2661	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2662	happen after the watcher has been started will be reported.
		2663
		2664	The path does not need to exist: changing from "path exists" to "path does
		2665	not exist" is a status change like any other. The condition "path does not
		2666	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2667	C<st_nlink> field being zero (which is otherwise always forced to be at
		2668	least one) and all the other fields of the stat buffer having unspecified
		2669	contents.
		2670
		2671	The path I<must not> end in a slash or contain special components such as
		2672	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2673	your working directory changes, then the behaviour is undefined.
		2674
		2675	Since there is no portable change notification interface available, the
		2676	portable implementation simply calls C<stat(2)> regularly on the path
		2677	to see if it changed somehow. You can specify a recommended polling
		2678	interval for this case. If you specify a polling interval of C<0> (highly
		2679	recommended!) then a I<suitable, unspecified default> value will be used
		2680	(which you can expect to be around five seconds, although this might
		2681	change dynamically). Libev will also impose a minimum interval which is
		2682	currently around C<0.1>, but that's usually overkill.
		2683
		2684	This watcher type is not meant for massive numbers of stat watchers,
		2685	as even with OS-supported change notifications, this can be
		2686	resource-intensive.
		2687
		2688	At the time of this writing, the only OS-specific interface implemented
		2689	is the Linux inotify interface (implementing kqueue support is left as an
		2690	exercise for the reader. Note, however, that the author sees no way of
		2691	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2692
		2693	=head3 ABI Issues (Largefile Support)
		2694
		2695	Libev by default (unless the user overrides this) uses the default
		2696	compilation environment, which means that on systems with large file
		2697	support disabled by default, you get the 32 bit version of the stat
		2698	structure. When using the library from programs that change the ABI to
		2699	use 64 bit file offsets the programs will fail. In that case you have to
		2700	compile libev with the same flags to get binary compatibility. This is
		2701	obviously the case with any flags that change the ABI, but the problem is
		2702	most noticeably displayed with ev_stat and large file support.
		2703
		2704	The solution for this is to lobby your distribution maker to make large
		2705	file interfaces available by default (as e.g. FreeBSD does) and not
		2706	optional. Libev cannot simply switch on large file support because it has
		2707	to exchange stat structures with application programs compiled using the
		2708	default compilation environment.
		2709
		2710	=head3 Inotify and Kqueue
		2711
		2712	When C<inotify (7)> support has been compiled into libev and present at
		2713	runtime, it will be used to speed up change detection where possible. The
		2714	inotify descriptor will be created lazily when the first C<ev_stat>
		2715	watcher is being started.
		2716
		2717	Inotify presence does not change the semantics of C<ev_stat> watchers
		2718	except that changes might be detected earlier, and in some cases, to avoid
		2719	making regular C<stat> calls. Even in the presence of inotify support
		2720	there are many cases where libev has to resort to regular C<stat> polling,
		2721	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2722	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2723	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2724	xfs are fully working) libev usually gets away without polling.
		2725
		2726	There is no support for kqueue, as apparently it cannot be used to
		2727	implement this functionality, due to the requirement of having a file
		2728	descriptor open on the object at all times, and detecting renames, unlinks
		2729	etc. is difficult.
		2730
		2731	=head3 C<stat ()> is a synchronous operation
		2732
		2733	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2734	the process. The exception are C<ev_stat> watchers - those call C<stat
		2735	()>, which is a synchronous operation.
		2736
		2737	For local paths, this usually doesn't matter: unless the system is very
		2738	busy or the intervals between stat's are large, a stat call will be fast,
		2739	as the path data is usually in memory already (except when starting the
		2740	watcher).
		2741
		2742	For networked file systems, calling C<stat ()> can block an indefinite
		2743	time due to network issues, and even under good conditions, a stat call
		2744	often takes multiple milliseconds.
		2745
		2746	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2747	paths, although this is fully supported by libev.
		2748
		2749	=head3 The special problem of stat time resolution
		2750
		2751	The C<stat ()> system call only supports full-second resolution portably,
		2752	and even on systems where the resolution is higher, most file systems
		2753	still only support whole seconds.
		2754
		2755	That means that, if the time is the only thing that changes, you can
		2756	easily miss updates: on the first update, C<ev_stat> detects a change and
		2757	calls your callback, which does something. When there is another update
		2758	within the same second, C<ev_stat> will be unable to detect unless the
		2759	stat data does change in other ways (e.g. file size).
		2760
		2761	The solution to this is to delay acting on a change for slightly more
		2762	than a second (or till slightly after the next full second boundary), using
		2763	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2764	ev_timer_again (loop, w)>).
		2765
		2766	The C<.02> offset is added to work around small timing inconsistencies
		2767	of some operating systems (where the second counter of the current time
		2768	might be be delayed. One such system is the Linux kernel, where a call to
		2769	C<gettimeofday> might return a timestamp with a full second later than
		2770	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2771	update file times then there will be a small window where the kernel uses
		2772	the previous second to update file times but libev might already execute
		2773	the timer callback).
		2774
		2775	=head3 Watcher-Specific Functions and Data Members
		2776
		2777	=over 4
		2778
		2779	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2780
		2781	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2782
		2783	Configures the watcher to wait for status changes of the given
		2784	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2785	be detected and should normally be specified as C<0> to let libev choose
		2786	a suitable value. The memory pointed to by C<path> must point to the same
		2787	path for as long as the watcher is active.
		2788
		2789	The callback will receive an C<EV_STAT> event when a change was detected,
		2790	relative to the attributes at the time the watcher was started (or the
		2791	last change was detected).
		2792
		2793	=item ev_stat_stat (loop, ev_stat *)
		2794
		2795	Updates the stat buffer immediately with new values. If you change the
		2796	watched path in your callback, you could call this function to avoid
		2797	detecting this change (while introducing a race condition if you are not
		2798	the only one changing the path). Can also be useful simply to find out the
		2799	new values.
		2800
		2801	=item ev_statdata attr [read-only]
		2802
		2803	The most-recently detected attributes of the file. Although the type is
		2804	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2805	suitable for your system, but you can only rely on the POSIX-standardised
		2806	members to be present. If the C<st_nlink> member is C<0>, then there was
		2807	some error while C<stat>ing the file.
		2808
		2809	=item ev_statdata prev [read-only]
		2810
		2811	The previous attributes of the file. The callback gets invoked whenever
		2812	C<prev> != C<attr>, or, more precisely, one or more of these members
		2813	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2814	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2815
		2816	=item ev_tstamp interval [read-only]
		2817
		2818	The specified interval.
		2819
		2820	=item const char *path [read-only]
		2821
		2822	The file system path that is being watched.
		2823
		2824	=back
		2825
		2826	=head3 Examples
		2827
		2828	Example: Watch C</etc/passwd> for attribute changes.
		2829
		2830	static void
		2831	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2832	{
		2833	/* /etc/passwd changed in some way */
		2834	if (w->attr.st_nlink)
		2835	{
		2836	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2837	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2838	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2839	}
		2840	else
		2841	/* you shalt not abuse printf for puts */
		2842	puts ("wow, /etc/passwd is not there, expect problems. "
		2843	"if this is windows, they already arrived\n");
		2844	}
		2845
		2846	...
		2847	ev_stat passwd;
		2848
		2849	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2850	ev_stat_start (loop, &passwd);
		2851
		2852	Example: Like above, but additionally use a one-second delay so we do not
		2853	miss updates (however, frequent updates will delay processing, too, so
		2854	one might do the work both on C<ev_stat> callback invocation I<and> on
		2855	C<ev_timer> callback invocation).
		2856
		2857	static ev_stat passwd;
		2858	static ev_timer timer;
		2859
		2860	static void
		2861	timer_cb (EV_P_ ev_timer *w, int revents)
		2862	{
		2863	ev_timer_stop (EV_A_ w);
		2864
		2865	/* now it's one second after the most recent passwd change */
		2866	}
		2867
		2868	static void
		2869	stat_cb (EV_P_ ev_stat *w, int revents)
		2870	{
		2871	/* reset the one-second timer */
		2872	ev_timer_again (EV_A_ &timer);
		2873	}
		2874
		2875	...
		2876	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2877	ev_stat_start (loop, &passwd);
		2878	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2879
		2880
630	=head2 ev_idle - when you've got nothing better to do	2881	=head2 C<ev_idle> - when you've got nothing better to do...
631		2882
632	Idle watchers trigger events when there are no other I/O or timer (or	2883	Idle watchers trigger events when no other events of the same or higher
633	periodic) events pending. That is, as long as your process is busy	2884	priority are pending (prepare, check and other idle watchers do not count
634	handling sockets or timeouts it will not be called. But when your process	2885	as receiving "events").
635	is idle all idle watchers are being called again and again - until	2886
		2887	That is, as long as your process is busy handling sockets or timeouts
		2888	(or even signals, imagine) of the same or higher priority it will not be
		2889	triggered. But when your process is idle (or only lower-priority watchers
		2890	are pending), the idle watchers are being called once per event loop
636	stopped, that is, or your process receives more events.	2891	iteration - until stopped, that is, or your process receives more events
		2892	and becomes busy again with higher priority stuff.
637		2893
638	The most noteworthy effect is that as long as any idle watchers are	2894	The most noteworthy effect is that as long as any idle watchers are
639	active, the process will not block when waiting for new events.	2895	active, the process will not block when waiting for new events.
640		2896
641	Apart from keeping your process non-blocking (which is a useful	2897	Apart from keeping your process non-blocking (which is a useful
642	effect on its own sometimes), idle watchers are a good place to do	2898	effect on its own sometimes), idle watchers are a good place to do
643	"pseudo-background processing", or delay processing stuff to after the	2899	"pseudo-background processing", or delay processing stuff to after the
644	event loop has handled all outstanding events.	2900	event loop has handled all outstanding events.
645		2901
		2902	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2903
		2904	As long as there is at least one active idle watcher, libev will never
		2905	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2906	For this to work, the idle watcher doesn't need to be invoked at all - the
		2907	lowest priority will do.
		2908
		2909	This mode of operation can be useful together with an C<ev_check> watcher,
		2910	to do something on each event loop iteration - for example to balance load
		2911	between different connections.
		2912
		2913	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2914	example.
		2915
		2916	=head3 Watcher-Specific Functions and Data Members
		2917
646	=over 4	2918	=over 4
647		2919
648	=item ev_idle_init (ev_signal *, callback)	2920	=item ev_idle_init (ev_idle *, callback)
649		2921
650	Initialises and configures the idle watcher - it has no parameters of any	2922	Initialises and configures the idle watcher - it has no parameters of any
651	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2923	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
652	believe me.	2924	believe me.
653		2925
654	=back	2926	=back
655		2927
656	=head2 prepare and check - your hooks into the event loop	2928	=head3 Examples
657		2929
		2930	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2931	callback, free it. Also, use no error checking, as usual.
		2932
		2933	static void
		2934	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2935	{
		2936	// stop the watcher
		2937	ev_idle_stop (loop, w);
		2938
		2939	// now we can free it
		2940	free (w);
		2941
		2942	// now do something you wanted to do when the program has
		2943	// no longer anything immediate to do.
		2944	}
		2945
		2946	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2947	ev_idle_init (idle_watcher, idle_cb);
		2948	ev_idle_start (loop, idle_watcher);
		2949
		2950
		2951	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
		2952
658	Prepare and check watchers usually (but not always) are used in	2953	Prepare and check watchers are often (but not always) used in pairs:
659	tandom. Prepare watchers get invoked before the process blocks and check	2954	prepare watchers get invoked before the process blocks and check watchers
660	watchers afterwards.	2955	afterwards.
661		2956
		2957	You I<must not> call C<ev_run> (or similar functions that enter the
		2958	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
		2959	C<ev_check> watchers. Other loops than the current one are fine,
		2960	however. The rationale behind this is that you do not need to check
		2961	for recursion in those watchers, i.e. the sequence will always be
		2962	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
		2963	kind they will always be called in pairs bracketing the blocking call.
		2964
662	Their main purpose is to integrate other event mechanisms into libev. This	2965	Their main purpose is to integrate other event mechanisms into libev and
663	could be used, for example, to track variable changes, implement your own	2966	their use is somewhat advanced. They could be used, for example, to track
664	watchers, integrate net-snmp or a coroutine library and lots more.	2967	variable changes, implement your own watchers, integrate net-snmp or a
		2968	coroutine library and lots more. They are also occasionally useful if
		2969	you cache some data and want to flush it before blocking (for example,
		2970	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2971	watcher).
665		2972
666	This is done by examining in each prepare call which file descriptors need	2973	This is done by examining in each prepare call which file descriptors
667	to be watched by the other library, registering ev_io watchers for them	2974	need to be watched by the other library, registering C<ev_io> watchers
668	and starting an ev_timer watcher for any timeouts (many libraries provide	2975	for them and starting an C<ev_timer> watcher for any timeouts (many
669	just this functionality). Then, in the check watcher you check for any	2976	libraries provide exactly this functionality). Then, in the check watcher,
670	events that occured (by making your callbacks set soem flags for example)	2977	you check for any events that occurred (by checking the pending status
671	and call back into the library.	2978	of all watchers and stopping them) and call back into the library. The
		2979	I/O and timer callbacks will never actually be called (but must be valid
		2980	nevertheless, because you never know, you know?).
