ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.30 by root, Fri Nov 23 04:36:03 2007 UTC vs.
Revision 1.428 by root, Thu May 30 18:51:57 2013 UTC

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

Diff Legend

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