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 1.367 by root, Sun Feb 20 02:56:23 2011 UTC

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

Diff Legend

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