672		2981
673	As another example, the perl Coro module uses these hooks to integrate	2982	As another example, the Perl Coro module uses these hooks to integrate
674	coroutines into libev programs, by yielding to other active coroutines	2983	coroutines into libev programs, by yielding to other active coroutines
675	during each prepare and only letting the process block if no coroutines	2984	during each prepare and only letting the process block if no coroutines
676	are ready to run.	2985	are ready to run (it's actually more complicated: it only runs coroutines
		2986	with priority higher than or equal to the event loop and one coroutine
		2987	of lower priority, but only once, using idle watchers to keep the event
		2988	loop from blocking if lower-priority coroutines are active, thus mapping
		2989	low-priority coroutines to idle/background tasks).
		2990
		2991	When used for this purpose, it is recommended to give C<ev_check> watchers
		2992	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
		2993	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		2994	watchers).
		2995
		2996	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2997	activate ("feed") events into libev. While libev fully supports this, they
		2998	might get executed before other C<ev_check> watchers did their job. As
		2999	C<ev_check> watchers are often used to embed other (non-libev) event
		3000	loops those other event loops might be in an unusable state until their
		3001	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		3002	others).
		3003
		3004	=head3 Abusing an C<ev_check> watcher for its side-effect
		3005
		3006	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		3007	useful because they are called once per event loop iteration. For
		3008	example, if you want to handle a large number of connections fairly, you
		3009	normally only do a bit of work for each active connection, and if there
		3010	is more work to do, you wait for the next event loop iteration, so other
		3011	connections have a chance of making progress.
		3012
		3013	Using an C<ev_check> watcher is almost enough: it will be called on the
		3014	next event loop iteration. However, that isn't as soon as possible -
		3015	without external events, your C<ev_check> watcher will not be invoked.
		3016
		3017	This is where C<ev_idle> watchers come in handy - all you need is a
		3018	single global idle watcher that is active as long as you have one active
		3019	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		3020	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		3021	invoked. Neither watcher alone can do that.
		3022
		3023	=head3 Watcher-Specific Functions and Data Members
677		3024
678	=over 4	3025	=over 4
679		3026
680	=item ev_prepare_init (ev_prepare *, callback)	3027	=item ev_prepare_init (ev_prepare *, callback)
681		3028
682	=item ev_check_init (ev_check *, callback)	3029	=item ev_check_init (ev_check *, callback)
683		3030
684	Initialises and configures the prepare or check watcher - they have no	3031	Initialises and configures the prepare or check watcher - they have no
685	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	3032	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
686	macros, but using them is utterly, utterly pointless.	3033	macros, but using them is utterly, utterly, utterly and completely
		3034	pointless.
687		3035
688	=back	3036	=back
689		3037
		3038	=head3 Examples
		3039
		3040	There are a number of principal ways to embed other event loops or modules
		3041	into libev. Here are some ideas on how to include libadns into libev
		3042	(there is a Perl module named C<EV::ADNS> that does this, which you could
		3043	use as a working example. Another Perl module named C<EV::Glib> embeds a
		3044	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		3045	Glib event loop).
		3046
		3047	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		3048	and in a check watcher, destroy them and call into libadns. What follows
		3049	is pseudo-code only of course. This requires you to either use a low
		3050	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		3051	the callbacks for the IO/timeout watchers might not have been called yet.
		3052
		3053	static ev_io iow [nfd];
		3054	static ev_timer tw;
		3055
		3056	static void
		3057	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		3058	{
		3059	}
		3060
		3061	// create io watchers for each fd and a timer before blocking
		3062	static void
		3063	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		3064	{
		3065	int timeout = 3600000;
		3066	struct pollfd fds [nfd];
		3067	// actual code will need to loop here and realloc etc.
		3068	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		3069
		3070	/* the callback is illegal, but won't be called as we stop during check */
		3071	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		3072	ev_timer_start (loop, &tw);
		3073
		3074	// create one ev_io per pollfd
		3075	for (int i = 0; i < nfd; ++i)
		3076	{
		3077	ev_io_init (iow + i, io_cb, fds [i].fd,
		3078	((fds [i].events & POLLIN ? EV_READ : 0)
		3079	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		3080
		3081	fds [i].revents = 0;
		3082	ev_io_start (loop, iow + i);
		3083	}
		3084	}
		3085
		3086	// stop all watchers after blocking
		3087	static void
		3088	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		3089	{
		3090	ev_timer_stop (loop, &tw);
		3091
		3092	for (int i = 0; i < nfd; ++i)
		3093	{
		3094	// set the relevant poll flags
		3095	// could also call adns_processreadable etc. here
		3096	struct pollfd *fd = fds + i;
		3097	int revents = ev_clear_pending (iow + i);
		3098	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		3099	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		3100
		3101	// now stop the watcher
		3102	ev_io_stop (loop, iow + i);
		3103	}
		3104
		3105	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		3106	}
		3107
		3108	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		3109	in the prepare watcher and would dispose of the check watcher.
		3110
		3111	Method 3: If the module to be embedded supports explicit event
		3112	notification (libadns does), you can also make use of the actual watcher
		3113	callbacks, and only destroy/create the watchers in the prepare watcher.
		3114
		3115	static void
		3116	timer_cb (EV_P_ ev_timer *w, int revents)
		3117	{
		3118	adns_state ads = (adns_state)w->data;
		3119	update_now (EV_A);
		3120
		3121	adns_processtimeouts (ads, &tv_now);
		3122	}
		3123
		3124	static void
		3125	io_cb (EV_P_ ev_io *w, int revents)
		3126	{
		3127	adns_state ads = (adns_state)w->data;
		3128	update_now (EV_A);
		3129
		3130	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		3131	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		3132	}
		3133
		3134	// do not ever call adns_afterpoll
		3135
		3136	Method 4: Do not use a prepare or check watcher because the module you
		3137	want to embed is not flexible enough to support it. Instead, you can
		3138	override their poll function. The drawback with this solution is that the
		3139	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		3140	this approach, effectively embedding EV as a client into the horrible
		3141	libglib event loop.
		3142
		3143	static gint
		3144	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		3145	{
		3146	int got_events = 0;
		3147
		3148	for (n = 0; n < nfds; ++n)
		3149	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		3150
		3151	if (timeout >= 0)
		3152	// create/start timer
		3153
		3154	// poll
		3155	ev_run (EV_A_ 0);
		3156
		3157	// stop timer again
		3158	if (timeout >= 0)
		3159	ev_timer_stop (EV_A_ &to);
		3160
		3161	// stop io watchers again - their callbacks should have set
		3162	for (n = 0; n < nfds; ++n)
		3163	ev_io_stop (EV_A_ iow [n]);
		3164
		3165	return got_events;
		3166	}
		3167
		3168
		3169	=head2 C<ev_embed> - when one backend isn't enough...
		3170
		3171	This is a rather advanced watcher type that lets you embed one event loop
		3172	into another (currently only C<ev_io> events are supported in the embedded
		3173	loop, other types of watchers might be handled in a delayed or incorrect
		3174	fashion and must not be used).
		3175
		3176	There are primarily two reasons you would want that: work around bugs and
		3177	prioritise I/O.
		3178
		3179	As an example for a bug workaround, the kqueue backend might only support
		3180	sockets on some platform, so it is unusable as generic backend, but you
		3181	still want to make use of it because you have many sockets and it scales
		3182	so nicely. In this case, you would create a kqueue-based loop and embed
		3183	it into your default loop (which might use e.g. poll). Overall operation
		3184	will be a bit slower because first libev has to call C<poll> and then
		3185	C<kevent>, but at least you can use both mechanisms for what they are
		3186	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		3187
		3188	As for prioritising I/O: under rare circumstances you have the case where
		3189	some fds have to be watched and handled very quickly (with low latency),
		3190	and even priorities and idle watchers might have too much overhead. In
		3191	this case you would put all the high priority stuff in one loop and all
		3192	the rest in a second one, and embed the second one in the first.
		3193
		3194	As long as the watcher is active, the callback will be invoked every
		3195	time there might be events pending in the embedded loop. The callback
		3196	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		3197	sweep and invoke their callbacks (the callback doesn't need to invoke the
		3198	C<ev_embed_sweep> function directly, it could also start an idle watcher
		3199	to give the embedded loop strictly lower priority for example).
		3200
		3201	You can also set the callback to C<0>, in which case the embed watcher
		3202	will automatically execute the embedded loop sweep whenever necessary.
		3203
		3204	Fork detection will be handled transparently while the C<ev_embed> watcher
		3205	is active, i.e., the embedded loop will automatically be forked when the
		3206	embedding loop forks. In other cases, the user is responsible for calling
		3207	C<ev_loop_fork> on the embedded loop.
		3208
		3209	Unfortunately, not all backends are embeddable: only the ones returned by
		3210	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		3211	portable one.
		3212
		3213	So when you want to use this feature you will always have to be prepared
		3214	that you cannot get an embeddable loop. The recommended way to get around
		3215	this is to have a separate variables for your embeddable loop, try to
		3216	create it, and if that fails, use the normal loop for everything.
		3217
		3218	=head3 C<ev_embed> and fork
		3219
		3220	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3221	automatically be applied to the embedded loop as well, so no special
		3222	fork handling is required in that case. When the watcher is not running,
		3223	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3224	as applicable.
		3225
		3226	=head3 Watcher-Specific Functions and Data Members
		3227
		3228	=over 4
		3229
		3230	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		3231
		3232	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
		3233
		3234	Configures the watcher to embed the given loop, which must be
		3235	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		3236	invoked automatically, otherwise it is the responsibility of the callback
		3237	to invoke it (it will continue to be called until the sweep has been done,
		3238	if you do not want that, you need to temporarily stop the embed watcher).
		3239
		3240	=item ev_embed_sweep (loop, ev_embed *)
		3241
		3242	Make a single, non-blocking sweep over the embedded loop. This works
		3243	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3244	appropriate way for embedded loops.
		3245
		3246	=item struct ev_loop *other [read-only]
		3247
		3248	The embedded event loop.
		3249
		3250	=back
		3251
		3252	=head3 Examples
		3253
		3254	Example: Try to get an embeddable event loop and embed it into the default
		3255	event loop. If that is not possible, use the default loop. The default
		3256	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3257	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3258	used).
		3259
		3260	struct ev_loop *loop_hi = ev_default_init (0);
		3261	struct ev_loop *loop_lo = 0;
		3262	ev_embed embed;
		3263
		3264	// see if there is a chance of getting one that works
		3265	// (remember that a flags value of 0 means autodetection)
		3266	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3267	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3268	: 0;
		3269
		3270	// if we got one, then embed it, otherwise default to loop_hi
		3271	if (loop_lo)
		3272	{
		3273	ev_embed_init (&embed, 0, loop_lo);
		3274	ev_embed_start (loop_hi, &embed);
		3275	}
		3276	else
		3277	loop_lo = loop_hi;
		3278
		3279	Example: Check if kqueue is available but not recommended and create
		3280	a kqueue backend for use with sockets (which usually work with any
		3281	kqueue implementation). Store the kqueue/socket-only event loop in
		3282	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3283
		3284	struct ev_loop *loop = ev_default_init (0);
		3285	struct ev_loop *loop_socket = 0;
		3286	ev_embed embed;
		3287
		3288	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3289	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3290	{
		3291	ev_embed_init (&embed, 0, loop_socket);
		3292	ev_embed_start (loop, &embed);
		3293	}
		3294
		3295	if (!loop_socket)
		3296	loop_socket = loop;
		3297
		3298	// now use loop_socket for all sockets, and loop for everything else
		3299
		3300
		3301	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3302
		3303	Fork watchers are called when a C<fork ()> was detected (usually because
		3304	whoever is a good citizen cared to tell libev about it by calling
		3305	C<ev_loop_fork>). The invocation is done before the event loop blocks next
		3306	and before C<ev_check> watchers are being called, and only in the child
		3307	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
		3308	and calls it in the wrong process, the fork handlers will be invoked, too,
		3309	of course.
		3310
		3311	=head3 The special problem of life after fork - how is it possible?
		3312
		3313	Most uses of C<fork ()> consist of forking, then some simple calls to set
		3314	up/change the process environment, followed by a call to C<exec()>. This
		3315	sequence should be handled by libev without any problems.
		3316
		3317	This changes when the application actually wants to do event handling
		3318	in the child, or both parent in child, in effect "continuing" after the
		3319	fork.
		3320
		3321	The default mode of operation (for libev, with application help to detect
		3322	forks) is to duplicate all the state in the child, as would be expected
		3323	when I<either> the parent I<or> the child process continues.
		3324
		3325	When both processes want to continue using libev, then this is usually the
		3326	wrong result. In that case, usually one process (typically the parent) is
		3327	supposed to continue with all watchers in place as before, while the other
		3328	process typically wants to start fresh, i.e. without any active watchers.
		3329
		3330	The cleanest and most efficient way to achieve that with libev is to
		3331	simply create a new event loop, which of course will be "empty", and
		3332	use that for new watchers. This has the advantage of not touching more
		3333	memory than necessary, and thus avoiding the copy-on-write, and the
		3334	disadvantage of having to use multiple event loops (which do not support
		3335	signal watchers).
		3336
		3337	When this is not possible, or you want to use the default loop for
		3338	other reasons, then in the process that wants to start "fresh", call
		3339	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3340	Destroying the default loop will "orphan" (not stop) all registered
		3341	watchers, so you have to be careful not to execute code that modifies
		3342	those watchers. Note also that in that case, you have to re-register any
		3343	signal watchers.
