ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.43 by root, Sat Nov 24 16:33:23 2007 UTC vs.
Revision 1.459 by root, Wed Jan 22 01:50:42 2020 UTC

		1	=encoding utf-8
		2
1	=head1 NAME	3	=head1 NAME
2		4
3	libev - a high performance full-featured event loop written in C	5	libev - a high performance full-featured event loop written in C
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
6		8
7	#include <ev.h>	9	#include <ev.h>
8		10
9	=head1 DESCRIPTION	11	=head2 EXAMPLE PROGRAM
		12
		13	// a single header file is required
		14	#include <ev.h>
		15
		16	#include <stdio.h> // for puts
		17
		18	// every watcher type has its own typedef'd struct
		19	// with the name ev_TYPE
		20	ev_io stdin_watcher;
		21	ev_timer timeout_watcher;
		22
		23	// all watcher callbacks have a similar signature
		24	// this callback is called when data is readable on stdin
		25	static void
		26	stdin_cb (EV_P_ ev_io *w, int revents)
		27	{
		28	puts ("stdin ready");
		29	// for one-shot events, one must manually stop the watcher
		30	// with its corresponding stop function.
		31	ev_io_stop (EV_A_ w);
		32
		33	// this causes all nested ev_run's to stop iterating
		34	ev_break (EV_A_ EVBREAK_ALL);
		35	}
		36
		37	// another callback, this time for a time-out
		38	static void
		39	timeout_cb (EV_P_ ev_timer *w, int revents)
		40	{
		41	puts ("timeout");
		42	// this causes the innermost ev_run to stop iterating
		43	ev_break (EV_A_ EVBREAK_ONE);
		44	}
		45
		46	int
		47	main (void)
		48	{
		49	// use the default event loop unless you have special needs
		50	struct ev_loop *loop = EV_DEFAULT;
		51
		52	// initialise an io watcher, then start it
		53	// this one will watch for stdin to become readable
		54	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		55	ev_io_start (loop, &stdin_watcher);
		56
		57	// initialise a timer watcher, then start it
		58	// simple non-repeating 5.5 second timeout
		59	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		60	ev_timer_start (loop, &timeout_watcher);
		61
		62	// now wait for events to arrive
		63	ev_run (loop, 0);
		64
		65	// break was called, so exit
		66	return 0;
		67	}
		68
		69	=head1 ABOUT THIS DOCUMENT
		70
		71	This document documents the libev software package.
		72
		73	The newest version of this document is also available as an html-formatted
		74	web page you might find easier to navigate when reading it for the first
		75	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		76
		77	While this document tries to be as complete as possible in documenting
		78	libev, its usage and the rationale behind its design, it is not a tutorial
		79	on event-based programming, nor will it introduce event-based programming
		80	with libev.
		81
		82	Familiarity with event based programming techniques in general is assumed
		83	throughout this document.
		84
		85	=head1 WHAT TO READ WHEN IN A HURRY
		86
		87	This manual tries to be very detailed, but unfortunately, this also makes
		88	it very long. If you just want to know the basics of libev, I suggest
		89	reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
		90	look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
		91	C<ev_timer> sections in L</WATCHER TYPES>.
		92
		93	=head1 ABOUT LIBEV
10		94
11	Libev is an event loop: you register interest in certain events (such as a	95	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	96	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	97	these event sources and provide your program with events.
14		98
15	To do this, it must take more or less complete control over your process	99	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	100	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	101	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	103	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	104	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	105	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	106	watcher.
23		107
24	=head1 FEATURES	108	=head2 FEATURES
25		109
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	110	Libev supports C<select>, C<poll>, the Linux-specific aio and C<epoll>
27	kqueue mechanisms for file descriptor events, relative timers, absolute	111	interfaces, the BSD-specific C<kqueue> and the Solaris-specific event port
28	timers with customised rescheduling, signal events, process status change	112	mechanisms for file descriptor events (C<ev_io>), the Linux C<inotify>
		113	interface (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		114	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		115	timers (C<ev_timer>), absolute timers with customised rescheduling
		116	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	117	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	118	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		119	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		120	limited support for fork events (C<ev_fork>).
		121
		122	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	123	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	124	for example).
33		125
34	=head1 CONVENTIONS	126	=head2 CONVENTIONS
35		127
36	Libev is very configurable. In this manual the default configuration	128	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	129	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	130	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	131	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	132	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	133	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	134	this argument.
43		135
44	=head1 TIME REPRESENTATION	136	=head2 TIME REPRESENTATION
45		137
46	Libev represents time as a single floating point number, representing the	138	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	139	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	140	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	141	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the C<double> type in C, and when you need to do any calculations on	142	too. It usually aliases to the C<double> type in C. When you need to do
51	it, you should treat it as such.	143	any calculations on it, you should treat it as some floating point value.
		144
		145	Unlike the name component C<stamp> might indicate, it is also used for
		146	time differences (e.g. delays) throughout libev.
		147
		148	=head1 ERROR HANDLING
		149
		150	Libev knows three classes of errors: operating system errors, usage errors
		151	and internal errors (bugs).
		152
		153	When libev catches an operating system error it cannot handle (for example
		154	a system call indicating a condition libev cannot fix), it calls the callback
		155	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		156	abort. The default is to print a diagnostic message and to call C<abort
		157	()>.
		158
		159	When libev detects a usage error such as a negative timer interval, then
		160	it will print a diagnostic message and abort (via the C<assert> mechanism,
		161	so C<NDEBUG> will disable this checking): these are programming errors in
		162	the libev caller and need to be fixed there.
		163
		164	Via the C<EV_FREQUENT> macro you can compile in and/or enable extensive
		165	consistency checking code inside libev that can be used to check for
		166	internal inconsistencies, suually caused by application bugs.
		167
		168	Libev also has a few internal error-checking C<assert>ions. These do not
		169	trigger under normal circumstances, as they indicate either a bug in libev
		170	or worse.
52		171
53		172
54	=head1 GLOBAL FUNCTIONS	173	=head1 GLOBAL FUNCTIONS
55		174
56	These functions can be called anytime, even before initialising the	175	These functions can be called anytime, even before initialising the
…		…
60		179
61	=item ev_tstamp ev_time ()	180	=item ev_tstamp ev_time ()
62		181
63	Returns the current time as libev would use it. Please note that the	182	Returns the current time as libev would use it. Please note that the
64	C<ev_now> function is usually faster and also often returns the timestamp	183	C<ev_now> function is usually faster and also often returns the timestamp
65	you actually want to know.	184	you actually want to know. Also interesting is the combination of
		185	C<ev_now_update> and C<ev_now>.
		186
		187	=item ev_sleep (ev_tstamp interval)
		188
		189	Sleep for the given interval: The current thread will be blocked
		190	until either it is interrupted or the given time interval has
		191	passed (approximately - it might return a bit earlier even if not
		192	interrupted). Returns immediately if C<< interval <= 0 >>.
		193
		194	Basically this is a sub-second-resolution C<sleep ()>.
		195
		196	The range of the C<interval> is limited - libev only guarantees to work
		197	with sleep times of up to one day (C<< interval <= 86400 >>).
66		198
67	=item int ev_version_major ()	199	=item int ev_version_major ()
68		200
69	=item int ev_version_minor ()	201	=item int ev_version_minor ()
70		202
71	You can find out the major and minor version numbers of the library	203	You can find out the major and minor ABI version numbers of the library
72	you linked against by calling the functions C<ev_version_major> and	204	you linked against by calling the functions C<ev_version_major> and
73	C<ev_version_minor>. If you want, you can compare against the global	205	C<ev_version_minor>. If you want, you can compare against the global
74	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	206	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
75	version of the library your program was compiled against.	207	version of the library your program was compiled against.
76		208
		209	These version numbers refer to the ABI version of the library, not the
		210	release version.
		211
77	Usually, it's a good idea to terminate if the major versions mismatch,	212	Usually, it's a good idea to terminate if the major versions mismatch,
78	as this indicates an incompatible change. Minor versions are usually	213	as this indicates an incompatible change. Minor versions are usually
79	compatible to older versions, so a larger minor version alone is usually	214	compatible to older versions, so a larger minor version alone is usually
80	not a problem.	215	not a problem.
81		216
82	Example: make sure we haven't accidentally been linked against the wrong	217	Example: Make sure we haven't accidentally been linked against the wrong
83	version:	218	version (note, however, that this will not detect other ABI mismatches,
		219	such as LFS or reentrancy).
84		220
85	assert (("libev version mismatch",	221	assert (("libev version mismatch",
86	ev_version_major () == EV_VERSION_MAJOR	222	ev_version_major () == EV_VERSION_MAJOR
87	&& ev_version_minor () >= EV_VERSION_MINOR));	223	&& ev_version_minor () >= EV_VERSION_MINOR));
88		224
89	=item unsigned int ev_supported_backends ()	225	=item unsigned int ev_supported_backends ()
90		226
91	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>	227	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
92	value) compiled into this binary of libev (independent of their	228	value) compiled into this binary of libev (independent of their
…		…
94	a description of the set values.	230	a description of the set values.
95		231
96	Example: make sure we have the epoll method, because yeah this is cool and	232	Example: make sure we have the epoll method, because yeah this is cool and
97	a must have and can we have a torrent of it please!!!11	233	a must have and can we have a torrent of it please!!!11
98		234
99	assert (("sorry, no epoll, no sex",	235	assert (("sorry, no epoll, no sex",
100	ev_supported_backends () & EVBACKEND_EPOLL));	236	ev_supported_backends () & EVBACKEND_EPOLL));
101		237
102	=item unsigned int ev_recommended_backends ()	238	=item unsigned int ev_recommended_backends ()
103		239
104	Return the set of all backends compiled into this binary of libev and also	240	Return the set of all backends compiled into this binary of libev and
105	recommended for this platform. This set is often smaller than the one	241	also recommended for this platform, meaning it will work for most file
		242	descriptor types. This set is often smaller than the one returned by
106	returned by C<ev_supported_backends>, as for example kqueue is broken on	243	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
107	most BSDs and will not be autodetected unless you explicitly request it	244	and will not be auto-detected unless you explicitly request it (assuming
108	(assuming you know what you are doing). This is the set of backends that	245	you know what you are doing). This is the set of backends that libev will
109	libev will probe for if you specify no backends explicitly.	246	probe for if you specify no backends explicitly.
