ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.8 by root, Mon Nov 12 08:20:02 2007 UTC vs.
Revision 1.433 by root, Fri May 2 07:05:42 2014 UTC

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

Diff Legend

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