		3344
		3345	=head3 Watcher-Specific Functions and Data Members
		3346
		3347	=over 4
		3348
		3349	=item ev_fork_init (ev_fork *, callback)
		3350
		3351	Initialises and configures the fork watcher - it has no parameters of any
		3352	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3353	really.
		3354
		3355	=back
		3356
		3357
		3358	=head2 C<ev_cleanup> - even the best things end
		3359
		3360	Cleanup watchers are called just before the event loop is being destroyed
		3361	by a call to C<ev_loop_destroy>.
		3362
		3363	While there is no guarantee that the event loop gets destroyed, cleanup
		3364	watchers provide a convenient method to install cleanup hooks for your
		3365	program, worker threads and so on - you just to make sure to destroy the
		3366	loop when you want them to be invoked.
		3367
		3368	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3369	all other watchers, they do not keep a reference to the event loop (which
		3370	makes a lot of sense if you think about it). Like all other watchers, you
		3371	can call libev functions in the callback, except C<ev_cleanup_start>.
		3372
		3373	=head3 Watcher-Specific Functions and Data Members
		3374
		3375	=over 4
		3376
		3377	=item ev_cleanup_init (ev_cleanup *, callback)
		3378
		3379	Initialises and configures the cleanup watcher - it has no parameters of
		3380	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3381	pointless, I assure you.
		3382
		3383	=back
		3384
		3385	Example: Register an atexit handler to destroy the default loop, so any
		3386	cleanup functions are called.
		3387
		3388	static void
		3389	program_exits (void)
		3390	{
		3391	ev_loop_destroy (EV_DEFAULT_UC);
		3392	}
		3393
		3394	...
		3395	atexit (program_exits);
		3396
		3397
		3398	=head2 C<ev_async> - how to wake up an event loop
		3399
		3400	In general, you cannot use an C<ev_loop> from multiple threads or other
		3401	asynchronous sources such as signal handlers (as opposed to multiple event
		3402	loops - those are of course safe to use in different threads).
		3403
		3404	Sometimes, however, you need to wake up an event loop you do not control,
		3405	for example because it belongs to another thread. This is what C<ev_async>
		3406	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3407	it by calling C<ev_async_send>, which is thread- and signal safe.
		3408
		3409	This functionality is very similar to C<ev_signal> watchers, as signals,
		3410	too, are asynchronous in nature, and signals, too, will be compressed
		3411	(i.e. the number of callback invocations may be less than the number of
		3412	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
		3413	of "global async watchers" by using a watcher on an otherwise unused
		3414	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3415	even without knowing which loop owns the signal.
		3416
		3417	=head3 Queueing
		3418
		3419	C<ev_async> does not support queueing of data in any way. The reason
		3420	is that the author does not know of a simple (or any) algorithm for a
		3421	multiple-writer-single-reader queue that works in all cases and doesn't
		3422	need elaborate support such as pthreads or unportable memory access
		3423	semantics.
		3424
		3425	That means that if you want to queue data, you have to provide your own
		3426	queue. But at least I can tell you how to implement locking around your
		3427	queue:
		3428
		3429	=over 4
		3430
		3431	=item queueing from a signal handler context
		3432
		3433	To implement race-free queueing, you simply add to the queue in the signal
		3434	handler but you block the signal handler in the watcher callback. Here is
		3435	an example that does that for some fictitious SIGUSR1 handler:
		3436
		3437	static ev_async mysig;
		3438
		3439	static void
		3440	sigusr1_handler (void)
		3441	{
		3442	sometype data;
		3443
		3444	// no locking etc.
		3445	queue_put (data);
		3446	ev_async_send (EV_DEFAULT_ &mysig);
		3447	}
		3448
		3449	static void
		3450	mysig_cb (EV_P_ ev_async *w, int revents)
		3451	{
		3452	sometype data;
		3453	sigset_t block, prev;
		3454
		3455	sigemptyset (&block);
		3456	sigaddset (&block, SIGUSR1);
		3457	sigprocmask (SIG_BLOCK, &block, &prev);
		3458
		3459	while (queue_get (&data))
		3460	process (data);
		3461
		3462	if (sigismember (&prev, SIGUSR1)
		3463	sigprocmask (SIG_UNBLOCK, &block, 0);
		3464	}
		3465
		3466	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3467	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3468	either...).
		3469
		3470	=item queueing from a thread context
		3471
		3472	The strategy for threads is different, as you cannot (easily) block
		3473	threads but you can easily preempt them, so to queue safely you need to
		3474	employ a traditional mutex lock, such as in this pthread example:
		3475
		3476	static ev_async mysig;
		3477	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3478
		3479	static void
		3480	otherthread (void)
		3481	{
		3482	// only need to lock the actual queueing operation
		3483	pthread_mutex_lock (&mymutex);
		3484	queue_put (data);
		3485	pthread_mutex_unlock (&mymutex);
		3486
		3487	ev_async_send (EV_DEFAULT_ &mysig);
		3488	}
		3489
		3490	static void
		3491	mysig_cb (EV_P_ ev_async *w, int revents)
		3492	{
		3493	pthread_mutex_lock (&mymutex);
		3494
		3495	while (queue_get (&data))
		3496	process (data);
		3497
		3498	pthread_mutex_unlock (&mymutex);
		3499	}
		3500
		3501	=back
		3502
		3503
		3504	=head3 Watcher-Specific Functions and Data Members
		3505
		3506	=over 4
		3507
		3508	=item ev_async_init (ev_async *, callback)
		3509
		3510	Initialises and configures the async watcher - it has no parameters of any
		3511	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3512	trust me.
		3513
		3514	=item ev_async_send (loop, ev_async *)
		3515
		3516	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3517	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3518	returns.
		3519
		3520	Unlike C<ev_feed_event>, this call is safe to do from other threads,
		3521	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
		3522	embedding section below on what exactly this means).
		3523
		3524	Note that, as with other watchers in libev, multiple events might get
		3525	compressed into a single callback invocation (another way to look at
		3526	this is that C<ev_async> watchers are level-triggered: they are set on
		3527	C<ev_async_send>, reset when the event loop detects that).
		3528
		3529	This call incurs the overhead of at most one extra system call per event
		3530	loop iteration, if the event loop is blocked, and no syscall at all if
		3531	the event loop (or your program) is processing events. That means that
		3532	repeated calls are basically free (there is no need to avoid calls for
		3533	performance reasons) and that the overhead becomes smaller (typically
		3534	zero) under load.
		3535
		3536	=item bool = ev_async_pending (ev_async *)
		3537
		3538	Returns a non-zero value when C<ev_async_send> has been called on the
		3539	watcher but the event has not yet been processed (or even noted) by the
		3540	event loop.
		3541
		3542	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3543	the loop iterates next and checks for the watcher to have become active,
		3544	it will reset the flag again. C<ev_async_pending> can be used to very
		3545	quickly check whether invoking the loop might be a good idea.
		3546
		3547	Not that this does I<not> check whether the watcher itself is pending,
		3548	only whether it has been requested to make this watcher pending: there
		3549	is a time window between the event loop checking and resetting the async
		3550	notification, and the callback being invoked.
		3551
		3552	=back
		3553
		3554
690	=head1 OTHER FUNCTIONS	3555	=head1 OTHER FUNCTIONS
691		3556
692	There are some other fucntions of possible interest. Described. Here. Now.	3557	There are some other functions of possible interest. Described. Here. Now.
693		3558
694	=over 4	3559	=over 4
695		3560
696	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3561	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
697		3562
698	This function combines a simple timer and an I/O watcher, calls your	3563	This function combines a simple timer and an I/O watcher, calls your
699	callback on whichever event happens first and automatically stop both	3564	callback on whichever event happens first and automatically stops both
700	watchers. This is useful if you want to wait for a single event on an fd	3565	watchers. This is useful if you want to wait for a single event on an fd
701	or timeout without havign to allocate/configure/start/stop/free one or	3566	or timeout without having to allocate/configure/start/stop/free one or
702	more watchers yourself.	3567	more watchers yourself.
703		3568
704	If C<fd> is less than 0, then no I/O watcher will be started and events is	3569	If C<fd> is less than 0, then no I/O watcher will be started and the
705	ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set	3570	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
706	will be craeted and started.	3571	the given C<fd> and C<events> set will be created and started.
707		3572
708	If C<timeout> is less than 0, then no timeout watcher will be	3573	If C<timeout> is less than 0, then no timeout watcher will be
709	started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat	3574	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
710	= 0) will be started.	3575	repeat = 0) will be started. C<0> is a valid timeout.
711		3576
712	The callback has the type C<void (*cb)(int revents, void *arg)> and	3577	The callback has the type C<void (*cb)(int revents, void *arg)> and is
713	gets passed an events set (normally a combination of EV_ERROR, EV_READ,	3578	passed an C<revents> set like normal event callbacks (a combination of
714	EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>:	3579	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
		3580	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3581	a timeout and an io event at the same time - you probably should give io
		3582	events precedence.
715		3583
		3584	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3585
716	static void stdin_ready (int revents, void *arg)	3586	static void stdin_ready (int revents, void *arg)
		3587	{
		3588	if (revents & EV_READ)
		3589	/* stdin might have data for us, joy! */;
		3590	else if (revents & EV_TIMER)
		3591	/* doh, nothing entered */;
		3592	}
		3593
		3594	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
		3595
		3596	=item ev_feed_fd_event (loop, int fd, int revents)
		3597
		3598	Feed an event on the given fd, as if a file descriptor backend detected
		3599	the given events.
		3600
		3601	=item ev_feed_signal_event (loop, int signum)
		3602
		3603	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3604	which is async-safe.
		3605
		3606	=back
		3607
		3608
		3609	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3610
		3611	This section explains some common idioms that are not immediately
		3612	obvious. Note that examples are sprinkled over the whole manual, and this
		3613	section only contains stuff that wouldn't fit anywhere else.
		3614
		3615	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3616
		3617	Each watcher has, by default, a C<void *data> member that you can read
		3618	or modify at any time: libev will completely ignore it. This can be used
		3619	to associate arbitrary data with your watcher. If you need more data and
		3620	don't want to allocate memory separately and store a pointer to it in that
		3621	data member, you can also "subclass" the watcher type and provide your own
		3622	data:
		3623
		3624	struct my_io
		3625	{
		3626	ev_io io;
		3627	int otherfd;
		3628	void *somedata;
		3629	struct whatever *mostinteresting;
		3630	};
		3631
		3632	...
		3633	struct my_io w;
		3634	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3635
		3636	And since your callback will be called with a pointer to the watcher, you
		3637	can cast it back to your own type:
		3638
		3639	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3640	{
		3641	struct my_io *w = (struct my_io *)w_;
		3642	...
		3643	}
		3644
		3645	More interesting and less C-conformant ways of casting your callback
		3646	function type instead have been omitted.
		3647
		3648	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3649
		3650	Another common scenario is to use some data structure with multiple
		3651	embedded watchers, in effect creating your own watcher that combines
		3652	multiple libev event sources into one "super-watcher":
		3653
		3654	struct my_biggy
		3655	{
		3656	int some_data;
		3657	ev_timer t1;
		3658	ev_timer t2;
		3659	}
		3660
		3661	In this case getting the pointer to C<my_biggy> is a bit more
		3662	complicated: Either you store the address of your C<my_biggy> struct in
		3663	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3664	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3665	real programmers):
		3666
		3667	#include <stddef.h>
		3668
		3669	static void
		3670	t1_cb (EV_P_ ev_timer *w, int revents)
		3671	{
		3672	struct my_biggy big = (struct my_biggy *)
		3673	(((char *)w) - offsetof (struct my_biggy, t1));
		3674	}
		3675
		3676	static void
		3677	t2_cb (EV_P_ ev_timer *w, int revents)
		3678	{
		3679	struct my_biggy big = (struct my_biggy *)
		3680	(((char *)w) - offsetof (struct my_biggy, t2));
		3681	}
		3682
		3683	=head2 AVOIDING FINISHING BEFORE RETURNING
		3684
		3685	Often you have structures like this in event-based programs:
		3686
		3687	callback ()
717	{	3688	{
718	if (revents & EV_TIMEOUT)	3689	free (request);
719	/* doh, nothing entered */
720	else if (revents & EV_READ)
721	/* stdin might have data for us, joy! */
722	}	3690	}
723		3691
724	ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0);	3692	request = start_new_request (..., callback);
725		3693
726	=item ev_feed_event (loop, watcher, int events)	3694	The intent is to start some "lengthy" operation. The C<request> could be
		3695	used to cancel the operation, or do other things with it.
727		3696
728	Feeds the given event set into the event loop, as if the specified event	3697	It's not uncommon to have code paths in C<start_new_request> that
729	has happened for the specified watcher (which must be a pointer to an	3698	immediately invoke the callback, for example, to report errors. Or you add
730	initialised but not necessarily active event watcher).	3699	some caching layer that finds that it can skip the lengthy aspects of the
		3700	operation and simply invoke the callback with the result.
731		3701
732	=item ev_feed_fd_event (loop, int fd, int revents)	3702	The problem here is that this will happen I<before> C<start_new_request>
		3703	has returned, so C<request> is not set.
733		3704
734	Feed an event on the given fd, as if a file descriptor backend detected it.	3705	Even if you pass the request by some safer means to the callback, you
		3706	might want to do something to the request after starting it, such as
		3707	canceling it, which probably isn't working so well when the callback has
		3708	already been invoked.