110		247
111	=item unsigned int ev_embeddable_backends ()	248	=item unsigned int ev_embeddable_backends ()
112		249
113	Returns the set of backends that are embeddable in other event loops. This	250	Returns the set of backends that are embeddable in other event loops. This
114	is the theoretical, all-platform, value. To find which backends	251	value is platform-specific but can include backends not available on the
115	might be supported on the current system, you would need to look at	252	current system. To find which embeddable backends might be supported on
116	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for	253	the current system, you would need to look at C<ev_embeddable_backends ()
117	recommended ones.	254	& ev_supported_backends ()>, likewise for recommended ones.
118		255
119	See the description of C<ev_embed> watchers for more info.	256	See the description of C<ev_embed> watchers for more info.
120		257
121	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	258	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
122		259
123	Sets the allocation function to use (the prototype is similar to the	260	Sets the allocation function to use (the prototype is similar - the
124	realloc C function, the semantics are identical). It is used to allocate	261	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
125	and free memory (no surprises here). If it returns zero when memory	262	used to allocate and free memory (no surprises here). If it returns zero
126	needs to be allocated, the library might abort or take some potentially	263	when memory needs to be allocated (C<size != 0>), the library might abort
127	destructive action. The default is your system realloc function.	264	or take some potentially destructive action.
		265
		266	Since some systems (at least OpenBSD and Darwin) fail to implement
		267	correct C<realloc> semantics, libev will use a wrapper around the system
		268	C<realloc> and C<free> functions by default.
128		269
129	You could override this function in high-availability programs to, say,	270	You could override this function in high-availability programs to, say,
130	free some memory if it cannot allocate memory, to use a special allocator,	271	free some memory if it cannot allocate memory, to use a special allocator,
131	or even to sleep a while and retry until some memory is available.	272	or even to sleep a while and retry until some memory is available.
132		273
133	Example: replace the libev allocator with one that waits a bit and then	274	Example: The following is the C<realloc> function that libev itself uses
134	retries: better than mine).	275	which should work with C<realloc> and C<free> functions of all kinds and
		276	is probably a good basis for your own implementation.
135		277
136	static void *	278	static void *
137	persistent_realloc (void *ptr, long size)	279	ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
138	{	280	{
		281	if (size)
		282	return realloc (ptr, size);
		283
		284	free (ptr);
		285	return 0;
		286	}
		287
		288	Example: Replace the libev allocator with one that waits a bit and then
		289	retries.
		290
		291	static void *
		292	persistent_realloc (void *ptr, size_t size)
		293	{
		294	if (!size)
		295	{
		296	free (ptr);
		297	return 0;
		298	}
		299
139	for (;;)	300	for (;;)
140	{	301	{
141	void *newptr = realloc (ptr, size);	302	void *newptr = realloc (ptr, size);
142		303
143	if (newptr)	304	if (newptr)
…		…
148	}	309	}
149		310
150	...	311	...
151	ev_set_allocator (persistent_realloc);	312	ev_set_allocator (persistent_realloc);
152		313
153	=item ev_set_syserr_cb (void (*cb)(const char *msg));	314	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
154		315
155	Set the callback function to call on a retryable syscall error (such	316	Set the callback function to call on a retryable system call error (such
156	as failed select, poll, epoll_wait). The message is a printable string	317	as failed select, poll, epoll_wait). The message is a printable string
157	indicating the system call or subsystem causing the problem. If this	318	indicating the system call or subsystem causing the problem. If this
158	callback is set, then libev will expect it to remedy the sitution, no	319	callback is set, then libev will expect it to remedy the situation, no
159	matter what, when it returns. That is, libev will generally retry the	320	matter what, when it returns. That is, libev will generally retry the
160	requested operation, or, if the condition doesn't go away, do bad stuff	321	requested operation, or, if the condition doesn't go away, do bad stuff
161	(such as abort).	322	(such as abort).
162		323
163	Example: do the same thing as libev does internally:	324	Example: This is basically the same thing that libev does internally, too.
164		325
165	static void	326	static void
166	fatal_error (const char *msg)	327	fatal_error (const char *msg)
167	{	328	{
168	perror (msg);	329	perror (msg);
…		…
170	}	331	}
171		332
172	...	333	...
173	ev_set_syserr_cb (fatal_error);	334	ev_set_syserr_cb (fatal_error);
174		335
		336	=item ev_feed_signal (int signum)
		337
		338	This function can be used to "simulate" a signal receive. It is completely
		339	safe to call this function at any time, from any context, including signal
		340	handlers or random threads.
		341
		342	Its main use is to customise signal handling in your process, especially
		343	in the presence of threads. For example, you could block signals
		344	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		345	creating any loops), and in one thread, use C<sigwait> or any other
		346	mechanism to wait for signals, then "deliver" them to libev by calling
		347	C<ev_feed_signal>.
		348
175	=back	349	=back
176		350
177	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	351	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
178		352
179	An event loop is described by a C<struct ev_loop *>. The library knows two	353	An event loop is described by a C<struct ev_loop *> (the C<struct> is
180	types of such loops, the I<default> loop, which supports signals and child	354	I<not> optional in this case unless libev 3 compatibility is disabled, as
181	events, and dynamically created loops which do not.	355	libev 3 had an C<ev_loop> function colliding with the struct name).
182		356
183	If you use threads, a common model is to run the default event loop	357	The library knows two types of such loops, the I<default> loop, which
184	in your main thread (or in a separate thread) and for each thread you	358	supports child process events, and dynamically created event loops which
185	create, you also create another event loop. Libev itself does no locking	359	do not.
186	whatsoever, so if you mix calls to the same event loop in different
187	threads, make sure you lock (this is usually a bad idea, though, even if
188	done correctly, because it's hideous and inefficient).
189		360
190	=over 4	361	=over 4
191		362
192	=item struct ev_loop *ev_default_loop (unsigned int flags)	363	=item struct ev_loop *ev_default_loop (unsigned int flags)
193		364
194	This will initialise the default event loop if it hasn't been initialised	365	This returns the "default" event loop object, which is what you should
195	yet and return it. If the default loop could not be initialised, returns	366	normally use when you just need "the event loop". Event loop objects and
196	false. If it already was initialised it simply returns it (and ignores the	367	the C<flags> parameter are described in more detail in the entry for
197	flags. If that is troubling you, check C<ev_backend ()> afterwards).	368	C<ev_loop_new>.
		369
		370	If the default loop is already initialised then this function simply
		371	returns it (and ignores the flags. If that is troubling you, check
		372	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		373	flags, which should almost always be C<0>, unless the caller is also the
		374	one calling C<ev_run> or otherwise qualifies as "the main program".
198		375
199	If you don't know what event loop to use, use the one returned from this	376	If you don't know what event loop to use, use the one returned from this
200	function.	377	function (or via the C<EV_DEFAULT> macro).
		378
		379	Note that this function is I<not> thread-safe, so if you want to use it
		380	from multiple threads, you have to employ some kind of mutex (note also
		381	that this case is unlikely, as loops cannot be shared easily between
		382	threads anyway).
		383
		384	The default loop is the only loop that can handle C<ev_child> watchers,
		385	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		386	a problem for your application you can either create a dynamic loop with
		387	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		388	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		389
		390	Example: This is the most typical usage.
		391
		392	if (!ev_default_loop (0))
		393	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		394
		395	Example: Restrict libev to the select and poll backends, and do not allow
		396	environment settings to be taken into account:
		397
		398	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		399
		400	=item struct ev_loop *ev_loop_new (unsigned int flags)
		401
		402	This will create and initialise a new event loop object. If the loop
		403	could not be initialised, returns false.
		404
		405	This function is thread-safe, and one common way to use libev with
		406	threads is indeed to create one loop per thread, and using the default
		407	loop in the "main" or "initial" thread.
201		408
202	The flags argument can be used to specify special behaviour or specific	409	The flags argument can be used to specify special behaviour or specific
203	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).	410	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
204		411
205	The following flags are supported:	412	The following flags are supported:
…		…
211	The default flags value. Use this if you have no clue (it's the right	418	The default flags value. Use this if you have no clue (it's the right
212	thing, believe me).	419	thing, believe me).
213		420
214	=item C<EVFLAG_NOENV>	421	=item C<EVFLAG_NOENV>
215		422
216	If this flag bit is ored into the flag value (or the program runs setuid	423	If this flag bit is or'ed into the flag value (or the program runs setuid
217	or setgid) then libev will I<not> look at the environment variable	424	or setgid) then libev will I<not> look at the environment variable
218	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	425	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
219	override the flags completely if it is found in the environment. This is	426	override the flags completely if it is found in the environment. This is
220	useful to try out specific backends to test their performance, or to work	427	useful to try out specific backends to test their performance, to work
221	around bugs.	428	around bugs, or to make libev threadsafe (accessing environment variables
		429	cannot be done in a threadsafe way, but usually it works if no other
		430	thread modifies them).
		431
		432	=item C<EVFLAG_FORKCHECK>
		433
		434	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		435	make libev check for a fork in each iteration by enabling this flag.
		436
		437	This works by calling C<getpid ()> on every iteration of the loop,
		438	and thus this might slow down your event loop if you do a lot of loop
		439	iterations and little real work, but is usually not noticeable (on my
		440	GNU/Linux system for example, C<getpid> is actually a simple 5-insn
		441	sequence without a system call and thus I<very> fast, but my GNU/Linux
		442	system also has C<pthread_atfork> which is even faster). (Update: glibc
		443	versions 2.25 apparently removed the C<getpid> optimisation again).
		444
		445	The big advantage of this flag is that you can forget about fork (and
		446	forget about forgetting to tell libev about forking, although you still
		447	have to ignore C<SIGPIPE>) when you use this flag.
		448
		449	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		450	environment variable.
		451
		452	=item C<EVFLAG_NOINOTIFY>
		453
		454	When this flag is specified, then libev will not attempt to use the
		455	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		456	testing, this flag can be useful to conserve inotify file descriptors, as
		457	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		458
		459	=item C<EVFLAG_SIGNALFD>
		460
		461	When this flag is specified, then libev will attempt to use the
		462	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		463	delivers signals synchronously, which makes it both faster and might make
		464	it possible to get the queued signal data. It can also simplify signal
		465	handling with threads, as long as you properly block signals in your
		466	threads that are not interested in handling them.
		467
		468	Signalfd will not be used by default as this changes your signal mask, and
		469	there are a lot of shoddy libraries and programs (glib's threadpool for
		470	example) that can't properly initialise their signal masks.
		471
		472	=item C<EVFLAG_NOSIGMASK>
		473
		474	When this flag is specified, then libev will avoid to modify the signal
		475	mask. Specifically, this means you have to make sure signals are unblocked
		476	when you want to receive them.
		477
		478	This behaviour is useful when you want to do your own signal handling, or
		479	want to handle signals only in specific threads and want to avoid libev
		480	unblocking the signals.