735		3709
736	=item ev_feed_signal_event (loop, int signum)	3710	A common way around all these issues is to make sure that
		3711	C<start_new_request> I<always> returns before the callback is invoked. If
		3712	C<start_new_request> immediately knows the result, it can artificially
		3713	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3714	example, or more sneakily, by reusing an existing (stopped) watcher and
		3715	pushing it into the pending queue:
737		3716
738	Feed an event as if the given signal occured (loop must be the default loop!).	3717	ev_set_cb (watcher, callback);
		3718	ev_feed_event (EV_A_ watcher, 0);
		3719
		3720	This way, C<start_new_request> can safely return before the callback is
		3721	invoked, while not delaying callback invocation too much.
		3722
		3723	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3724
		3725	Often (especially in GUI toolkits) there are places where you have
		3726	I<modal> interaction, which is most easily implemented by recursively
		3727	invoking C<ev_run>.
		3728
		3729	This brings the problem of exiting - a callback might want to finish the
		3730	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3731	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3732	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3733	other combination: In these cases, a simple C<ev_break> will not work.
		3734
		3735	The solution is to maintain "break this loop" variable for each C<ev_run>
		3736	invocation, and use a loop around C<ev_run> until the condition is
		3737	triggered, using C<EVRUN_ONCE>:
		3738
		3739	// main loop
		3740	int exit_main_loop = 0;
		3741
		3742	while (!exit_main_loop)
		3743	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3744
		3745	// in a modal watcher
		3746	int exit_nested_loop = 0;
		3747
		3748	while (!exit_nested_loop)
		3749	ev_run (EV_A_ EVRUN_ONCE);
		3750
		3751	To exit from any of these loops, just set the corresponding exit variable:
		3752
		3753	// exit modal loop
		3754	exit_nested_loop = 1;
		3755
		3756	// exit main program, after modal loop is finished
		3757	exit_main_loop = 1;
		3758
		3759	// exit both
		3760	exit_main_loop = exit_nested_loop = 1;
		3761
		3762	=head2 THREAD LOCKING EXAMPLE
		3763
		3764	Here is a fictitious example of how to run an event loop in a different
		3765	thread from where callbacks are being invoked and watchers are
		3766	created/added/removed.
		3767
		3768	For a real-world example, see the C<EV::Loop::Async> perl module,
		3769	which uses exactly this technique (which is suited for many high-level
		3770	languages).
		3771
		3772	The example uses a pthread mutex to protect the loop data, a condition
		3773	variable to wait for callback invocations, an async watcher to notify the
		3774	event loop thread and an unspecified mechanism to wake up the main thread.
		3775
		3776	First, you need to associate some data with the event loop:
		3777
		3778	typedef struct {
		3779	mutex_t lock; /* global loop lock */
		3780	ev_async async_w;
		3781	thread_t tid;
		3782	cond_t invoke_cv;
		3783	} userdata;
		3784
		3785	void prepare_loop (EV_P)
		3786	{
		3787	// for simplicity, we use a static userdata struct.
		3788	static userdata u;
		3789
		3790	ev_async_init (&u->async_w, async_cb);
		3791	ev_async_start (EV_A_ &u->async_w);
		3792
		3793	pthread_mutex_init (&u->lock, 0);
		3794	pthread_cond_init (&u->invoke_cv, 0);
		3795
		3796	// now associate this with the loop
		3797	ev_set_userdata (EV_A_ u);
		3798	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3799	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3800
		3801	// then create the thread running ev_run
		3802	pthread_create (&u->tid, 0, l_run, EV_A);
		3803	}
		3804
		3805	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3806	solely to wake up the event loop so it takes notice of any new watchers
		3807	that might have been added:
		3808
		3809	static void
		3810	async_cb (EV_P_ ev_async *w, int revents)
		3811	{
		3812	// just used for the side effects
		3813	}
		3814
		3815	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3816	protecting the loop data, respectively.
		3817
		3818	static void
		3819	l_release (EV_P)
		3820	{
		3821	userdata *u = ev_userdata (EV_A);
		3822	pthread_mutex_unlock (&u->lock);
		3823	}
		3824
		3825	static void
		3826	l_acquire (EV_P)
		3827	{
		3828	userdata *u = ev_userdata (EV_A);
		3829	pthread_mutex_lock (&u->lock);
		3830	}
		3831
		3832	The event loop thread first acquires the mutex, and then jumps straight
		3833	into C<ev_run>:
		3834
		3835	void *
		3836	l_run (void *thr_arg)
		3837	{
		3838	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3839
		3840	l_acquire (EV_A);
		3841	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3842	ev_run (EV_A_ 0);
		3843	l_release (EV_A);
		3844
		3845	return 0;
		3846	}
		3847
		3848	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3849	signal the main thread via some unspecified mechanism (signals? pipe
		3850	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3851	have been called (in a while loop because a) spurious wakeups are possible
		3852	and b) skipping inter-thread-communication when there are no pending
		3853	watchers is very beneficial):
		3854
		3855	static void
		3856	l_invoke (EV_P)
		3857	{
		3858	userdata *u = ev_userdata (EV_A);
		3859
		3860	while (ev_pending_count (EV_A))
		3861	{
		3862	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3863	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3864	}
		3865	}
		3866
		3867	Now, whenever the main thread gets told to invoke pending watchers, it
		3868	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3869	thread to continue:
		3870
		3871	static void
		3872	real_invoke_pending (EV_P)
		3873	{
		3874	userdata *u = ev_userdata (EV_A);
		3875
		3876	pthread_mutex_lock (&u->lock);
		3877	ev_invoke_pending (EV_A);
		3878	pthread_cond_signal (&u->invoke_cv);
		3879	pthread_mutex_unlock (&u->lock);
		3880	}
		3881
		3882	Whenever you want to start/stop a watcher or do other modifications to an
		3883	event loop, you will now have to lock:
		3884
		3885	ev_timer timeout_watcher;
		3886	userdata *u = ev_userdata (EV_A);
		3887
		3888	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3889
		3890	pthread_mutex_lock (&u->lock);
		3891	ev_timer_start (EV_A_ &timeout_watcher);
		3892	ev_async_send (EV_A_ &u->async_w);
		3893	pthread_mutex_unlock (&u->lock);
		3894
		3895	Note that sending the C<ev_async> watcher is required because otherwise
		3896	an event loop currently blocking in the kernel will have no knowledge
		3897	about the newly added timer. By waking up the loop it will pick up any new
		3898	watchers in the next event loop iteration.
		3899
		3900	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3901
		3902	While the overhead of a callback that e.g. schedules a thread is small, it
		3903	is still an overhead. If you embed libev, and your main usage is with some
		3904	kind of threads or coroutines, you might want to customise libev so that
		3905	doesn't need callbacks anymore.
		3906
		3907	Imagine you have coroutines that you can switch to using a function
		3908	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3909	and that due to some magic, the currently active coroutine is stored in a
		3910	global called C<current_coro>. Then you can build your own "wait for libev
		3911	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3912	the differing C<;> conventions):
		3913
		3914	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3915	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3916
		3917	That means instead of having a C callback function, you store the
		3918	coroutine to switch to in each watcher, and instead of having libev call
		3919	your callback, you instead have it switch to that coroutine.
		3920
		3921	A coroutine might now wait for an event with a function called
		3922	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3923	matter when, or whether the watcher is active or not when this function is
		3924	called):
		3925
		3926	void
		3927	wait_for_event (ev_watcher *w)
		3928	{
		3929	ev_set_cb (w, current_coro);
		3930	switch_to (libev_coro);
		3931	}
		3932
		3933	That basically suspends the coroutine inside C<wait_for_event> and
		3934	continues the libev coroutine, which, when appropriate, switches back to
		3935	this or any other coroutine.
		3936
		3937	You can do similar tricks if you have, say, threads with an event queue -
		3938	instead of storing a coroutine, you store the queue object and instead of
		3939	switching to a coroutine, you push the watcher onto the queue and notify
		3940	any waiters.
		3941
		3942	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		3943	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3944
		3945	// my_ev.h
		3946	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3947	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3948	#include "../libev/ev.h"
		3949
		3950	// my_ev.c
		3951	#define EV_H "my_ev.h"
		3952	#include "../libev/ev.c"
		3953
		3954	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		3955	F<my_ev.c> into your project. When properly specifying include paths, you
		3956	can even use F<ev.h> as header file name directly.
		3957
		3958
		3959	=head1 LIBEVENT EMULATION
		3960
		3961	Libev offers a compatibility emulation layer for libevent. It cannot
		3962	emulate the internals of libevent, so here are some usage hints:
		3963
		3964	=over 4
		3965
		3966	=item * Only the libevent-1.4.1-beta API is being emulated.
		3967
		3968	This was the newest libevent version available when libev was implemented,
		3969	and is still mostly unchanged in 2010.
		3970
		3971	=item * Use it by including <event.h>, as usual.
		3972
		3973	=item * The following members are fully supported: ev_base, ev_callback,
		3974	ev_arg, ev_fd, ev_res, ev_events.
		3975
		3976	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		3977	maintained by libev, it does not work exactly the same way as in libevent (consider
		3978	it a private API).
		3979
		3980	=item * Priorities are not currently supported. Initialising priorities
		3981	will fail and all watchers will have the same priority, even though there
		3982	is an ev_pri field.
		3983
		3984	=item * In libevent, the last base created gets the signals, in libev, the
		3985	base that registered the signal gets the signals.
		3986
		3987	=item * Other members are not supported.
		3988
		3989	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		3990	to use the libev header file and library.
739		3991
740	=back	3992	=back
741		3993
		3994	=head1 C++ SUPPORT
		3995
		3996	=head2 C API
		3997
		3998	The normal C API should work fine when used from C++: both ev.h and the
		3999	libev sources can be compiled as C++. Therefore, code that uses the C API
		4000	will work fine.
		4001
		4002	Proper exception specifications might have to be added to callbacks passed
		4003	to libev: exceptions may be thrown only from watcher callbacks, all other
		4004	callbacks (allocator, syserr, loop acquire/release and periodic reschedule
		4005	callbacks) must not throw exceptions, and might need a C<noexcept>
		4006	specification. If you have code that needs to be compiled as both C and
		4007	C++ you can use the C<EV_NOEXCEPT> macro for this:
		4008
		4009	static void
		4010	fatal_error (const char *msg) EV_NOEXCEPT
		4011	{
		4012	perror (msg);
		4013	abort ();
		4014	}
		4015
		4016	...
		4017	ev_set_syserr_cb (fatal_error);
		4018
		4019	The only API functions that can currently throw exceptions are C<ev_run>,
		4020	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		4021	because it runs cleanup watchers).
		4022
		4023	Throwing exceptions in watcher callbacks is only supported if libev itself
		4024	is compiled with a C++ compiler or your C and C++ environments allow
		4025	throwing exceptions through C libraries (most do).
		4026
		4027	=head2 C++ API
		4028
		4029	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		4030	you to use some convenience methods to start/stop watchers and also change
		4031	the callback model to a model using method callbacks on objects.
		4032
		4033	To use it,
		4034
		4035	#include <ev++.h>
		4036
		4037	This automatically includes F<ev.h> and puts all of its definitions (many
		4038	of them macros) into the global namespace. All C++ specific things are
		4039	put into the C<ev> namespace. It should support all the same embedding
		4040	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		4041
		4042	Care has been taken to keep the overhead low. The only data member the C++
		4043	classes add (compared to plain C-style watchers) is the event loop pointer
		4044	that the watcher is associated with (or no additional members at all if
		4045	you disable C<EV_MULTIPLICITY> when embedding libev).
		4046
		4047	Currently, functions, static and non-static member functions and classes
		4048	with C<operator ()> can be used as callbacks. Other types should be easy
		4049	to add as long as they only need one additional pointer for context. If
		4050	you need support for other types of functors please contact the author
		4051	(preferably after implementing it).
		4052
		4053	For all this to work, your C++ compiler either has to use the same calling
		4054	conventions as your C compiler (for static member functions), or you have
		4055	to embed libev and compile libev itself as C++.
		4056
		4057	Here is a list of things available in the C<ev> namespace:
		4058
		4059	=over 4
		4060
		4061	=item C<ev::READ>, C<ev::WRITE> etc.
		4062
		4063	These are just enum values with the same values as the C<EV_READ> etc.
		4064	macros from F<ev.h>.
		4065
		4066	=item C<ev::tstamp>, C<ev::now>
		4067
		4068	Aliases to the same types/functions as with the C<ev_> prefix.
		4069
		4070	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		4071
		4072	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		4073	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		4074	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		4075	defined by many implementations.
		4076
		4077	All of those classes have these methods:
		4078
		4079	=over 4
		4080
		4081	=item ev::TYPE::TYPE ()
		4082
		4083	=item ev::TYPE::TYPE (loop)
		4084
		4085	=item ev::TYPE::~TYPE
		4086
		4087	The constructor (optionally) takes an event loop to associate the watcher
		4088	with. If it is omitted, it will use C<EV_DEFAULT>.
		4089
		4090	The constructor calls C<ev_init> for you, which means you have to call the
		4091	C<set> method before starting it.
		4092
		4093	It will not set a callback, however: You have to call the templated C<set>
		4094	method to set a callback before you can start the watcher.
		4095
		4096	(The reason why you have to use a method is a limitation in C++ which does
		4097	not allow explicit template arguments for constructors).