		481
		482	It's also required by POSIX in a threaded program, as libev calls
		483	C<sigprocmask>, whose behaviour is officially unspecified.
		484
		485	=item C<EVFLAG_NOTIMERFD>
		486
		487	When this flag is specified, the libev will avoid using a C<timerfd> to
		488	detect time jumps. It will still be able to detect time jumps, but takes
		489	longer and has a lower accuracy in doing so, but saves a file descriptor
		490	per loop.
		491
		492	The current implementation only tries to use a C<timerfd> when the first
		493	C<ev_periodic> watcher is started and falls back on other methods if it
		494	cannot be created, but this behaviour might change in the future.
222		495
223	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	496	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
224		497
225	This is your standard select(2) backend. Not I<completely> standard, as	498	This is your standard select(2) backend. Not I<completely> standard, as
226	libev tries to roll its own fd_set with no limits on the number of fds,	499	libev tries to roll its own fd_set with no limits on the number of fds,
227	but if that fails, expect a fairly low limit on the number of fds when	500	but if that fails, expect a fairly low limit on the number of fds when
228	using this backend. It doesn't scale too well (O(highest_fd)), but its usually	501	using this backend. It doesn't scale too well (O(highest_fd)), but its
229	the fastest backend for a low number of fds.	502	usually the fastest backend for a low number of (low-numbered :) fds.
		503
		504	To get good performance out of this backend you need a high amount of
		505	parallelism (most of the file descriptors should be busy). If you are
		506	writing a server, you should C<accept ()> in a loop to accept as many
		507	connections as possible during one iteration. You might also want to have
		508	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		509	readiness notifications you get per iteration.
		510
		511	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		512	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		513	C<exceptfds> set on that platform).
230		514
231	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	515	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
232		516
233	And this is your standard poll(2) backend. It's more complicated than	517	And this is your standard poll(2) backend. It's more complicated
234	select, but handles sparse fds better and has no artificial limit on the	518	than select, but handles sparse fds better and has no artificial
235	number of fds you can use (except it will slow down considerably with a	519	limit on the number of fds you can use (except it will slow down
236	lot of inactive fds). It scales similarly to select, i.e. O(total_fds).	520	considerably with a lot of inactive fds). It scales similarly to select,
		521	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		522	performance tips.
		523
		524	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		525	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
237		526
238	=item C<EVBACKEND_EPOLL> (value 4, Linux)	527	=item C<EVBACKEND_EPOLL> (value 4, Linux)
239		528
		529	Use the Linux-specific epoll(7) interface (for both pre- and post-2.6.9
		530	kernels).
		531
240	For few fds, this backend is a bit little slower than poll and select,	532	For few fds, this backend is a bit little slower than poll and select, but
241	but it scales phenomenally better. While poll and select usually scale like	533	it scales phenomenally better. While poll and select usually scale like
242	O(total_fds) where n is the total number of fds (or the highest fd), epoll scales	534	O(total_fds) where total_fds is the total number of fds (or the highest
243	either O(1) or O(active_fds).	535	fd), epoll scales either O(1) or O(active_fds).
244		536
		537	The epoll mechanism deserves honorable mention as the most misdesigned
		538	of the more advanced event mechanisms: mere annoyances include silently
		539	dropping file descriptors, requiring a system call per change per file
		540	descriptor (and unnecessary guessing of parameters), problems with dup,
		541	returning before the timeout value, resulting in additional iterations
		542	(and only giving 5ms accuracy while select on the same platform gives
		543	0.1ms) and so on. The biggest issue is fork races, however - if a program
		544	forks then I<both> parent and child process have to recreate the epoll
		545	set, which can take considerable time (one syscall per file descriptor)
		546	and is of course hard to detect.
		547
		548	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
		549	but of course I<doesn't>, and epoll just loves to report events for
		550	totally I<different> file descriptors (even already closed ones, so
		551	one cannot even remove them from the set) than registered in the set
		552	(especially on SMP systems). Libev tries to counter these spurious
		553	notifications by employing an additional generation counter and comparing
		554	that against the events to filter out spurious ones, recreating the set
		555	when required. Epoll also erroneously rounds down timeouts, but gives you
		556	no way to know when and by how much, so sometimes you have to busy-wait
		557	because epoll returns immediately despite a nonzero timeout. And last
		558	not least, it also refuses to work with some file descriptors which work
		559	perfectly fine with C<select> (files, many character devices...).
		560
		561	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		562	cobbled together in a hurry, no thought to design or interaction with
		563	others. Oh, the pain, will it ever stop...
		564
245	While stopping and starting an I/O watcher in the same iteration will	565	While stopping, setting and starting an I/O watcher in the same iteration
246	result in some caching, there is still a syscall per such incident	566	will result in some caching, there is still a system call per such
247	(because the fd could point to a different file description now), so its	567	incident (because the same I<file descriptor> could point to a different
248	best to avoid that. Also, dup()ed file descriptors might not work very	568	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
249	well if you register events for both fds.	569	file descriptors might not work very well if you register events for both
		570	file descriptors.
250		571
251	Please note that epoll sometimes generates spurious notifications, so you	572	Best performance from this backend is achieved by not unregistering all
252	need to use non-blocking I/O or other means to avoid blocking when no data	573	watchers for a file descriptor until it has been closed, if possible,
253	(or space) is available.	574	i.e. keep at least one watcher active per fd at all times. Stopping and
		575	starting a watcher (without re-setting it) also usually doesn't cause
		576	extra overhead. A fork can both result in spurious notifications as well
		577	as in libev having to destroy and recreate the epoll object, which can
		578	take considerable time and thus should be avoided.
		579
		580	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		581	faster than epoll for maybe up to a hundred file descriptors, depending on
		582	the usage. So sad.
		583
		584	While nominally embeddable in other event loops, this feature is broken in
		585	a lot of kernel revisions, but probably(!) works in current versions.
		586
		587	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		588	C<EVBACKEND_POLL>.
		589
		590	=item C<EVBACKEND_LINUXAIO> (value 64, Linux)
		591
		592	Use the Linux-specific Linux AIO (I<not> C<< aio(7) >> but C<<
		593	io_submit(2) >>) event interface available in post-4.18 kernels (but libev
		594	only tries to use it in 4.19+).
		595
		596	This is another Linux train wreck of an event interface.
		597
		598	If this backend works for you (as of this writing, it was very
		599	experimental), it is the best event interface available on Linux and might
		600	be well worth enabling it - if it isn't available in your kernel this will
		601	be detected and this backend will be skipped.
		602
		603	This backend can batch oneshot requests and supports a user-space ring
		604	buffer to receive events. It also doesn't suffer from most of the design
		605	problems of epoll (such as not being able to remove event sources from
		606	the epoll set), and generally sounds too good to be true. Because, this
		607	being the Linux kernel, of course it suffers from a whole new set of
		608	limitations, forcing you to fall back to epoll, inheriting all its design
		609	issues.
		610
		611	For one, it is not easily embeddable (but probably could be done using
		612	an event fd at some extra overhead). It also is subject to a system wide
		613	limit that can be configured in F</proc/sys/fs/aio-max-nr>. If no AIO
		614	requests are left, this backend will be skipped during initialisation, and
		615	will switch to epoll when the loop is active.
		616
		617	Most problematic in practice, however, is that not all file descriptors
		618	work with it. For example, in Linux 5.1, TCP sockets, pipes, event fds,
		619	files, F</dev/null> and many others are supported, but ttys do not work
		620	properly (a known bug that the kernel developers don't care about, see
		621	L<https://lore.kernel.org/patchwork/patch/1047453/>), so this is not
		622	(yet?) a generic event polling interface.
		623
		624	Overall, it seems the Linux developers just don't want it to have a
		625	generic event handling mechanism other than C<select> or C<poll>.
		626
		627	To work around all these problem, the current version of libev uses its
		628	epoll backend as a fallback for file descriptor types that do not work. Or
		629	falls back completely to epoll if the kernel acts up.
		630
		631	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		632	C<EVBACKEND_POLL>.
254		633
255	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	634	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
256		635
257	Kqueue deserves special mention, as at the time of this writing, it	636	Kqueue deserves special mention, as at the time this backend was
258	was broken on all BSDs except NetBSD (usually it doesn't work with	637	implemented, it was broken on all BSDs except NetBSD (usually it doesn't
259	anything but sockets and pipes, except on Darwin, where of course its	638	work reliably with anything but sockets and pipes, except on Darwin,
260	completely useless). For this reason its not being "autodetected"	639	where of course it's completely useless). Unlike epoll, however, whose
261	unless you explicitly specify it explicitly in the flags (i.e. using	640	brokenness is by design, these kqueue bugs can be (and mostly have been)
262	C<EVBACKEND_KQUEUE>).	641	fixed without API changes to existing programs. For this reason it's not
		642	being "auto-detected" on all platforms unless you explicitly specify it
		643	in the flags (i.e. using C<EVBACKEND_KQUEUE>) or libev was compiled on a
		644	known-to-be-good (-enough) system like NetBSD.
		645
		646	You still can embed kqueue into a normal poll or select backend and use it
		647	only for sockets (after having made sure that sockets work with kqueue on
		648	the target platform). See C<ev_embed> watchers for more info.
263		649
264	It scales in the same way as the epoll backend, but the interface to the	650	It scales in the same way as the epoll backend, but the interface to the
265	kernel is more efficient (which says nothing about its actual speed, of	651	kernel is more efficient (which says nothing about its actual speed, of
266	course). While starting and stopping an I/O watcher does not cause an	652	course). While stopping, setting and starting an I/O watcher does never
267	extra syscall as with epoll, it still adds up to four event changes per	653	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
268	incident, so its best to avoid that.	654	two event changes per incident. Support for C<fork ()> is very bad (you
		655	might have to leak fds on fork, but it's more sane than epoll) and it
		656	drops fds silently in similarly hard-to-detect cases.
		657
		658	This backend usually performs well under most conditions.
		659
		660	While nominally embeddable in other event loops, this doesn't work
		661	everywhere, so you might need to test for this. And since it is broken
		662	almost everywhere, you should only use it when you have a lot of sockets
		663	(for which it usually works), by embedding it into another event loop
		664	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		665	also broken on OS X)) and, did I mention it, using it only for sockets.
		666
		667	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		668	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		669	C<NOTE_EOF>.
269		670
270	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)	671	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
271		672
272	This is not implemented yet (and might never be).	673	This is not implemented yet (and might never be, unless you send me an
		674	implementation). According to reports, C</dev/poll> only supports sockets
		675	and is not embeddable, which would limit the usefulness of this backend
		676	immensely.
273		677
274	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	678	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
275		679
276	This uses the Solaris 10 port mechanism. As with everything on Solaris,	680	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
277	it's really slow, but it still scales very well (O(active_fds)).	681	it's really slow, but it still scales very well (O(active_fds)).
278		682
279	Please note that solaris ports can result in a lot of spurious	683	While this backend scales well, it requires one system call per active
280	notifications, so you need to use non-blocking I/O or other means to avoid	684	file descriptor per loop iteration. For small and medium numbers of file
281	blocking when no data (or space) is available.	685	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		686	might perform better.
		687
		688	On the positive side, this backend actually performed fully to
		689	specification in all tests and is fully embeddable, which is a rare feat
		690	among the OS-specific backends (I vastly prefer correctness over speed
		691	hacks).
		692
		693	On the negative side, the interface is I<bizarre> - so bizarre that
		694	even sun itself gets it wrong in their code examples: The event polling
		695	function sometimes returns events to the caller even though an error
		696	occurred, but with no indication whether it has done so or not (yes, it's
		697	even documented that way) - deadly for edge-triggered interfaces where you
		698	absolutely have to know whether an event occurred or not because you have
		699	to re-arm the watcher.
		700
		701	Fortunately libev seems to be able to work around these idiocies.
		702
		703	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		704	C<EVBACKEND_POLL>.
282		705
283	=item C<EVBACKEND_ALL>	706	=item C<EVBACKEND_ALL>
284		707
285	Try all backends (even potentially broken ones that wouldn't be tried	708	Try all backends (even potentially broken ones that wouldn't be tried
286	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	709	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
287	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	710	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
288		711
		712	It is definitely not recommended to use this flag, use whatever
		713	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		714	at all.
		715
		716	=item C<EVBACKEND_MASK>
		717
		718	Not a backend at all, but a mask to select all backend bits from a
		719	C<flags> value, in case you want to mask out any backends from a flags
		720	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
		721
289	=back	722	=back
290		723
291	If one or more of these are ored into the flags value, then only these	724	If one or more of the backend flags are or'ed into the flags value,
292	backends will be tried (in the reverse order as given here). If none are	725	then only these backends will be tried (in the reverse order as listed
293	specified, most compiled-in backend will be tried, usually in reverse	726	here). If none are specified, all backends in C<ev_recommended_backends
294	order of their flag values :)	727	()> will be tried.
295		728
296	The most typical usage is like this:
297
298	if (!ev_default_loop (0))
299	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
300
301	Restrict libev to the select and poll backends, and do not allow
302	environment settings to be taken into account:
303
304	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
305
306	Use whatever libev has to offer, but make sure that kqueue is used if
307	available (warning, breaks stuff, best use only with your own private
308	event loop and only if you know the OS supports your types of fds):
309
310	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
311
312	=item struct ev_loop *ev_loop_new (unsigned int flags)
313
314	Similar to C<ev_default_loop>, but always creates a new event loop that is
315	always distinct from the default loop. Unlike the default loop, it cannot
316	handle signal and child watchers, and attempts to do so will be greeted by
317	undefined behaviour (or a failed assertion if assertions are enabled).
318
319	Example: try to create a event loop that uses epoll and nothing else.	729	Example: Try to create a event loop that uses epoll and nothing else.
320		730
321	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);	731	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
322	if (!epoller)	732	if (!epoller)
323	fatal ("no epoll found here, maybe it hides under your chair");	733	fatal ("no epoll found here, maybe it hides under your chair");
324		734
		735	Example: Use whatever libev has to offer, but make sure that kqueue is
		736	used if available.
		737
		738	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		739
		740	Example: Similarly, on linux, you mgiht want to take advantage of the
		741	linux aio backend if possible, but fall back to something else if that
		742	isn't available.
		743
		744	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
		745
325	=item ev_default_destroy ()	746	=item ev_loop_destroy (loop)
326		747
327	Destroys the default loop again (frees all memory and kernel state	748	Destroys an event loop object (frees all memory and kernel state
328	etc.). None of the active event watchers will be stopped in the normal	749	etc.). None of the active event watchers will be stopped in the normal
329	sense, so e.g. C<ev_is_active> might still return true. It is your	750	sense, so e.g. C<ev_is_active> might still return true. It is your
330	responsibility to either stop all watchers cleanly yoursef I<before>	751	responsibility to either stop all watchers cleanly yourself I<before>
331	calling this function, or cope with the fact afterwards (which is usually	752	calling this function, or cope with the fact afterwards (which is usually
332	the easiest thing, youc na just ignore the watchers and/or C<free ()> them	753	the easiest thing, you can just ignore the watchers and/or C<free ()> them
333	for example).	754	for example).
334		755
		756	Note that certain global state, such as signal state (and installed signal
		757	handlers), will not be freed by this function, and related watchers (such
		758	as signal and child watchers) would need to be stopped manually.
		759
		760	This function is normally used on loop objects allocated by
		761	C<ev_loop_new>, but it can also be used on the default loop returned by
		762	C<ev_default_loop>, in which case it is not thread-safe.
		763
		764	Note that it is not advisable to call this function on the default loop
		765	except in the rare occasion where you really need to free its resources.
		766	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		767	and C<ev_loop_destroy>.
		768
335	=item ev_loop_destroy (loop)	769	=item ev_loop_fork (loop)
336		770
337	Like C<ev_default_destroy>, but destroys an event loop created by an	771	This function sets a flag that causes subsequent C<ev_run> iterations
338	earlier call to C<ev_loop_new>.
339
340	=item ev_default_fork ()
341
342	This function reinitialises the kernel state for backends that have	772	to reinitialise the kernel state for backends that have one. Despite
343	one. Despite the name, you can call it anytime, but it makes most sense	773	the name, you can call it anytime you are allowed to start or stop
344	after forking, in either the parent or child process (or both, but that	774	watchers (except inside an C<ev_prepare> callback), but it makes most
345	again makes little sense).	775	sense after forking, in the child process. You I<must> call it (or use
		776	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
346		777
347	You I<must> call this function in the child process after forking if and	778	In addition, if you want to reuse a loop (via this function or
348	only if you want to use the event library in both processes. If you just	779	C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
349	fork+exec, you don't have to call it.	780
		781	Again, you I<have> to call it on I<any> loop that you want to re-use after
		782	a fork, I<even if you do not plan to use the loop in the parent>. This is
		783	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		784	during fork.
		785
		786	On the other hand, you only need to call this function in the child
		787	process if and only if you want to use the event loop in the child. If
		788	you just fork+exec or create a new loop in the child, you don't have to
		789	call it at all (in fact, C<epoll> is so badly broken that it makes a
		790	difference, but libev will usually detect this case on its own and do a
		791	costly reset of the backend).
350		792
351	The function itself is quite fast and it's usually not a problem to call	793	The function itself is quite fast and it's usually not a problem to call
352	it just in case after a fork. To make this easy, the function will fit in	794	it just in case after a fork.
353	quite nicely into a call to C<pthread_atfork>:
354		795
		796	Example: Automate calling C<ev_loop_fork> on the default loop when
		797	using pthreads.
		798
		799	static void
		800	post_fork_child (void)
		801	{
		802	ev_loop_fork (EV_DEFAULT);
		803	}
		804
		805	...
355	pthread_atfork (0, 0, ev_default_fork);	806	pthread_atfork (0, 0, post_fork_child);
356		807
357	At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use	808	=item int ev_is_default_loop (loop)
358	without calling this function, so if you force one of those backends you
359	do not need to care.
360		809
361	=item ev_loop_fork (loop)	810	Returns true when the given loop is, in fact, the default loop, and false
		811	otherwise.
362		812
363	Like C<ev_default_fork>, but acts on an event loop created by	813	=item unsigned int ev_iteration (loop)
364	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	814
365	after fork, and how you do this is entirely your own problem.	815	Returns the current iteration count for the event loop, which is identical
		816	to the number of times libev did poll for new events. It starts at C<0>
		817	and happily wraps around with enough iterations.
		818
		819	This value can sometimes be useful as a generation counter of sorts (it
		820	"ticks" the number of loop iterations), as it roughly corresponds with
		821	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		822	prepare and check phases.
		823
		824	=item unsigned int ev_depth (loop)
		825
		826	Returns the number of times C<ev_run> was entered minus the number of
		827	times C<ev_run> was exited normally, in other words, the recursion depth.
		828
		829	Outside C<ev_run>, this number is zero. In a callback, this number is
		830	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		831	in which case it is higher.
		832
		833	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		834	throwing an exception etc.), doesn't count as "exit" - consider this
		835	as a hint to avoid such ungentleman-like behaviour unless it's really
		836	convenient, in which case it is fully supported.
366		837
367	=item unsigned int ev_backend (loop)	838	=item unsigned int ev_backend (loop)
368		839
369	Returns one of the C<EVBACKEND_*> flags indicating the event backend in	840	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
370	use.	841	use.
…		…
373		844
374	Returns the current "event loop time", which is the time the event loop	845	Returns the current "event loop time", which is the time the event loop
375	received events and started processing them. This timestamp does not	846	received events and started processing them. This timestamp does not
376	change as long as callbacks are being processed, and this is also the base	847	change as long as callbacks are being processed, and this is also the base
377	time used for relative timers. You can treat it as the timestamp of the	848	time used for relative timers. You can treat it as the timestamp of the
378	event occuring (or more correctly, libev finding out about it).	849	event occurring (or more correctly, libev finding out about it).
379		850
		851	=item ev_now_update (loop)
		852
		853	Establishes the current time by querying the kernel, updating the time
		854	returned by C<ev_now ()> in the progress. This is a costly operation and
		855	is usually done automatically within C<ev_run ()>.
		856
		857	This function is rarely useful, but when some event callback runs for a
		858	very long time without entering the event loop, updating libev's idea of
		859	the current time is a good idea.
		860
		861	See also L</The special problem of time updates> in the C<ev_timer> section.
		862
		863	=item ev_suspend (loop)
		864
		865	=item ev_resume (loop)
		866
		867	These two functions suspend and resume an event loop, for use when the
		868	loop is not used for a while and timeouts should not be processed.
		869
		870	A typical use case would be an interactive program such as a game: When
		871	the user presses C<^Z> to suspend the game and resumes it an hour later it
		872	would be best to handle timeouts as if no time had actually passed while
		873	the program was suspended. This can be achieved by calling C<ev_suspend>
		874	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		875	C<ev_resume> directly afterwards to resume timer processing.
		876
		877	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		878	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		879	will be rescheduled (that is, they will lose any events that would have
		880	occurred while suspended).
		881
		882	After calling C<ev_suspend> you B<must not> call I<any> function on the
		883	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		884	without a previous call to C<ev_suspend>.