		4098
		4099	The destructor automatically stops the watcher if it is active.
		4100
		4101	=item w->set<class, &class::method> (object *)
		4102
		4103	This method sets the callback method to call. The method has to have a
		4104	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		4105	first argument and the C<revents> as second. The object must be given as
		4106	parameter and is stored in the C<data> member of the watcher.
		4107
		4108	This method synthesizes efficient thunking code to call your method from
		4109	the C callback that libev requires. If your compiler can inline your
		4110	callback (i.e. it is visible to it at the place of the C<set> call and
		4111	your compiler is good :), then the method will be fully inlined into the
		4112	thunking function, making it as fast as a direct C callback.
		4113
		4114	Example: simple class declaration and watcher initialisation
		4115
		4116	struct myclass
		4117	{
		4118	void io_cb (ev::io &w, int revents) { }
		4119	}
		4120
		4121	myclass obj;
		4122	ev::io iow;
		4123	iow.set <myclass, &myclass::io_cb> (&obj);
		4124
		4125	=item w->set (object *)
		4126
		4127	This is a variation of a method callback - leaving out the method to call
		4128	will default the method to C<operator ()>, which makes it possible to use
		4129	functor objects without having to manually specify the C<operator ()> all
		4130	the time. Incidentally, you can then also leave out the template argument
		4131	list.
		4132
		4133	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		4134	int revents)>.
		4135
		4136	See the method-C<set> above for more details.
		4137
		4138	Example: use a functor object as callback.
		4139
		4140	struct myfunctor
		4141	{
		4142	void operator() (ev::io &w, int revents)
		4143	{
		4144	...
		4145	}
		4146	}
		4147
		4148	myfunctor f;
		4149
		4150	ev::io w;
		4151	w.set (&f);
		4152
		4153	=item w->set<function> (void *data = 0)
		4154
		4155	Also sets a callback, but uses a static method or plain function as
		4156	callback. The optional C<data> argument will be stored in the watcher's
		4157	C<data> member and is free for you to use.
		4158
		4159	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		4160
		4161	See the method-C<set> above for more details.
		4162
		4163	Example: Use a plain function as callback.
		4164
		4165	static void io_cb (ev::io &w, int revents) { }
		4166	iow.set <io_cb> ();
		4167
		4168	=item w->set (loop)
		4169
		4170	Associates a different C<struct ev_loop> with this watcher. You can only
		4171	do this when the watcher is inactive (and not pending either).
		4172
		4173	=item w->set ([arguments])
		4174
		4175	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
		4176	with the same arguments. Either this method or a suitable start method
		4177	must be called at least once. Unlike the C counterpart, an active watcher
		4178	gets automatically stopped and restarted when reconfiguring it with this
		4179	method.
		4180
		4181	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4182	clashing with the C<set (loop)> method.
		4183
		4184	=item w->start ()
		4185
		4186	Starts the watcher. Note that there is no C<loop> argument, as the
		4187	constructor already stores the event loop.
		4188
		4189	=item w->start ([arguments])
		4190
		4191	Instead of calling C<set> and C<start> methods separately, it is often
		4192	convenient to wrap them in one call. Uses the same type of arguments as
		4193	the configure C<set> method of the watcher.
		4194
		4195	=item w->stop ()
		4196
		4197	Stops the watcher if it is active. Again, no C<loop> argument.
		4198
		4199	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		4200
		4201	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		4202	C<ev_TYPE_again> function.
		4203
		4204	=item w->sweep () (C<ev::embed> only)
		4205
		4206	Invokes C<ev_embed_sweep>.
		4207
		4208	=item w->update () (C<ev::stat> only)
		4209
		4210	Invokes C<ev_stat_stat>.
		4211
		4212	=back
		4213
		4214	=back
		4215
		4216	Example: Define a class with two I/O and idle watchers, start the I/O
		4217	watchers in the constructor.
		4218
		4219	class myclass
		4220	{
		4221	ev::io io ; void io_cb (ev::io &w, int revents);
		4222	ev::io io2 ; void io2_cb (ev::io &w, int revents);
		4223	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		4224
		4225	myclass (int fd)
		4226	{
		4227	io .set <myclass, &myclass::io_cb > (this);
		4228	io2 .set <myclass, &myclass::io2_cb > (this);
		4229	idle.set <myclass, &myclass::idle_cb> (this);
		4230
		4231	io.set (fd, ev::WRITE); // configure the watcher
		4232	io.start (); // start it whenever convenient
		4233
		4234	io2.start (fd, ev::READ); // set + start in one call
		4235	}
		4236	};
		4237
		4238
		4239	=head1 OTHER LANGUAGE BINDINGS
		4240
		4241	Libev does not offer other language bindings itself, but bindings for a
		4242	number of languages exist in the form of third-party packages. If you know
		4243	any interesting language binding in addition to the ones listed here, drop
		4244	me a note.
		4245
		4246	=over 4
		4247
		4248	=item Perl
		4249
		4250	The EV module implements the full libev API and is actually used to test
		4251	libev. EV is developed together with libev. Apart from the EV core module,
		4252	there are additional modules that implement libev-compatible interfaces
		4253	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		4254	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		4255	and C<EV::Glib>).
		4256
		4257	It can be found and installed via CPAN, its homepage is at
		4258	L<http://software.schmorp.de/pkg/EV>.
		4259
		4260	=item Python
		4261
		4262	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		4263	seems to be quite complete and well-documented.
		4264
		4265	=item Ruby
		4266
		4267	Tony Arcieri has written a ruby extension that offers access to a subset
		4268	of the libev API and adds file handle abstractions, asynchronous DNS and
		4269	more on top of it. It can be found via gem servers. Its homepage is at
		4270	L<http://rev.rubyforge.org/>.
		4271
		4272	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		4273	makes rev work even on mingw.
		4274
		4275	=item Haskell
		4276
		4277	A haskell binding to libev is available at
		4278	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		4279
		4280	=item D
		4281
		4282	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		4283	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
		4284
		4285	=item Ocaml
		4286
		4287	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4288	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4289
		4290	=item Lua
		4291
		4292	Brian Maher has written a partial interface to libev for lua (at the
		4293	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4294	L<http://github.com/brimworks/lua-ev>.
		4295
		4296	=item Javascript
		4297
		4298	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4299
		4300	=item Others
		4301
		4302	There are others, and I stopped counting.
		4303
		4304	=back
		4305
		4306
		4307	=head1 MACRO MAGIC
		4308
		4309	Libev can be compiled with a variety of options, the most fundamental
		4310	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		4311	functions and callbacks have an initial C<struct ev_loop *> argument.
		4312
		4313	To make it easier to write programs that cope with either variant, the
		4314	following macros are defined:
		4315
		4316	=over 4
		4317
		4318	=item C<EV_A>, C<EV_A_>
		4319
		4320	This provides the loop I<argument> for functions, if one is required ("ev
		4321	loop argument"). The C<EV_A> form is used when this is the sole argument,
		4322	C<EV_A_> is used when other arguments are following. Example:
		4323
		4324	ev_unref (EV_A);
		4325	ev_timer_add (EV_A_ watcher);
		4326	ev_run (EV_A_ 0);
		4327
		4328	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		4329	which is often provided by the following macro.
		4330
		4331	=item C<EV_P>, C<EV_P_>
		4332
		4333	This provides the loop I<parameter> for functions, if one is required ("ev
		4334	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		4335	C<EV_P_> is used when other parameters are following. Example:
		4336
		4337	// this is how ev_unref is being declared
		4338	static void ev_unref (EV_P);
		4339
		4340	// this is how you can declare your typical callback
		4341	static void cb (EV_P_ ev_timer *w, int revents)
		4342
		4343	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		4344	suitable for use with C<EV_A>.
		4345
		4346	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		4347
		4348	Similar to the other two macros, this gives you the value of the default
		4349	loop, if multiple loops are supported ("ev loop default"). The default loop
		4350	will be initialised if it isn't already initialised.
		4351
		4352	For non-multiplicity builds, these macros do nothing, so you always have
		4353	to initialise the loop somewhere.
		4354
		4355	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		4356
		4357	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		4358	default loop has been initialised (C<UC> == unchecked). Their behaviour
		4359	is undefined when the default loop has not been initialised by a previous
		4360	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		4361
		4362	It is often prudent to use C<EV_DEFAULT> when initialising the first
		4363	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		4364
		4365	=back
		4366
		4367	Example: Declare and initialise a check watcher, utilising the above
		4368	macros so it will work regardless of whether multiple loops are supported
		4369	or not.
		4370
		4371	static void
		4372	check_cb (EV_P_ ev_timer *w, int revents)
		4373	{
		4374	ev_check_stop (EV_A_ w);
		4375	}
		4376
		4377	ev_check check;
		4378	ev_check_init (&check, check_cb);
		4379	ev_check_start (EV_DEFAULT_ &check);
		4380	ev_run (EV_DEFAULT_ 0);
		4381
		4382	=head1 EMBEDDING
		4383
		4384	Libev can (and often is) directly embedded into host
		4385	applications. Examples of applications that embed it include the Deliantra
		4386	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		4387	and rxvt-unicode.
		4388
		4389	The goal is to enable you to just copy the necessary files into your
		4390	source directory without having to change even a single line in them, so
		4391	you can easily upgrade by simply copying (or having a checked-out copy of
		4392	libev somewhere in your source tree).
		4393
		4394	=head2 FILESETS
		4395
		4396	Depending on what features you need you need to include one or more sets of files
		4397	in your application.
		4398
		4399	=head3 CORE EVENT LOOP
		4400
		4401	To include only the libev core (all the C<ev_*> functions), with manual
		4402	configuration (no autoconf):
		4403
		4404	#define EV_STANDALONE 1
		4405	#include "ev.c"
		4406
		4407	This will automatically include F<ev.h>, too, and should be done in a
		4408	single C source file only to provide the function implementations. To use
		4409	it, do the same for F<ev.h> in all files wishing to use this API (best
		4410	done by writing a wrapper around F<ev.h> that you can include instead and
		4411	where you can put other configuration options):
		4412
		4413	#define EV_STANDALONE 1
		4414	#include "ev.h"
		4415
		4416	Both header files and implementation files can be compiled with a C++
		4417	compiler (at least, that's a stated goal, and breakage will be treated
		4418	as a bug).
		4419
		4420	You need the following files in your source tree, or in a directory
		4421	in your include path (e.g. in libev/ when using -Ilibev):
		4422
		4423	ev.h
		4424	ev.c
		4425	ev_vars.h
		4426	ev_wrap.h
		4427
		4428	ev_win32.c required on win32 platforms only
		4429
		4430	ev_select.c only when select backend is enabled
		4431	ev_poll.c only when poll backend is enabled
		4432	ev_epoll.c only when the epoll backend is enabled
		4433	ev_kqueue.c only when the kqueue backend is enabled
		4434	ev_port.c only when the solaris port backend is enabled
		4435
		4436	F<ev.c> includes the backend files directly when enabled, so you only need
		4437	to compile this single file.
		4438
		4439	=head3 LIBEVENT COMPATIBILITY API
		4440
		4441	To include the libevent compatibility API, also include:
		4442
		4443	#include "event.c"
		4444
		4445	in the file including F<ev.c>, and:
		4446
		4447	#include "event.h"
		4448
		4449	in the files that want to use the libevent API. This also includes F<ev.h>.
		4450
		4451	You need the following additional files for this:
		4452
		4453	event.h
		4454	event.c
		4455
		4456	=head3 AUTOCONF SUPPORT
		4457
		4458	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		4459	whatever way you want, you can also C<m4_include([libev.m4])> in your
		4460	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		4461	include F<config.h> and configure itself accordingly.
		4462
		4463	For this of course you need the m4 file:
		4464
		4465	libev.m4
		4466
		4467	=head2 PREPROCESSOR SYMBOLS/MACROS
		4468
		4469	Libev can be configured via a variety of preprocessor symbols you have to
		4470	define before including (or compiling) any of its files. The default in
		4471	the absence of autoconf is documented for every option.
		4472
		4473	Symbols marked with "(h)" do not change the ABI, and can have different
		4474	values when compiling libev vs. including F<ev.h>, so it is permissible
		4475	to redefine them before including F<ev.h> without breaking compatibility
		4476	to a compiled library. All other symbols change the ABI, which means all
		4477	users of libev and the libev code itself must be compiled with compatible
		4478	settings.
		4479
		4480	=over 4
		4481
		4482	=item EV_COMPAT3 (h)
		4483
		4484	Backwards compatibility is a major concern for libev. This is why this
		4485	release of libev comes with wrappers for the functions and symbols that
		4486	have been renamed between libev version 3 and 4.
		4487
		4488	You can disable these wrappers (to test compatibility with future
		4489	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4490	sources. This has the additional advantage that you can drop the C<struct>
		4491	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4492	typedef in that case.
		4493
		4494	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4495	and in some even more future version the compatibility code will be
		4496	removed completely.
		4497
		4498	=item EV_STANDALONE (h)
		4499
		4500	Must always be C<1> if you do not use autoconf configuration, which
		4501	keeps libev from including F<config.h>, and it also defines dummy
		4502	implementations for some libevent functions (such as logging, which is not
		4503	supported). It will also not define any of the structs usually found in
		4504	F<event.h> that are not directly supported by the libev core alone.
		4505
		4506	In standalone mode, libev will still try to automatically deduce the
		4507	configuration, but has to be more conservative.