		885
		886	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		887	event loop time (see C<ev_now_update>).
		888
380	=item ev_loop (loop, int flags)	889	=item bool ev_run (loop, int flags)
381		890
382	Finally, this is it, the event handler. This function usually is called	891	Finally, this is it, the event handler. This function usually is called
383	after you initialised all your watchers and you want to start handling	892	after you have initialised all your watchers and you want to start
384	events.	893	handling events. It will ask the operating system for any new events, call
		894	the watcher callbacks, and then repeat the whole process indefinitely: This
		895	is why event loops are called I<loops>.
385		896
386	If the flags argument is specified as C<0>, it will not return until	897	If the flags argument is specified as C<0>, it will keep handling events
387	either no event watchers are active anymore or C<ev_unloop> was called.	898	until either no event watchers are active anymore or C<ev_break> was
		899	called.
388		900
		901	The return value is false if there are no more active watchers (which
		902	usually means "all jobs done" or "deadlock"), and true in all other cases
		903	(which usually means " you should call C<ev_run> again").
		904
389	Please note that an explicit C<ev_unloop> is usually better than	905	Please note that an explicit C<ev_break> is usually better than
390	relying on all watchers to be stopped when deciding when a program has	906	relying on all watchers to be stopped when deciding when a program has
391	finished (especially in interactive programs), but having a program that	907	finished (especially in interactive programs), but having a program
392	automatically loops as long as it has to and no longer by virtue of	908	that automatically loops as long as it has to and no longer by virtue
393	relying on its watchers stopping correctly is a thing of beauty.	909	of relying on its watchers stopping correctly, that is truly a thing of
		910	beauty.
394		911
		912	This function is I<mostly> exception-safe - you can break out of a
		913	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		914	exception and so on. This does not decrement the C<ev_depth> value, nor
		915	will it clear any outstanding C<EVBREAK_ONE> breaks.
		916
395	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	917	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
396	those events and any outstanding ones, but will not block your process in	918	those events and any already outstanding ones, but will not wait and
397	case there are no events and will return after one iteration of the loop.	919	block your process in case there are no events and will return after one
		920	iteration of the loop. This is sometimes useful to poll and handle new
		921	events while doing lengthy calculations, to keep the program responsive.
398		922
399	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	923	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
400	neccessary) and will handle those and any outstanding ones. It will block	924	necessary) and will handle those and any already outstanding ones. It
401	your process until at least one new event arrives, and will return after	925	will block your process until at least one new event arrives (which could
402	one iteration of the loop. This is useful if you are waiting for some	926	be an event internal to libev itself, so there is no guarantee that a
403	external event in conjunction with something not expressible using other	927	user-registered callback will be called), and will return after one
		928	iteration of the loop.
		929
		930	This is useful if you are waiting for some external event in conjunction
		931	with something not expressible using other libev watchers (i.e. "roll your
404	libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is	932	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
405	usually a better approach for this kind of thing.	933	usually a better approach for this kind of thing.
406		934
407	Here are the gory details of what C<ev_loop> does:	935	Here are the gory details of what C<ev_run> does (this is for your
		936	understanding, not a guarantee that things will work exactly like this in
		937	future versions):
408		938
409	* If there are no active watchers (reference count is zero), return.	939	- Increment loop depth.
410	- Queue prepare watchers and then call all outstanding watchers.	940	- Reset the ev_break status.
		941	- Before the first iteration, call any pending watchers.
		942	LOOP:
		943	- If EVFLAG_FORKCHECK was used, check for a fork.
		944	- If a fork was detected (by any means), queue and call all fork watchers.
		945	- Queue and call all prepare watchers.
		946	- If ev_break was called, goto FINISH.
411	- If we have been forked, recreate the kernel state.	947	- If we have been forked, detach and recreate the kernel state
		948	as to not disturb the other process.
412	- Update the kernel state with all outstanding changes.	949	- Update the kernel state with all outstanding changes.
413	- Update the "event loop time".	950	- Update the "event loop time" (ev_now ()).
414	- Calculate for how long to block.	951	- Calculate for how long to sleep or block, if at all
		952	(active idle watchers, EVRUN_NOWAIT or not having
		953	any active watchers at all will result in not sleeping).
		954	- Sleep if the I/O and timer collect interval say so.
		955	- Increment loop iteration counter.
415	- Block the process, waiting for any events.	956	- Block the process, waiting for any events.
416	- Queue all outstanding I/O (fd) events.	957	- Queue all outstanding I/O (fd) events.
417	- Update the "event loop time" and do time jump handling.	958	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
418	- Queue all outstanding timers.	959	- Queue all expired timers.
419	- Queue all outstanding periodics.	960	- Queue all expired periodics.
420	- If no events are pending now, queue all idle watchers.	961	- Queue all idle watchers with priority higher than that of pending events.
421	- Queue all check watchers.	962	- Queue all check watchers.
422	- Call all queued watchers in reverse order (i.e. check watchers first).	963	- Call all queued watchers in reverse order (i.e. check watchers first).
423	Signals and child watchers are implemented as I/O watchers, and will	964	Signals and child watchers are implemented as I/O watchers, and will
424	be handled here by queueing them when their watcher gets executed.	965	be handled here by queueing them when their watcher gets executed.
425	- If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	966	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
426	were used, return, otherwise continue with step *.	967	were used, or there are no active watchers, goto FINISH, otherwise
		968	continue with step LOOP.
		969	FINISH:
		970	- Reset the ev_break status iff it was EVBREAK_ONE.
		971	- Decrement the loop depth.
		972	- Return.
427		973
428	Example: queue some jobs and then loop until no events are outsanding	974	Example: Queue some jobs and then loop until no events are outstanding
429	anymore.	975	anymore.
430		976
431	... queue jobs here, make sure they register event watchers as long	977	... queue jobs here, make sure they register event watchers as long
432	... as they still have work to do (even an idle watcher will do..)	978	... as they still have work to do (even an idle watcher will do..)
433	ev_loop (my_loop, 0);	979	ev_run (my_loop, 0);
434	... jobs done. yeah!	980	... jobs done or somebody called break. yeah!
435		981
436	=item ev_unloop (loop, how)	982	=item ev_break (loop, how)
437		983
438	Can be used to make a call to C<ev_loop> return early (but only after it	984	Can be used to make a call to C<ev_run> return early (but only after it
439	has processed all outstanding events). The C<how> argument must be either	985	has processed all outstanding events). The C<how> argument must be either
440	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	986	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
441	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	987	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		988
		989	This "break state" will be cleared on the next call to C<ev_run>.
		990
		991	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		992	which case it will have no effect.
442		993
443	=item ev_ref (loop)	994	=item ev_ref (loop)
444		995
445	=item ev_unref (loop)	996	=item ev_unref (loop)
446		997
447	Ref/unref can be used to add or remove a reference count on the event	998	Ref/unref can be used to add or remove a reference count on the event
448	loop: Every watcher keeps one reference, and as long as the reference	999	loop: Every watcher keeps one reference, and as long as the reference
449	count is nonzero, C<ev_loop> will not return on its own. If you have	1000	count is nonzero, C<ev_run> will not return on its own.
450	a watcher you never unregister that should not keep C<ev_loop> from	1001
451	returning, ev_unref() after starting, and ev_ref() before stopping it. For	1002	This is useful when you have a watcher that you never intend to
		1003	unregister, but that nevertheless should not keep C<ev_run> from
		1004	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		1005	before stopping it.
		1006
452	example, libev itself uses this for its internal signal pipe: It is not	1007	As an example, libev itself uses this for its internal signal pipe: It
453	visible to the libev user and should not keep C<ev_loop> from exiting if	1008	is not visible to the libev user and should not keep C<ev_run> from
454	no event watchers registered by it are active. It is also an excellent	1009	exiting if no event watchers registered by it are active. It is also an
455	way to do this for generic recurring timers or from within third-party	1010	excellent way to do this for generic recurring timers or from within
456	libraries. Just remember to I<unref after start> and I<ref before stop>.	1011	third-party libraries. Just remember to I<unref after start> and I<ref
		1012	before stop> (but only if the watcher wasn't active before, or was active
		1013	before, respectively. Note also that libev might stop watchers itself
		1014	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		1015	in the callback).
457		1016
458	Example: create a signal watcher, but keep it from keeping C<ev_loop>	1017	Example: Create a signal watcher, but keep it from keeping C<ev_run>
459	running when nothing else is active.	1018	running when nothing else is active.
460		1019
461	struct dv_signal exitsig;	1020	ev_signal exitsig;
462	ev_signal_init (&exitsig, sig_cb, SIGINT);	1021	ev_signal_init (&exitsig, sig_cb, SIGINT);
463	ev_signal_start (myloop, &exitsig);	1022	ev_signal_start (loop, &exitsig);
464	evf_unref (myloop);	1023	ev_unref (loop);
465		1024
466	Example: for some weird reason, unregister the above signal handler again.	1025	Example: For some weird reason, unregister the above signal handler again.
467		1026
468	ev_ref (myloop);	1027	ev_ref (loop);
469	ev_signal_stop (myloop, &exitsig);	1028	ev_signal_stop (loop, &exitsig);
		1029
		1030	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		1031
		1032	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		1033
		1034	These advanced functions influence the time that libev will spend waiting
		1035	for events. Both time intervals are by default C<0>, meaning that libev
		1036	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		1037	latency.
		1038
		1039	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		1040	allows libev to delay invocation of I/O and timer/periodic callbacks
		1041	to increase efficiency of loop iterations (or to increase power-saving
		1042	opportunities).
		1043
		1044	The idea is that sometimes your program runs just fast enough to handle
		1045	one (or very few) event(s) per loop iteration. While this makes the
		1046	program responsive, it also wastes a lot of CPU time to poll for new
		1047	events, especially with backends like C<select ()> which have a high
		1048	overhead for the actual polling but can deliver many events at once.
		1049
		1050	By setting a higher I<io collect interval> you allow libev to spend more
		1051	time collecting I/O events, so you can handle more events per iteration,
		1052	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		1053	C<ev_timer>) will not be affected. Setting this to a non-null value will
		1054	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		1055	sleep time ensures that libev will not poll for I/O events more often then
		1056	once per this interval, on average (as long as the host time resolution is
		1057	good enough).
		1058
		1059	Likewise, by setting a higher I<timeout collect interval> you allow libev
		1060	to spend more time collecting timeouts, at the expense of increased
		1061	latency/jitter/inexactness (the watcher callback will be called
		1062	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		1063	value will not introduce any overhead in libev.
		1064
		1065	Many (busy) programs can usually benefit by setting the I/O collect
		1066	interval to a value near C<0.1> or so, which is often enough for
		1067	interactive servers (of course not for games), likewise for timeouts. It
		1068	usually doesn't make much sense to set it to a lower value than C<0.01>,
		1069	as this approaches the timing granularity of most systems. Note that if
		1070	you do transactions with the outside world and you can't increase the
		1071	parallelity, then this setting will limit your transaction rate (if you
		1072	need to poll once per transaction and the I/O collect interval is 0.01,
		1073	then you can't do more than 100 transactions per second).
		1074
		1075	Setting the I<timeout collect interval> can improve the opportunity for
		1076	saving power, as the program will "bundle" timer callback invocations that
		1077	are "near" in time together, by delaying some, thus reducing the number of
		1078	times the process sleeps and wakes up again. Another useful technique to
		1079	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		1080	they fire on, say, one-second boundaries only.
		1081
		1082	Example: we only need 0.1s timeout granularity, and we wish not to poll
		1083	more often than 100 times per second:
		1084
		1085	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		1086	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		1087
		1088	=item ev_invoke_pending (loop)
		1089
		1090	This call will simply invoke all pending watchers while resetting their
		1091	pending state. Normally, C<ev_run> does this automatically when required,
		1092	but when overriding the invoke callback this call comes handy. This
		1093	function can be invoked from a watcher - this can be useful for example
		1094	when you want to do some lengthy calculation and want to pass further
		1095	event handling to another thread (you still have to make sure only one
		1096	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		1097
		1098	=item int ev_pending_count (loop)
		1099
		1100	Returns the number of pending watchers - zero indicates that no watchers
		1101	are pending.
		1102
		1103	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1104
		1105	This overrides the invoke pending functionality of the loop: Instead of
		1106	invoking all pending watchers when there are any, C<ev_run> will call
		1107	this callback instead. This is useful, for example, when you want to
		1108	invoke the actual watchers inside another context (another thread etc.).
		1109
		1110	If you want to reset the callback, use C<ev_invoke_pending> as new
		1111	callback.
		1112
		1113	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
		1114
		1115	Sometimes you want to share the same loop between multiple threads. This
		1116	can be done relatively simply by putting mutex_lock/unlock calls around
		1117	each call to a libev function.
		1118
		1119	However, C<ev_run> can run an indefinite time, so it is not feasible
		1120	to wait for it to return. One way around this is to wake up the event
		1121	loop via C<ev_break> and C<ev_async_send>, another way is to set these
		1122	I<release> and I<acquire> callbacks on the loop.
		1123
		1124	When set, then C<release> will be called just before the thread is
		1125	suspended waiting for new events, and C<acquire> is called just
		1126	afterwards.
		1127
		1128	Ideally, C<release> will just call your mutex_unlock function, and
		1129	C<acquire> will just call the mutex_lock function again.
		1130
		1131	While event loop modifications are allowed between invocations of
		1132	C<release> and C<acquire> (that's their only purpose after all), no
		1133	modifications done will affect the event loop, i.e. adding watchers will
		1134	have no effect on the set of file descriptors being watched, or the time
		1135	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1136	to take note of any changes you made.
		1137
		1138	In theory, threads executing C<ev_run> will be async-cancel safe between
		1139	invocations of C<release> and C<acquire>.
		1140
		1141	See also the locking example in the C<THREADS> section later in this
		1142	document.
		1143
		1144	=item ev_set_userdata (loop, void *data)
		1145
		1146	=item void *ev_userdata (loop)
		1147
		1148	Set and retrieve a single C<void *> associated with a loop. When
		1149	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1150	C<0>.
		1151
		1152	These two functions can be used to associate arbitrary data with a loop,
		1153	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1154	C<acquire> callbacks described above, but of course can be (ab-)used for
		1155	any other purpose as well.
		1156
		1157	=item ev_verify (loop)
		1158
		1159	This function only does something when C<EV_VERIFY> support has been
		1160	compiled in, which is the default for non-minimal builds. It tries to go
		1161	through all internal structures and checks them for validity. If anything
		1162	is found to be inconsistent, it will print an error message to standard
		1163	error and call C<abort ()>.
		1164
		1165	This can be used to catch bugs inside libev itself: under normal
		1166	circumstances, this function will never abort as of course libev keeps its
		1167	data structures consistent.
470		1168
471	=back	1169	=back
472		1170
473		1171
474	=head1 ANATOMY OF A WATCHER	1172	=head1 ANATOMY OF A WATCHER
475		1173
		1174	In the following description, uppercase C<TYPE> in names stands for the
		1175	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1176	watchers and C<ev_io_start> for I/O watchers.
		1177
476	A watcher is a structure that you create and register to record your	1178	A watcher is an opaque structure that you allocate and register to record
477	interest in some event. For instance, if you want to wait for STDIN to	1179	your interest in some event. To make a concrete example, imagine you want
478	become readable, you would create an C<ev_io> watcher for that:	1180	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1181	for that:
479		1182
480	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1183	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
481	{	1184	{
482	ev_io_stop (w);	1185	ev_io_stop (w);
483	ev_unloop (loop, EVUNLOOP_ALL);	1186	ev_break (loop, EVBREAK_ALL);
484	}	1187	}
485		1188
486	struct ev_loop *loop = ev_default_loop (0);	1189	struct ev_loop *loop = ev_default_loop (0);
		1190
487	struct ev_io stdin_watcher;	1191	ev_io stdin_watcher;
		1192
488	ev_init (&stdin_watcher, my_cb);	1193	ev_init (&stdin_watcher, my_cb);
489	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1194	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
490	ev_io_start (loop, &stdin_watcher);	1195	ev_io_start (loop, &stdin_watcher);
		1196
491	ev_loop (loop, 0);	1197	ev_run (loop, 0);
492		1198
493	As you can see, you are responsible for allocating the memory for your	1199	As you can see, you are responsible for allocating the memory for your
494	watcher structures (and it is usually a bad idea to do this on the stack,	1200	watcher structures (and it is I<usually> a bad idea to do this on the
495	although this can sometimes be quite valid).	1201	stack).
496		1202
		1203	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1204	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1205
497	Each watcher structure must be initialised by a call to C<ev_init	1206	Each watcher structure must be initialised by a call to C<ev_init (watcher
498	(watcher *, callback)>, which expects a callback to be provided. This	1207	*, callback)>, which expects a callback to be provided. This callback is
499	callback gets invoked each time the event occurs (or, in the case of io	1208	invoked each time the event occurs (or, in the case of I/O watchers, each
500	watchers, each time the event loop detects that the file descriptor given	1209	time the event loop detects that the file descriptor given is readable
501	is readable and/or writable).	1210	and/or writable).
502		1211
503	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1212	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
504	with arguments specific to this watcher type. There is also a macro	1213	macro to configure it, with arguments specific to the watcher type. There
505	to combine initialisation and setting in one call: C<< ev_<type>_init	1214	is also a macro to combine initialisation and setting in one call: C<<
506	(watcher *, callback, ...) >>.	1215	ev_TYPE_init (watcher *, callback, ...) >>.
507		1216
508	To make the watcher actually watch out for events, you have to start it	1217	To make the watcher actually watch out for events, you have to start it
509	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1218	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
510	*) >>), and you can stop watching for events at any time by calling the	1219	*) >>), and you can stop watching for events at any time by calling the
511	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1220	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
512		1221
513	As long as your watcher is active (has been started but not stopped) you	1222	As long as your watcher is active (has been started but not stopped) you
514	must not touch the values stored in it. Most specifically you must never	1223	must not touch the values stored in it. Most specifically you must never
515	reinitialise it or call its C<set> macro.	1224	reinitialise it or call its C<ev_TYPE_set> macro.
516		1225
517	Each and every callback receives the event loop pointer as first, the	1226	Each and every callback receives the event loop pointer as first, the
518	registered watcher structure as second, and a bitset of received events as	1227	registered watcher structure as second, and a bitset of received events as
519	third argument.	1228	third argument.
520		1229
…		…
529	=item C<EV_WRITE>	1238	=item C<EV_WRITE>
530		1239
531	The file descriptor in the C<ev_io> watcher has become readable and/or	1240	The file descriptor in the C<ev_io> watcher has become readable and/or
532	writable.	1241	writable.
533		1242
534	=item C<EV_TIMEOUT>	1243	=item C<EV_TIMER>
535		1244
536	The C<ev_timer> watcher has timed out.	1245	The C<ev_timer> watcher has timed out.
537		1246
538	=item C<EV_PERIODIC>	1247	=item C<EV_PERIODIC>
539		1248
…		…
545		1254
546	=item C<EV_CHILD>	1255	=item C<EV_CHILD>
547		1256
548	The pid specified in the C<ev_child> watcher has received a status change.	1257	The pid specified in the C<ev_child> watcher has received a status change.
549		1258
		1259	=item C<EV_STAT>
		1260
		1261	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1262
550	=item C<EV_IDLE>	1263	=item C<EV_IDLE>
551		1264
552	The C<ev_idle> watcher has determined that you have nothing better to do.	1265	The C<ev_idle> watcher has determined that you have nothing better to do.
553		1266
554	=item C<EV_PREPARE>	1267	=item C<EV_PREPARE>
555		1268
556	=item C<EV_CHECK>	1269	=item C<EV_CHECK>
557		1270
558	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1271	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
559	to gather new events, and all C<ev_check> watchers are invoked just after	1272	gather new events, and all C<ev_check> watchers are queued (not invoked)
560	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1273	just after C<ev_run> has gathered them, but before it queues any callbacks
		1274	for any received events. That means C<ev_prepare> watchers are the last
		1275	watchers invoked before the event loop sleeps or polls for new events, and
		1276	C<ev_check> watchers will be invoked before any other watchers of the same
		1277	or lower priority within an event loop iteration.
		1278
561	received events. Callbacks of both watcher types can start and stop as	1279	Callbacks of both watcher types can start and stop as many watchers as
562	many watchers as they want, and all of them will be taken into account	1280	they want, and all of them will be taken into account (for example, a
563	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1281	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
564	C<ev_loop> from blocking).	1282	blocking).
		1283
		1284	=item C<EV_EMBED>
		1285
		1286	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1287
		1288	=item C<EV_FORK>
		1289
		1290	The event loop has been resumed in the child process after fork (see
		1291	C<ev_fork>).
		1292
		1293	=item C<EV_CLEANUP>
		1294
		1295	The event loop is about to be destroyed (see C<ev_cleanup>).
		1296
		1297	=item C<EV_ASYNC>
		1298
		1299	The given async watcher has been asynchronously notified (see C<ev_async>).
		1300
		1301	=item C<EV_CUSTOM>
		1302
		1303	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1304	by libev users to signal watchers (e.g. via C<ev_feed_event>).