		4508
		4509	=item EV_USE_FLOOR
		4510
		4511	If defined to be C<1>, libev will use the C<floor ()> function for its
		4512	periodic reschedule calculations, otherwise libev will fall back on a
		4513	portable (slower) implementation. If you enable this, you usually have to
		4514	link against libm or something equivalent. Enabling this when the C<floor>
		4515	function is not available will fail, so the safe default is to not enable
		4516	this.
		4517
		4518	=item EV_USE_MONOTONIC
		4519
		4520	If defined to be C<1>, libev will try to detect the availability of the
		4521	monotonic clock option at both compile time and runtime. Otherwise no
		4522	use of the monotonic clock option will be attempted. If you enable this,
		4523	you usually have to link against librt or something similar. Enabling it
		4524	when the functionality isn't available is safe, though, although you have
		4525	to make sure you link against any libraries where the C<clock_gettime>
		4526	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		4527
		4528	=item EV_USE_REALTIME
		4529
		4530	If defined to be C<1>, libev will try to detect the availability of the
		4531	real-time clock option at compile time (and assume its availability
		4532	at runtime if successful). Otherwise no use of the real-time clock
		4533	option will be attempted. This effectively replaces C<gettimeofday>
		4534	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		4535	correctness. See the note about libraries in the description of
		4536	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4537	C<EV_USE_CLOCK_SYSCALL>.
		4538
		4539	=item EV_USE_CLOCK_SYSCALL
		4540
		4541	If defined to be C<1>, libev will try to use a direct syscall instead
		4542	of calling the system-provided C<clock_gettime> function. This option
		4543	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4544	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4545	programs needlessly. Using a direct syscall is slightly slower (in
		4546	theory), because no optimised vdso implementation can be used, but avoids
		4547	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4548	higher, as it simplifies linking (no need for C<-lrt>).
		4549
		4550	=item EV_USE_NANOSLEEP
		4551
		4552	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4553	and will use it for delays. Otherwise it will use C<select ()>.
		4554
		4555	=item EV_USE_EVENTFD
		4556
		4557	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4558	available and will probe for kernel support at runtime. This will improve
		4559	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4560	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4561	2.7 or newer, otherwise disabled.
		4562
		4563	=item EV_USE_SELECT
		4564
		4565	If undefined or defined to be C<1>, libev will compile in support for the
		4566	C<select>(2) backend. No attempt at auto-detection will be done: if no
		4567	other method takes over, select will be it. Otherwise the select backend
		4568	will not be compiled in.
		4569
		4570	=item EV_SELECT_USE_FD_SET
		4571
		4572	If defined to C<1>, then the select backend will use the system C<fd_set>
		4573	structure. This is useful if libev doesn't compile due to a missing
		4574	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		4575	on exotic systems. This usually limits the range of file descriptors to
		4576	some low limit such as 1024 or might have other limitations (winsocket
		4577	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		4578	configures the maximum size of the C<fd_set>.
		4579
		4580	=item EV_SELECT_IS_WINSOCKET
		4581
		4582	When defined to C<1>, the select backend will assume that
		4583	select/socket/connect etc. don't understand file descriptors but
		4584	wants osf handles on win32 (this is the case when the select to
		4585	be used is the winsock select). This means that it will call
		4586	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		4587	it is assumed that all these functions actually work on fds, even
		4588	on win32. Should not be defined on non-win32 platforms.
		4589
		4590	=item EV_FD_TO_WIN32_HANDLE(fd)
		4591
		4592	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4593	file descriptors to socket handles. When not defining this symbol (the
		4594	default), then libev will call C<_get_osfhandle>, which is usually
		4595	correct. In some cases, programs use their own file descriptor management,
		4596	in which case they can provide this function to map fds to socket handles.
		4597
		4598	=item EV_WIN32_HANDLE_TO_FD(handle)
		4599
		4600	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4601	using the standard C<_open_osfhandle> function. For programs implementing
		4602	their own fd to handle mapping, overwriting this function makes it easier
		4603	to do so. This can be done by defining this macro to an appropriate value.
		4604
		4605	=item EV_WIN32_CLOSE_FD(fd)
		4606
		4607	If programs implement their own fd to handle mapping on win32, then this
		4608	macro can be used to override the C<close> function, useful to unregister
		4609	file descriptors again. Note that the replacement function has to close
		4610	the underlying OS handle.
		4611
		4612	=item EV_USE_WSASOCKET
		4613
		4614	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4615	communication socket, which works better in some environments. Otherwise,
		4616	the normal C<socket> function will be used, which works better in other
		4617	environments.
		4618
		4619	=item EV_USE_POLL
		4620
		4621	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		4622	backend. Otherwise it will be enabled on non-win32 platforms. It
		4623	takes precedence over select.
		4624
		4625	=item EV_USE_EPOLL
		4626
		4627	If defined to be C<1>, libev will compile in support for the Linux
		4628	C<epoll>(7) backend. Its availability will be detected at runtime,
		4629	otherwise another method will be used as fallback. This is the preferred
		4630	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4631	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4632
		4633	=item EV_USE_KQUEUE
		4634
		4635	If defined to be C<1>, libev will compile in support for the BSD style
		4636	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		4637	otherwise another method will be used as fallback. This is the preferred
		4638	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		4639	supports some types of fds correctly (the only platform we found that
		4640	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		4641	not be used unless explicitly requested. The best way to use it is to find
		4642	out whether kqueue supports your type of fd properly and use an embedded
		4643	kqueue loop.
		4644
		4645	=item EV_USE_PORT
		4646
		4647	If defined to be C<1>, libev will compile in support for the Solaris
		4648	10 port style backend. Its availability will be detected at runtime,
		4649	otherwise another method will be used as fallback. This is the preferred
		4650	backend for Solaris 10 systems.
		4651
		4652	=item EV_USE_DEVPOLL
		4653
		4654	Reserved for future expansion, works like the USE symbols above.
		4655
		4656	=item EV_USE_INOTIFY
		4657
		4658	If defined to be C<1>, libev will compile in support for the Linux inotify
		4659	interface to speed up C<ev_stat> watchers. Its actual availability will
		4660	be detected at runtime. If undefined, it will be enabled if the headers
		4661	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4662
		4663	=item EV_NO_SMP
		4664
		4665	If defined to be C<1>, libev will assume that memory is always coherent
		4666	between threads, that is, threads can be used, but threads never run on
		4667	different cpus (or different cpu cores). This reduces dependencies
		4668	and makes libev faster.
		4669
		4670	=item EV_NO_THREADS
		4671
		4672	If defined to be C<1>, libev will assume that it will never be called from
		4673	different threads (that includes signal handlers), which is a stronger
		4674	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4675	libev faster.
		4676
		4677	=item EV_ATOMIC_T
		4678
		4679	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4680	access is atomic with respect to other threads or signal contexts. No
		4681	such type is easily found in the C language, so you can provide your own
		4682	type that you know is safe for your purposes. It is used both for signal
		4683	handler "locking" as well as for signal and thread safety in C<ev_async>
		4684	watchers.
		4685
		4686	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4687	(from F<signal.h>), which is usually good enough on most platforms.
		4688
		4689	=item EV_H (h)
		4690
		4691	The name of the F<ev.h> header file used to include it. The default if
		4692	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4693	used to virtually rename the F<ev.h> header file in case of conflicts.
		4694
		4695	=item EV_CONFIG_H (h)
		4696
		4697	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4698	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4699	C<EV_H>, above.
		4700
		4701	=item EV_EVENT_H (h)
		4702
		4703	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4704	of how the F<event.h> header can be found, the default is C<"event.h">.
		4705
		4706	=item EV_PROTOTYPES (h)
		4707
		4708	If defined to be C<0>, then F<ev.h> will not define any function
		4709	prototypes, but still define all the structs and other symbols. This is
		4710	occasionally useful if you want to provide your own wrapper functions
		4711	around libev functions.
		4712
		4713	=item EV_MULTIPLICITY
		4714
		4715	If undefined or defined to C<1>, then all event-loop-specific functions
		4716	will have the C<struct ev_loop *> as first argument, and you can create
		4717	additional independent event loops. Otherwise there will be no support
		4718	for multiple event loops and there is no first event loop pointer
		4719	argument. Instead, all functions act on the single default loop.
		4720
		4721	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4722	default loop when multiplicity is switched off - you always have to
		4723	initialise the loop manually in this case.
		4724
		4725	=item EV_MINPRI
		4726
		4727	=item EV_MAXPRI
		4728
		4729	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4730	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4731	provide for more priorities by overriding those symbols (usually defined
		4732	to be C<-2> and C<2>, respectively).
		4733
		4734	When doing priority-based operations, libev usually has to linearly search
		4735	all the priorities, so having many of them (hundreds) uses a lot of space
		4736	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4737	fine.
		4738
		4739	If your embedding application does not need any priorities, defining these
		4740	both to C<0> will save some memory and CPU.
		4741
		4742	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4743	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4744	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4745
		4746	If undefined or defined to be C<1> (and the platform supports it), then
		4747	the respective watcher type is supported. If defined to be C<0>, then it
		4748	is not. Disabling watcher types mainly saves code size.
		4749
		4750	=item EV_FEATURES
		4751
		4752	If you need to shave off some kilobytes of code at the expense of some
		4753	speed (but with the full API), you can define this symbol to request
		4754	certain subsets of functionality. The default is to enable all features
		4755	that can be enabled on the platform.
		4756
		4757	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4758	with some broad features you want) and then selectively re-enable
		4759	additional parts you want, for example if you want everything minimal,
		4760	but multiple event loop support, async and child watchers and the poll
		4761	backend, use this:
		4762
		4763	#define EV_FEATURES 0
		4764	#define EV_MULTIPLICITY 1
		4765	#define EV_USE_POLL 1
		4766	#define EV_CHILD_ENABLE 1
		4767	#define EV_ASYNC_ENABLE 1
		4768
		4769	The actual value is a bitset, it can be a combination of the following
		4770	values (by default, all of these are enabled):
		4771
		4772	=over 4
		4773
		4774	=item C<1> - faster/larger code
		4775
		4776	Use larger code to speed up some operations.
		4777
		4778	Currently this is used to override some inlining decisions (enlarging the
		4779	code size by roughly 30% on amd64).
		4780
		4781	When optimising for size, use of compiler flags such as C<-Os> with
		4782	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4783	assertions.
		4784
		4785	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4786	(e.g. gcc with C<-Os>).
		4787
		4788	=item C<2> - faster/larger data structures
		4789
		4790	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4791	hash table sizes and so on. This will usually further increase code size
		4792	and can additionally have an effect on the size of data structures at
		4793	runtime.
		4794
		4795	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4796	(e.g. gcc with C<-Os>).
		4797
		4798	=item C<4> - full API configuration
		4799
		4800	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4801	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4802
		4803	=item C<8> - full API
		4804
		4805	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4806	details on which parts of the API are still available without this
		4807	feature, and do not complain if this subset changes over time.
		4808
		4809	=item C<16> - enable all optional watcher types
		4810
		4811	Enables all optional watcher types. If you want to selectively enable
		4812	only some watcher types other than I/O and timers (e.g. prepare,
		4813	embed, async, child...) you can enable them manually by defining
		4814	C<EV_watchertype_ENABLE> to C<1> instead.
		4815
		4816	=item C<32> - enable all backends
		4817
		4818	This enables all backends - without this feature, you need to enable at
		4819	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4820
		4821	=item C<64> - enable OS-specific "helper" APIs
		4822
		4823	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4824	default.
		4825
		4826	=back
		4827
		4828	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4829	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4830	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4831	watchers, timers and monotonic clock support.
		4832
		4833	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4834	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4835	your program might be left out as well - a binary starting a timer and an
		4836	I/O watcher then might come out at only 5Kb.
		4837
		4838	=item EV_API_STATIC
		4839
		4840	If this symbol is defined (by default it is not), then all identifiers
		4841	will have static linkage. This means that libev will not export any
		4842	identifiers, and you cannot link against libev anymore. This can be useful
		4843	when you embed libev, only want to use libev functions in a single file,
		4844	and do not want its identifiers to be visible.
		4845
		4846	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4847	wants to use libev.
		4848
		4849	This option only works when libev is compiled with a C compiler, as C++
		4850	doesn't support the required declaration syntax.
		4851
		4852	=item EV_AVOID_STDIO
		4853
		4854	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4855	functions (printf, scanf, perror etc.). This will increase the code size
		4856	somewhat, but if your program doesn't otherwise depend on stdio and your
		4857	libc allows it, this avoids linking in the stdio library which is quite
		4858	big.
		4859
		4860	Note that error messages might become less precise when this option is
		4861	enabled.
		4862
		4863	=item EV_NSIG
		4864
		4865	The highest supported signal number, +1 (or, the number of
		4866	signals): Normally, libev tries to deduce the maximum number of signals
		4867	automatically, but sometimes this fails, in which case it can be
		4868	specified. Also, using a lower number than detected (C<32> should be
		4869	good for about any system in existence) can save some memory, as libev
		4870	statically allocates some 12-24 bytes per signal number.
		4871
		4872	=item EV_PID_HASHSIZE
		4873
		4874	C<ev_child> watchers use a small hash table to distribute workload by
		4875	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4876	usually more than enough. If you need to manage thousands of children you
		4877	might want to increase this value (I<must> be a power of two).