565		1305
566	=item C<EV_ERROR>	1306	=item C<EV_ERROR>
567		1307
568	An unspecified error has occured, the watcher has been stopped. This might	1308	An unspecified error has occurred, the watcher has been stopped. This might
569	happen because the watcher could not be properly started because libev	1309	happen because the watcher could not be properly started because libev
570	ran out of memory, a file descriptor was found to be closed or any other	1310	ran out of memory, a file descriptor was found to be closed or any other
		1311	problem. Libev considers these application bugs.
		1312
571	problem. You best act on it by reporting the problem and somehow coping	1313	You best act on it by reporting the problem and somehow coping with the
572	with the watcher being stopped.	1314	watcher being stopped. Note that well-written programs should not receive
		1315	an error ever, so when your watcher receives it, this usually indicates a
		1316	bug in your program.
573		1317
574	Libev will usually signal a few "dummy" events together with an error,	1318	Libev will usually signal a few "dummy" events together with an error, for
575	for example it might indicate that a fd is readable or writable, and if	1319	example it might indicate that a fd is readable or writable, and if your
576	your callbacks is well-written it can just attempt the operation and cope	1320	callbacks is well-written it can just attempt the operation and cope with
577	with the error from read() or write(). This will not work in multithreaded	1321	the error from read() or write(). This will not work in multi-threaded
578	programs, though, so beware.	1322	programs, though, as the fd could already be closed and reused for another
		1323	thing, so beware.
579		1324
580	=back	1325	=back
581		1326
582	=head2 GENERIC WATCHER FUNCTIONS	1327	=head2 GENERIC WATCHER FUNCTIONS
583
584	In the following description, C<TYPE> stands for the watcher type,
585	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
586		1328
587	=over 4	1329	=over 4
588		1330
589	=item C<ev_init> (ev_TYPE *watcher, callback)	1331	=item C<ev_init> (ev_TYPE *watcher, callback)
590		1332
…		…
596	which rolls both calls into one.	1338	which rolls both calls into one.
597		1339
598	You can reinitialise a watcher at any time as long as it has been stopped	1340	You can reinitialise a watcher at any time as long as it has been stopped
599	(or never started) and there are no pending events outstanding.	1341	(or never started) and there are no pending events outstanding.
600		1342
601	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,	1343	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
602	int revents)>.	1344	int revents)>.
603		1345
		1346	Example: Initialise an C<ev_io> watcher in two steps.
		1347
		1348	ev_io w;
		1349	ev_init (&w, my_cb);
		1350	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1351
604	=item C<ev_TYPE_set> (ev_TYPE *, [args])	1352	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
605		1353
606	This macro initialises the type-specific parts of a watcher. You need to	1354	This macro initialises the type-specific parts of a watcher. You need to
607	call C<ev_init> at least once before you call this macro, but you can	1355	call C<ev_init> at least once before you call this macro, but you can
608	call C<ev_TYPE_set> any number of times. You must not, however, call this	1356	call C<ev_TYPE_set> any number of times. You must not, however, call this
609	macro on a watcher that is active (it can be pending, however, which is a	1357	macro on a watcher that is active (it can be pending, however, which is a
610	difference to the C<ev_init> macro).	1358	difference to the C<ev_init> macro).
611		1359
612	Although some watcher types do not have type-specific arguments	1360	Although some watcher types do not have type-specific arguments
613	(e.g. C<ev_prepare>) you still need to call its C<set> macro.	1361	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
614		1362
		1363	See C<ev_init>, above, for an example.
		1364
615	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])	1365	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
616		1366
617	This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro	1367	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
618	calls into a single call. This is the most convinient method to initialise	1368	calls into a single call. This is the most convenient method to initialise
619	a watcher. The same limitations apply, of course.	1369	a watcher. The same limitations apply, of course.
620		1370
		1371	Example: Initialise and set an C<ev_io> watcher in one step.
		1372
		1373	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1374
621	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)	1375	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
622		1376
623	Starts (activates) the given watcher. Only active watchers will receive	1377	Starts (activates) the given watcher. Only active watchers will receive
624	events. If the watcher is already active nothing will happen.	1378	events. If the watcher is already active nothing will happen.
625		1379
		1380	Example: Start the C<ev_io> watcher that is being abused as example in this
		1381	whole section.
		1382
		1383	ev_io_start (EV_DEFAULT_UC, &w);
		1384
626	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)	1385	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
627		1386
628	Stops the given watcher again (if active) and clears the pending	1387	Stops the given watcher if active, and clears the pending status (whether
		1388	the watcher was active or not).
		1389
629	status. It is possible that stopped watchers are pending (for example,	1390	It is possible that stopped watchers are pending - for example,
630	non-repeating timers are being stopped when they become pending), but	1391	non-repeating timers are being stopped when they become pending - but
631	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If	1392	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
632	you want to free or reuse the memory used by the watcher it is therefore a	1393	pending. If you want to free or reuse the memory used by the watcher it is
633	good idea to always call its C<ev_TYPE_stop> function.	1394	therefore a good idea to always call its C<ev_TYPE_stop> function.
634		1395
635	=item bool ev_is_active (ev_TYPE *watcher)	1396	=item bool ev_is_active (ev_TYPE *watcher)
636		1397
637	Returns a true value iff the watcher is active (i.e. it has been started	1398	Returns a true value iff the watcher is active (i.e. it has been started
638	and not yet been stopped). As long as a watcher is active you must not modify	1399	and not yet been stopped). As long as a watcher is active you must not modify
…		…
641	=item bool ev_is_pending (ev_TYPE *watcher)	1402	=item bool ev_is_pending (ev_TYPE *watcher)
642		1403
643	Returns a true value iff the watcher is pending, (i.e. it has outstanding	1404	Returns a true value iff the watcher is pending, (i.e. it has outstanding
644	events but its callback has not yet been invoked). As long as a watcher	1405	events but its callback has not yet been invoked). As long as a watcher
645	is pending (but not active) you must not call an init function on it (but	1406	is pending (but not active) you must not call an init function on it (but
646	C<ev_TYPE_set> is safe) and you must make sure the watcher is available to	1407	C<ev_TYPE_set> is safe), you must not change its priority, and you must
647	libev (e.g. you cnanot C<free ()> it).	1408	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1409	it).
648		1410
649	=item callback = ev_cb (ev_TYPE *watcher)	1411	=item callback ev_cb (ev_TYPE *watcher)
650		1412
651	Returns the callback currently set on the watcher.	1413	Returns the callback currently set on the watcher.
652		1414
653	=item ev_cb_set (ev_TYPE *watcher, callback)	1415	=item ev_set_cb (ev_TYPE *watcher, callback)
654		1416
655	Change the callback. You can change the callback at virtually any time	1417	Change the callback. You can change the callback at virtually any time
656	(modulo threads).	1418	(modulo threads).
657		1419
		1420	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1421
		1422	=item int ev_priority (ev_TYPE *watcher)
		1423
		1424	Set and query the priority of the watcher. The priority is a small
		1425	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1426	(default: C<-2>). Pending watchers with higher priority will be invoked
		1427	before watchers with lower priority, but priority will not keep watchers
		1428	from being executed (except for C<ev_idle> watchers).
		1429
		1430	If you need to suppress invocation when higher priority events are pending
		1431	you need to look at C<ev_idle> watchers, which provide this functionality.
		1432
		1433	You I<must not> change the priority of a watcher as long as it is active or
		1434	pending.
		1435
		1436	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1437	fine, as long as you do not mind that the priority value you query might
		1438	or might not have been clamped to the valid range.
		1439
		1440	The default priority used by watchers when no priority has been set is
		1441	always C<0>, which is supposed to not be too high and not be too low :).
		1442
		1443	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1444	priorities.
		1445
		1446	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1447
		1448	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1449	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1450	can deal with that fact, as both are simply passed through to the
		1451	callback.
		1452
		1453	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1454
		1455	If the watcher is pending, this function clears its pending status and
		1456	returns its C<revents> bitset (as if its callback was invoked). If the
		1457	watcher isn't pending it does nothing and returns C<0>.
		1458
		1459	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1460	callback to be invoked, which can be accomplished with this function.
		1461
		1462	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1463
		1464	Feeds the given event set into the event loop, as if the specified event
		1465	had happened for the specified watcher (which must be a pointer to an
		1466	initialised but not necessarily started event watcher). Obviously you must
		1467	not free the watcher as long as it has pending events.
		1468
		1469	Stopping the watcher, letting libev invoke it, or calling
		1470	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1471	not started in the first place.
		1472
		1473	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1474	functions that do not need a watcher.
		1475
658	=back	1476	=back
659		1477
		1478	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
		1479	OWN COMPOSITE WATCHERS> idioms.
660		1480
661	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1481	=head2 WATCHER STATES
662		1482
663	Each watcher has, by default, a member C<void *data> that you can change	1483	There are various watcher states mentioned throughout this manual -
664	and read at any time, libev will completely ignore it. This can be used	1484	active, pending and so on. In this section these states and the rules to
665	to associate arbitrary data with your watcher. If you need more data and	1485	transition between them will be described in more detail - and while these
666	don't want to allocate memory and store a pointer to it in that data	1486	rules might look complicated, they usually do "the right thing".
667	member, you can also "subclass" the watcher type and provide your own
668	data:
669		1487
670	struct my_io	1488	=over 4
		1489
		1490	=item initialised
		1491
		1492	Before a watcher can be registered with the event loop it has to be
		1493	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1494	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1495
		1496	In this state it is simply some block of memory that is suitable for
		1497	use in an event loop. It can be moved around, freed, reused etc. at
		1498	will - as long as you either keep the memory contents intact, or call
		1499	C<ev_TYPE_init> again.
		1500
		1501	=item started/running/active
		1502
		1503	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1504	property of the event loop, and is actively waiting for events. While in
		1505	this state it cannot be accessed (except in a few documented ways), moved,
		1506	freed or anything else - the only legal thing is to keep a pointer to it,
		1507	and call libev functions on it that are documented to work on active watchers.
		1508
		1509	=item pending
		1510
		1511	If a watcher is active and libev determines that an event it is interested
		1512	in has occurred (such as a timer expiring), it will become pending. It will
		1513	stay in this pending state until either it is stopped or its callback is
		1514	about to be invoked, so it is not normally pending inside the watcher
		1515	callback.
		1516
		1517	The watcher might or might not be active while it is pending (for example,
		1518	an expired non-repeating timer can be pending but no longer active). If it
		1519	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1520	but it is still property of the event loop at this time, so cannot be
		1521	moved, freed or reused. And if it is active the rules described in the
		1522	previous item still apply.