		4878
		4879	=item EV_INOTIFY_HASHSIZE
		4880
		4881	C<ev_stat> watchers use a small hash table to distribute workload by
		4882	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4883	disabled), usually more than enough. If you need to manage thousands of
		4884	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4885	power of two).
		4886
		4887	=item EV_USE_4HEAP
		4888
		4889	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4890	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4891	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4892	faster performance with many (thousands) of watchers.
		4893
		4894	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4895	will be C<0>.
		4896
		4897	=item EV_HEAP_CACHE_AT
		4898
		4899	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4900	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4901	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4902	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4903	but avoids random read accesses on heap changes. This improves performance
		4904	noticeably with many (hundreds) of watchers.
		4905
		4906	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4907	will be C<0>.
		4908
		4909	=item EV_VERIFY
		4910
		4911	Controls how much internal verification (see C<ev_verify ()>) will
		4912	be done: If set to C<0>, no internal verification code will be compiled
		4913	in. If set to C<1>, then verification code will be compiled in, but not
		4914	called. If set to C<2>, then the internal verification code will be
		4915	called once per loop, which can slow down libev. If set to C<3>, then the
		4916	verification code will be called very frequently, which will slow down
		4917	libev considerably.
		4918
		4919	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4920	will be C<0>.
		4921
		4922	=item EV_COMMON
		4923
		4924	By default, all watchers have a C<void *data> member. By redefining
		4925	this macro to something else you can include more and other types of
		4926	members. You have to define it each time you include one of the files,
		4927	though, and it must be identical each time.
		4928
		4929	For example, the perl EV module uses something like this:
		4930
		4931	#define EV_COMMON \
		4932	SV *self; /* contains this struct */ \
		4933	SV *cb_sv, *fh /* note no trailing ";" */
		4934
		4935	=item EV_CB_DECLARE (type)
		4936
		4937	=item EV_CB_INVOKE (watcher, revents)
		4938
		4939	=item ev_set_cb (ev, cb)
		4940
		4941	Can be used to change the callback member declaration in each watcher,
		4942	and the way callbacks are invoked and set. Must expand to a struct member
		4943	definition and a statement, respectively. See the F<ev.h> header file for
		4944	their default definitions. One possible use for overriding these is to
		4945	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4946	method calls instead of plain function calls in C++.
		4947
		4948	=back
		4949
		4950	=head2 EXPORTED API SYMBOLS
		4951
		4952	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4953	exported symbols, you can use the provided F<Symbol.*> files which list
		4954	all public symbols, one per line:
		4955
		4956	Symbols.ev for libev proper
		4957	Symbols.event for the libevent emulation
		4958
		4959	This can also be used to rename all public symbols to avoid clashes with
		4960	multiple versions of libev linked together (which is obviously bad in
		4961	itself, but sometimes it is inconvenient to avoid this).
		4962
		4963	A sed command like this will create wrapper C<#define>'s that you need to
		4964	include before including F<ev.h>:
		4965
		4966	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4967
		4968	This would create a file F<wrap.h> which essentially looks like this:
		4969
		4970	#define ev_backend myprefix_ev_backend
		4971	#define ev_check_start myprefix_ev_check_start
		4972	#define ev_check_stop myprefix_ev_check_stop
		4973	...
		4974
		4975	=head2 EXAMPLES
		4976
		4977	For a real-world example of a program the includes libev
		4978	verbatim, you can have a look at the EV perl module
		4979	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4980	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4981	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4982	will be compiled. It is pretty complex because it provides its own header
		4983	file.
		4984
		4985	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4986	that everybody includes and which overrides some configure choices:
		4987
		4988	#define EV_FEATURES 8
		4989	#define EV_USE_SELECT 1
		4990	#define EV_PREPARE_ENABLE 1
		4991	#define EV_IDLE_ENABLE 1
		4992	#define EV_SIGNAL_ENABLE 1
		4993	#define EV_CHILD_ENABLE 1
		4994	#define EV_USE_STDEXCEPT 0
		4995	#define EV_CONFIG_H <config.h>
		4996
		4997	#include "ev++.h"
		4998
		4999	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		5000
		5001	#include "ev_cpp.h"
		5002	#include "ev.c"
		5003
		5004	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
		5005
		5006	=head2 THREADS AND COROUTINES
		5007
		5008	=head3 THREADS
		5009
		5010	All libev functions are reentrant and thread-safe unless explicitly
		5011	documented otherwise, but libev implements no locking itself. This means
		5012	that you can use as many loops as you want in parallel, as long as there
		5013	are no concurrent calls into any libev function with the same loop
		5014	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		5015	of course): libev guarantees that different event loops share no data
		5016	structures that need any locking.
		5017
		5018	Or to put it differently: calls with different loop parameters can be done
		5019	concurrently from multiple threads, calls with the same loop parameter
		5020	must be done serially (but can be done from different threads, as long as
		5021	only one thread ever is inside a call at any point in time, e.g. by using
		5022	a mutex per loop).
		5023
		5024	Specifically to support threads (and signal handlers), libev implements
		5025	so-called C<ev_async> watchers, which allow some limited form of
		5026	concurrency on the same event loop, namely waking it up "from the
		5027	outside".
		5028
		5029	If you want to know which design (one loop, locking, or multiple loops
		5030	without or something else still) is best for your problem, then I cannot
		5031	help you, but here is some generic advice:
		5032
		5033	=over 4
		5034
		5035	=item * most applications have a main thread: use the default libev loop
		5036	in that thread, or create a separate thread running only the default loop.
		5037
		5038	This helps integrating other libraries or software modules that use libev
		5039	themselves and don't care/know about threading.
		5040
		5041	=item * one loop per thread is usually a good model.
		5042
		5043	Doing this is almost never wrong, sometimes a better-performance model
		5044	exists, but it is always a good start.
		5045
		5046	=item * other models exist, such as the leader/follower pattern, where one
		5047	loop is handed through multiple threads in a kind of round-robin fashion.
		5048
		5049	Choosing a model is hard - look around, learn, know that usually you can do
		5050	better than you currently do :-)
		5051
		5052	=item * often you need to talk to some other thread which blocks in the
		5053	event loop.
		5054
		5055	C<ev_async> watchers can be used to wake them up from other threads safely
		5056	(or from signal contexts...).
		5057
		5058	An example use would be to communicate signals or other events that only
		5059	work in the default loop by registering the signal watcher with the
		5060	default loop and triggering an C<ev_async> watcher from the default loop
		5061	watcher callback into the event loop interested in the signal.
		5062
		5063	=back
		5064
		5065	See also L</THREAD LOCKING EXAMPLE>.
		5066
		5067	=head3 COROUTINES
		5068
		5069	Libev is very accommodating to coroutines ("cooperative threads"):
		5070	libev fully supports nesting calls to its functions from different
		5071	coroutines (e.g. you can call C<ev_run> on the same loop from two
		5072	different coroutines, and switch freely between both coroutines running
		5073	the loop, as long as you don't confuse yourself). The only exception is
		5074	that you must not do this from C<ev_periodic> reschedule callbacks.
		5075
		5076	Care has been taken to ensure that libev does not keep local state inside
		5077	C<ev_run>, and other calls do not usually allow for coroutine switches as
		5078	they do not call any callbacks.
		5079
		5080	=head2 COMPILER WARNINGS
		5081
		5082	Depending on your compiler and compiler settings, you might get no or a
		5083	lot of warnings when compiling libev code. Some people are apparently
		5084	scared by this.
		5085
		5086	However, these are unavoidable for many reasons. For one, each compiler
		5087	has different warnings, and each user has different tastes regarding
		5088	warning options. "Warn-free" code therefore cannot be a goal except when
		5089	targeting a specific compiler and compiler-version.
		5090
		5091	Another reason is that some compiler warnings require elaborate
		5092	workarounds, or other changes to the code that make it less clear and less
		5093	maintainable.
		5094
		5095	And of course, some compiler warnings are just plain stupid, or simply
		5096	wrong (because they don't actually warn about the condition their message
		5097	seems to warn about). For example, certain older gcc versions had some
		5098	warnings that resulted in an extreme number of false positives. These have
		5099	been fixed, but some people still insist on making code warn-free with
		5100	such buggy versions.
		5101
		5102	While libev is written to generate as few warnings as possible,
		5103	"warn-free" code is not a goal, and it is recommended not to build libev
		5104	with any compiler warnings enabled unless you are prepared to cope with
		5105	them (e.g. by ignoring them). Remember that warnings are just that:
		5106	warnings, not errors, or proof of bugs.
		5107
		5108
		5109	=head2 VALGRIND
		5110
		5111	Valgrind has a special section here because it is a popular tool that is
		5112	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		5113
		5114	If you think you found a bug (memory leak, uninitialised data access etc.)
		5115	in libev, then check twice: If valgrind reports something like:
		5116
		5117	==2274== definitely lost: 0 bytes in 0 blocks.
		5118	==2274== possibly lost: 0 bytes in 0 blocks.
		5119	==2274== still reachable: 256 bytes in 1 blocks.
		5120
		5121	Then there is no memory leak, just as memory accounted to global variables
		5122	is not a memleak - the memory is still being referenced, and didn't leak.
		5123
		5124	Similarly, under some circumstances, valgrind might report kernel bugs
		5125	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		5126	although an acceptable workaround has been found here), or it might be
		5127	confused.
		5128
		5129	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		5130	make it into some kind of religion.
		5131
		5132	If you are unsure about something, feel free to contact the mailing list
		5133	with the full valgrind report and an explanation on why you think this
		5134	is a bug in libev (best check the archives, too :). However, don't be
		5135	annoyed when you get a brisk "this is no bug" answer and take the chance
		5136	of learning how to interpret valgrind properly.
		5137
		5138	If you need, for some reason, empty reports from valgrind for your project
		5139	I suggest using suppression lists.
		5140
		5141
		5142	=head1 PORTABILITY NOTES
		5143
		5144	=head2 GNU/LINUX 32 BIT LIMITATIONS
		5145
		5146	GNU/Linux is the only common platform that supports 64 bit file/large file
		5147	interfaces but I<disables> them by default.
		5148
		5149	That means that libev compiled in the default environment doesn't support
		5150	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		5151
		5152	Unfortunately, many programs try to work around this GNU/Linux issue
		5153	by enabling the large file API, which makes them incompatible with the
		5154	standard libev compiled for their system.
		5155
		5156	Likewise, libev cannot enable the large file API itself as this would
		5157	suddenly make it incompatible to the default compile time environment,
		5158	i.e. all programs not using special compile switches.
		5159
		5160	=head2 OS/X AND DARWIN BUGS
		5161
		5162	The whole thing is a bug if you ask me - basically any system interface
		5163	you touch is broken, whether it is locales, poll, kqueue or even the
		5164	OpenGL drivers.
		5165
		5166	=head3 C<kqueue> is buggy
		5167
		5168	The kqueue syscall is broken in all known versions - most versions support
		5169	only sockets, many support pipes.
		5170
		5171	Libev tries to work around this by not using C<kqueue> by default on this
		5172	rotten platform, but of course you can still ask for it when creating a
		5173	loop - embedding a socket-only kqueue loop into a select-based one is
		5174	probably going to work well.
		5175
		5176	=head3 C<poll> is buggy
		5177
		5178	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		5179	implementation by something calling C<kqueue> internally around the 10.5.6
		5180	release, so now C<kqueue> I<and> C<poll> are broken.
		5181
		5182	Libev tries to work around this by not using C<poll> by default on
		5183	this rotten platform, but of course you can still ask for it when creating
		5184	a loop.
		5185
		5186	=head3 C<select> is buggy
		5187
		5188	All that's left is C<select>, and of course Apple found a way to fuck this
		5189	one up as well: On OS/X, C<select> actively limits the number of file
		5190	descriptors you can pass in to 1024 - your program suddenly crashes when
		5191	you use more.
		5192
		5193	There is an undocumented "workaround" for this - defining
		5194	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		5195	work on OS/X.
		5196
		5197	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		5198
		5199	=head3 C<errno> reentrancy
		5200
		5201	The default compile environment on Solaris is unfortunately so
		5202	thread-unsafe that you can't even use components/libraries compiled
		5203	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		5204	defined by default. A valid, if stupid, implementation choice.
		5205
		5206	If you want to use libev in threaded environments you have to make sure
		5207	it's compiled with C<_REENTRANT> defined.
		5208
		5209	=head3 Event port backend
		5210
		5211	The scalable event interface for Solaris is called "event
		5212	ports". Unfortunately, this mechanism is very buggy in all major
		5213	releases. If you run into high CPU usage, your program freezes or you get
		5214	a large number of spurious wakeups, make sure you have all the relevant
		5215	and latest kernel patches applied. No, I don't know which ones, but there
		5216	are multiple ones to apply, and afterwards, event ports actually work
		5217	great.
		5218
		5219	If you can't get it to work, you can try running the program by setting
		5220	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		5221	C<select> backends.
		5222
		5223	=head2 AIX POLL BUG
		5224
		5225	AIX unfortunately has a broken C<poll.h> header. Libev works around
		5226	this by trying to avoid the poll backend altogether (i.e. it's not even
		5227	compiled in), which normally isn't a big problem as C<select> works fine
		5228	with large bitsets on AIX, and AIX is dead anyway.
		5229
		5230	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		5231
		5232	=head3 General issues
		5233
		5234	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		5235	requires, and its I/O model is fundamentally incompatible with the POSIX
		5236	model. Libev still offers limited functionality on this platform in
		5237	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		5238	descriptors. This only applies when using Win32 natively, not when using
		5239	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		5240	as every compiler comes with a slightly differently broken/incompatible
		5241	environment.
		5242
		5243	Lifting these limitations would basically require the full
		5244	re-implementation of the I/O system. If you are into this kind of thing,
		5245	then note that glib does exactly that for you in a very portable way (note
		5246	also that glib is the slowest event library known to man).
		5247
		5248	There is no supported compilation method available on windows except
		5249	embedding it into other applications.
		5250
		5251	Sensible signal handling is officially unsupported by Microsoft - libev
		5252	tries its best, but under most conditions, signals will simply not work.
		5253
		5254	Not a libev limitation but worth mentioning: windows apparently doesn't
		5255	accept large writes: instead of resulting in a partial write, windows will
		5256	either accept everything or return C<ENOBUFS> if the buffer is too large,
		5257	so make sure you only write small amounts into your sockets (less than a
		5258	megabyte seems safe, but this apparently depends on the amount of memory
		5259	available).
		5260
		5261	Due to the many, low, and arbitrary limits on the win32 platform and
		5262	the abysmal performance of winsockets, using a large number of sockets
		5263	is not recommended (and not reasonable). If your program needs to use
		5264	more than a hundred or so sockets, then likely it needs to use a totally
		5265	different implementation for windows, as libev offers the POSIX readiness
		5266	notification model, which cannot be implemented efficiently on windows
		5267	(due to Microsoft monopoly games).
		5268
		5269	A typical way to use libev under windows is to embed it (see the embedding
		5270	section for details) and use the following F<evwrap.h> header file instead
		5271	of F<ev.h>:
		5272
		5273	#define EV_STANDALONE /* keeps ev from requiring config.h */
		5274	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		5275
		5276	#include "ev.h"
		5277
		5278	And compile the following F<evwrap.c> file into your project (make sure
		5279	you do I<not> compile the F<ev.c> or any other embedded source files!):
		5280
		5281	#include "evwrap.h"
		5282	#include "ev.c"
		5283
		5284	=head3 The winsocket C<select> function
		5285
		5286	The winsocket C<select> function doesn't follow POSIX in that it
		5287	requires socket I<handles> and not socket I<file descriptors> (it is
		5288	also extremely buggy). This makes select very inefficient, and also
		5289	requires a mapping from file descriptors to socket handles (the Microsoft
		5290	C runtime provides the function C<_open_osfhandle> for this). See the
		5291	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		5292	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		5293
		5294	The configuration for a "naked" win32 using the Microsoft runtime
		5295	libraries and raw winsocket select is:
		5296
		5297	#define EV_USE_SELECT 1
		5298	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		5299
		5300	Note that winsockets handling of fd sets is O(n), so you can easily get a
		5301	complexity in the O(n²) range when using win32.
		5302
		5303	=head3 Limited number of file descriptors
		5304
		5305	Windows has numerous arbitrary (and low) limits on things.
		5306
		5307	Early versions of winsocket's select only supported waiting for a maximum
		5308	of C<64> handles (probably owning to the fact that all windows kernels
		5309	can only wait for C<64> things at the same time internally; Microsoft
		5310	recommends spawning a chain of threads and wait for 63 handles and the
		5311	previous thread in each. Sounds great!).
		5312
		5313	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		5314	to some high number (e.g. C<2048>) before compiling the winsocket select
		5315	call (which might be in libev or elsewhere, for example, perl and many
		5316	other interpreters do their own select emulation on windows).
		5317
		5318	Another limit is the number of file descriptors in the Microsoft runtime
		5319	libraries, which by default is C<64> (there must be a hidden I<64>
		5320	fetish or something like this inside Microsoft). You can increase this
		5321	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		5322	(another arbitrary limit), but is broken in many versions of the Microsoft
		5323	runtime libraries. This might get you to about C<512> or C<2048> sockets
		5324	(depending on windows version and/or the phase of the moon). To get more,
		5325	you need to wrap all I/O functions and provide your own fd management, but
		5326	the cost of calling select (O(n²)) will likely make this unworkable.
		5327
		5328	=head2 PORTABILITY REQUIREMENTS
		5329
		5330	In addition to a working ISO-C implementation and of course the
		5331	backend-specific APIs, libev relies on a few additional extensions:
		5332
		5333	=over 4
		5334
		5335	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		5336	calling conventions regardless of C<ev_watcher_type *>.
		5337
		5338	Libev assumes not only that all watcher pointers have the same internal
		5339	structure (guaranteed by POSIX but not by ISO C for example), but it also
		5340	assumes that the same (machine) code can be used to call any watcher
		5341	callback: The watcher callbacks have different type signatures, but libev
		5342	calls them using an C<ev_watcher *> internally.
		5343
		5344	=item null pointers and integer zero are represented by 0 bytes
		5345
		5346	Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
		5347	relies on this setting pointers and integers to null.
		5348
		5349	=item pointer accesses must be thread-atomic
		5350
		5351	Accessing a pointer value must be atomic, it must both be readable and
		5352	writable in one piece - this is the case on all current architectures.
		5353
		5354	=item C<sig_atomic_t volatile> must be thread-atomic as well
		5355
		5356	The type C<sig_atomic_t volatile> (or whatever is defined as
		5357	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		5358	threads. This is not part of the specification for C<sig_atomic_t>, but is
		5359	believed to be sufficiently portable.
		5360
		5361	=item C<sigprocmask> must work in a threaded environment
		5362
		5363	Libev uses C<sigprocmask> to temporarily block signals. This is not
		5364	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		5365	pthread implementations will either allow C<sigprocmask> in the "main
		5366	thread" or will block signals process-wide, both behaviours would
		5367	be compatible with libev. Interaction between C<sigprocmask> and
		5368	C<pthread_sigmask> could complicate things, however.
		5369
		5370	The most portable way to handle signals is to block signals in all threads
		5371	except the initial one, and run the signal handling loop in the initial
		5372	thread as well.
		5373
		5374	=item C<long> must be large enough for common memory allocation sizes
		5375
		5376	To improve portability and simplify its API, libev uses C<long> internally
		5377	instead of C<size_t> when allocating its data structures. On non-POSIX
		5378	systems (Microsoft...) this might be unexpectedly low, but is still at
		5379	least 31 bits everywhere, which is enough for hundreds of millions of
		5380	watchers.
		5381
		5382	=item C<double> must hold a time value in seconds with enough accuracy
		5383
		5384	The type C<double> is used to represent timestamps. It is required to
		5385	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		5386	good enough for at least into the year 4000 with millisecond accuracy
		5387	(the design goal for libev). This requirement is overfulfilled by
		5388	implementations using IEEE 754, which is basically all existing ones.
		5389
		5390	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5391	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5392	is either obsolete or somebody patched it to use C<long double> or
		5393	something like that, just kidding).
		5394
		5395	=back
		5396
		5397	If you know of other additional requirements drop me a note.
		5398
		5399
		5400	=head1 ALGORITHMIC COMPLEXITIES
		5401
		5402	In this section the complexities of (many of) the algorithms used inside
		5403	libev will be documented. For complexity discussions about backends see
		5404	the documentation for C<ev_default_init>.
		5405
		5406	All of the following are about amortised time: If an array needs to be
		5407	extended, libev needs to realloc and move the whole array, but this
		5408	happens asymptotically rarer with higher number of elements, so O(1) might
		5409	mean that libev does a lengthy realloc operation in rare cases, but on
		5410	average it is much faster and asymptotically approaches constant time.
		5411
		5412	=over 4
		5413
		5414	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		5415
		5416	This means that, when you have a watcher that triggers in one hour and
		5417	there are 100 watchers that would trigger before that, then inserting will
		5418	have to skip roughly seven (C<ld 100>) of these watchers.
		5419
		5420	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		5421
		5422	That means that changing a timer costs less than removing/adding them,
		5423	as only the relative motion in the event queue has to be paid for.
		5424
		5425	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		5426
		5427	These just add the watcher into an array or at the head of a list.
		5428
		5429	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		5430
		5431	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		5432
		5433	These watchers are stored in lists, so they need to be walked to find the
		5434	correct watcher to remove. The lists are usually short (you don't usually
		5435	have many watchers waiting for the same fd or signal: one is typical, two
		5436	is rare).
		5437
		5438	=item Finding the next timer in each loop iteration: O(1)
		5439
		5440	By virtue of using a binary or 4-heap, the next timer is always found at a
		5441	fixed position in the storage array.
		5442
		5443	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5444
		5445	A change means an I/O watcher gets started or stopped, which requires
		5446	libev to recalculate its status (and possibly tell the kernel, depending
		5447	on backend and whether C<ev_io_set> was used).
		5448
		5449	=item Activating one watcher (putting it into the pending state): O(1)
		5450
		5451	=item Priority handling: O(number_of_priorities)
		5452
		5453	Priorities are implemented by allocating some space for each
		5454	priority. When doing priority-based operations, libev usually has to
		5455	linearly search all the priorities, but starting/stopping and activating
		5456	watchers becomes O(1) with respect to priority handling.
		5457
		5458	=item Sending an ev_async: O(1)
		5459
		5460	=item Processing ev_async_send: O(number_of_async_watchers)
		5461
		5462	=item Processing signals: O(max_signal_number)
		5463
		5464	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5465	calls in the current loop iteration and the loop is currently
		5466	blocked. Checking for async and signal events involves iterating over all
		5467	running async watchers or all signal numbers.
		5468
		5469	=back
		5470
		5471
		5472	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5473
		5474	The major version 4 introduced some incompatible changes to the API.
		5475
		5476	At the moment, the C<ev.h> header file provides compatibility definitions
		5477	for all changes, so most programs should still compile. The compatibility
		5478	layer might be removed in later versions of libev, so better update to the
		5479	new API early than late.
		5480
		5481	=over 4
		5482
		5483	=item C<EV_COMPAT3> backwards compatibility mechanism
		5484
		5485	The backward compatibility mechanism can be controlled by
		5486	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
		5487	section.
		5488
		5489	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5490
		5491	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5492
		5493	ev_loop_destroy (EV_DEFAULT_UC);
		5494	ev_loop_fork (EV_DEFAULT);
		5495
		5496	=item function/symbol renames
		5497
		5498	A number of functions and symbols have been renamed:
		5499
		5500	ev_loop => ev_run
		5501	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5502	EVLOOP_ONESHOT => EVRUN_ONCE
		5503
		5504	ev_unloop => ev_break
		5505	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5506	EVUNLOOP_ONE => EVBREAK_ONE
		5507	EVUNLOOP_ALL => EVBREAK_ALL
		5508
		5509	EV_TIMEOUT => EV_TIMER
		5510
		5511	ev_loop_count => ev_iteration
		5512	ev_loop_depth => ev_depth
		5513	ev_loop_verify => ev_verify
		5514
		5515	Most functions working on C<struct ev_loop> objects don't have an
		5516	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5517	associated constants have been renamed to not collide with the C<struct
		5518	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5519	as all other watcher types. Note that C<ev_loop_fork> is still called
		5520	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5521	typedef.
		5522
		5523	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5524
		5525	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5526	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5527	and work, but the library code will of course be larger.
		5528
		5529	=back
		5530
		5531
		5532	=head1 GLOSSARY
		5533
		5534	=over 4
		5535
		5536	=item active
		5537
		5538	A watcher is active as long as it has been started and not yet stopped.
		5539	See L</WATCHER STATES> for details.
		5540
		5541	=item application
		5542
		5543	In this document, an application is whatever is using libev.
		5544
		5545	=item backend
		5546
		5547	The part of the code dealing with the operating system interfaces.
		5548
		5549	=item callback
		5550
		5551	The address of a function that is called when some event has been
		5552	detected. Callbacks are being passed the event loop, the watcher that
		5553	received the event, and the actual event bitset.
		5554
		5555	=item callback/watcher invocation
		5556
		5557	The act of calling the callback associated with a watcher.
		5558
		5559	=item event
		5560
		5561	A change of state of some external event, such as data now being available
		5562	for reading on a file descriptor, time having passed or simply not having
		5563	any other events happening anymore.
		5564
		5565	In libev, events are represented as single bits (such as C<EV_READ> or
		5566	C<EV_TIMER>).
		5567
		5568	=item event library
		5569
		5570	A software package implementing an event model and loop.
		5571
		5572	=item event loop
		5573
		5574	An entity that handles and processes external events and converts them
		5575	into callback invocations.
		5576
		5577	=item event model
		5578
		5579	The model used to describe how an event loop handles and processes
		5580	watchers and events.
		5581
		5582	=item pending
		5583
		5584	A watcher is pending as soon as the corresponding event has been
		5585	detected. See L</WATCHER STATES> for details.
		5586
		5587	=item real time
		5588
		5589	The physical time that is observed. It is apparently strictly monotonic :)
		5590
		5591	=item wall-clock time
		5592
		5593	The time and date as shown on clocks. Unlike real time, it can actually
		5594	be wrong and jump forwards and backwards, e.g. when you adjust your
		5595	clock.
		5596
		5597	=item watcher
		5598
		5599	A data structure that describes interest in certain events. Watchers need
		5600	to be started (attached to an event loop) before they can receive events.
		5601
		5602	=back
		5603
742	=head1 AUTHOR	5604	=head1 AUTHOR
743		5605
744	Marc Lehmann <libev@schmorp.de>.	5606	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5607	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
745		5608

Diff Legend

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