		1523
		1524	It is also possible to feed an event on a watcher that is not active (e.g.
		1525	via C<ev_feed_event>), in which case it becomes pending without being
		1526	active.
		1527
		1528	=item stopped
		1529
		1530	A watcher can be stopped implicitly by libev (in which case it might still
		1531	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1532	latter will clear any pending state the watcher might be in, regardless
		1533	of whether it was active or not, so stopping a watcher explicitly before
		1534	freeing it is often a good idea.
		1535
		1536	While stopped (and not pending) the watcher is essentially in the
		1537	initialised state, that is, it can be reused, moved, modified in any way
		1538	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1539	it again).
		1540
		1541	=back
		1542
		1543	=head2 WATCHER PRIORITY MODELS
		1544
		1545	Many event loops support I<watcher priorities>, which are usually small
		1546	integers that influence the ordering of event callback invocation
		1547	between watchers in some way, all else being equal.
		1548
		1549	In libev, watcher priorities can be set using C<ev_set_priority>. See its
		1550	description for the more technical details such as the actual priority
		1551	range.
		1552
		1553	There are two common ways how these these priorities are being interpreted
		1554	by event loops:
		1555
		1556	In the more common lock-out model, higher priorities "lock out" invocation
		1557	of lower priority watchers, which means as long as higher priority
		1558	watchers receive events, lower priority watchers are not being invoked.
		1559
		1560	The less common only-for-ordering model uses priorities solely to order
		1561	callback invocation within a single event loop iteration: Higher priority
		1562	watchers are invoked before lower priority ones, but they all get invoked
		1563	before polling for new events.
		1564
		1565	Libev uses the second (only-for-ordering) model for all its watchers
		1566	except for idle watchers (which use the lock-out model).
		1567
		1568	The rationale behind this is that implementing the lock-out model for
		1569	watchers is not well supported by most kernel interfaces, and most event
		1570	libraries will just poll for the same events again and again as long as
		1571	their callbacks have not been executed, which is very inefficient in the
		1572	common case of one high-priority watcher locking out a mass of lower
		1573	priority ones.
		1574
		1575	Static (ordering) priorities are most useful when you have two or more
		1576	watchers handling the same resource: a typical usage example is having an
		1577	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1578	timeouts. Under load, data might be received while the program handles
		1579	other jobs, but since timers normally get invoked first, the timeout
		1580	handler will be executed before checking for data. In that case, giving
		1581	the timer a lower priority than the I/O watcher ensures that I/O will be
		1582	handled first even under adverse conditions (which is usually, but not
		1583	always, what you want).
		1584
		1585	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1586	will only be executed when no same or higher priority watchers have
		1587	received events, they can be used to implement the "lock-out" model when
		1588	required.
		1589
		1590	For example, to emulate how many other event libraries handle priorities,
		1591	you can associate an C<ev_idle> watcher to each such watcher, and in
		1592	the normal watcher callback, you just start the idle watcher. The real
		1593	processing is done in the idle watcher callback. This causes libev to
		1594	continuously poll and process kernel event data for the watcher, but when
		1595	the lock-out case is known to be rare (which in turn is rare :), this is
		1596	workable.
		1597
		1598	Usually, however, the lock-out model implemented that way will perform
		1599	miserably under the type of load it was designed to handle. In that case,
		1600	it might be preferable to stop the real watcher before starting the
		1601	idle watcher, so the kernel will not have to process the event in case
		1602	the actual processing will be delayed for considerable time.
		1603
		1604	Here is an example of an I/O watcher that should run at a strictly lower
		1605	priority than the default, and which should only process data when no
		1606	other events are pending:
		1607
		1608	ev_idle idle; // actual processing watcher
		1609	ev_io io; // actual event watcher
		1610
		1611	static void
		1612	io_cb (EV_P_ ev_io *w, int revents)
671	{	1613	{
672	struct ev_io io;	1614	// stop the I/O watcher, we received the event, but
673	int otherfd;	1615	// are not yet ready to handle it.
674	void *somedata;	1616	ev_io_stop (EV_A_ w);
675	struct whatever *mostinteresting;	1617
		1618	// start the idle watcher to handle the actual event.
		1619	// it will not be executed as long as other watchers
		1620	// with the default priority are receiving events.
		1621	ev_idle_start (EV_A_ &idle);
676	}	1622	}
677		1623
678	And since your callback will be called with a pointer to the watcher, you	1624	static void
679	can cast it back to your own type:	1625	idle_cb (EV_P_ ev_idle *w, int revents)
680
681	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
682	{	1626	{
683	struct my_io *w = (struct my_io *)w_;	1627	// actual processing
684	...	1628	read (STDIN_FILENO, ...);
		1629
		1630	// have to start the I/O watcher again, as
		1631	// we have handled the event
		1632	ev_io_start (EV_P_ &io);
685	}	1633	}
686		1634
687	More interesting and less C-conformant ways of catsing your callback type	1635	// initialisation
688	have been omitted....	1636	ev_idle_init (&idle, idle_cb);
		1637	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1638	ev_io_start (EV_DEFAULT_ &io);
		1639
		1640	In the "real" world, it might also be beneficial to start a timer, so that
		1641	low-priority connections can not be locked out forever under load. This
		1642	enables your program to keep a lower latency for important connections
		1643	during short periods of high load, while not completely locking out less
		1644	important ones.
689		1645
690		1646
691	=head1 WATCHER TYPES	1647	=head1 WATCHER TYPES
692		1648
693	This section describes each watcher in detail, but will not repeat	1649	This section describes each watcher in detail, but will not repeat
694	information given in the last section.	1650	information given in the last section. Any initialisation/set macros,
		1651	functions and members specific to the watcher type are explained.
695		1652
		1653	Most members are additionally marked with either I<[read-only]>, meaning
		1654	that, while the watcher is active, you can look at the member and expect
		1655	some sensible content, but you must not modify it (you can modify it while
		1656	the watcher is stopped to your hearts content), or I<[read-write]>, which
		1657	means you can expect it to have some sensible content while the watcher
		1658	is active, but you can also modify it. Modifying it may not do something
		1659	sensible or take immediate effect (or do anything at all), but libev will
		1660	not crash or malfunction in any way.
		1661
		1662	In any case, the documentation for each member will explain what the
		1663	effects are, and if there are any additional access restrictions.
696		1664
697	=head2 C<ev_io> - is this file descriptor readable or writable?	1665	=head2 C<ev_io> - is this file descriptor readable or writable?
698		1666
699	I/O watchers check whether a file descriptor is readable or writable	1667	I/O watchers check whether a file descriptor is readable or writable
700	in each iteration of the event loop, or, more precisely, when reading	1668	in each iteration of the event loop, or, more precisely, when reading
…		…
707	In general you can register as many read and/or write event watchers per	1675	In general you can register as many read and/or write event watchers per
708	fd as you want (as long as you don't confuse yourself). Setting all file	1676	fd as you want (as long as you don't confuse yourself). Setting all file
709	descriptors to non-blocking mode is also usually a good idea (but not	1677	descriptors to non-blocking mode is also usually a good idea (but not
710	required if you know what you are doing).	1678	required if you know what you are doing).
711		1679
712	You have to be careful with dup'ed file descriptors, though. Some backends
713	(the linux epoll backend is a notable example) cannot handle dup'ed file
714	descriptors correctly if you register interest in two or more fds pointing
715	to the same underlying file/socket/etc. description (that is, they share
716	the same underlying "file open").
717
718	If you must do this, then force the use of a known-to-be-good backend
719	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
720	C<EVBACKEND_POLL>).
721
722	Another thing you have to watch out for is that it is quite easy to	1680	Another thing you have to watch out for is that it is quite easy to
723	receive "spurious" readyness notifications, that is your callback might	1681	receive "spurious" readiness notifications, that is, your callback might
724	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1682	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
725	because there is no data. Not only are some backends known to create a	1683	because there is no data. It is very easy to get into this situation even
726	lot of those (for example solaris ports), it is very easy to get into	1684	with a relatively standard program structure. Thus it is best to always
727	this situation even with a relatively standard program structure. Thus	1685	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
728	it is best to always use non-blocking I/O: An extra C<read>(2) returning
729	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1686	preferable to a program hanging until some data arrives.
730		1687
731	If you cannot run the fd in non-blocking mode (for example you should not	1688	If you cannot run the fd in non-blocking mode (for example you should
732	play around with an Xlib connection), then you have to seperately re-test	1689	not play around with an Xlib connection), then you have to separately
733	wether a file descriptor is really ready with a known-to-be good interface	1690	re-test whether a file descriptor is really ready with a known-to-be good
734	such as poll (fortunately in our Xlib example, Xlib already does this on	1691	interface such as poll (fortunately in the case of Xlib, it already does
735	its own, so its quite safe to use).	1692	this on its own, so its quite safe to use). Some people additionally
		1693	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1694	indefinitely.
		1695
		1696	But really, best use non-blocking mode.
		1697
		1698	=head3 The special problem of disappearing file descriptors
		1699
		1700	Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
		1701	a file descriptor (either due to calling C<close> explicitly or any other
		1702	means, such as C<dup2>). The reason is that you register interest in some
		1703	file descriptor, but when it goes away, the operating system will silently
		1704	drop this interest. If another file descriptor with the same number then
		1705	is registered with libev, there is no efficient way to see that this is,
		1706	in fact, a different file descriptor.
		1707
		1708	To avoid having to explicitly tell libev about such cases, libev follows
		1709	the following policy: Each time C<ev_io_set> is being called, libev
		1710	will assume that this is potentially a new file descriptor, otherwise
		1711	it is assumed that the file descriptor stays the same. That means that
		1712	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1713	descriptor even if the file descriptor number itself did not change.
		1714
		1715	This is how one would do it normally anyway, the important point is that
		1716	the libev application should not optimise around libev but should leave
		1717	optimisations to libev.
		1718
		1719	=head3 The special problem of dup'ed file descriptors
		1720
		1721	Some backends (e.g. epoll), cannot register events for file descriptors,
		1722	but only events for the underlying file descriptions. That means when you
		1723	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1724	events for them, only one file descriptor might actually receive events.
		1725
		1726	There is no workaround possible except not registering events
		1727	for potentially C<dup ()>'ed file descriptors, or to resort to
		1728	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1729
		1730	=head3 The special problem of files
		1731
		1732	Many people try to use C<select> (or libev) on file descriptors
		1733	representing files, and expect it to become ready when their program
		1734	doesn't block on disk accesses (which can take a long time on their own).
		1735
		1736	However, this cannot ever work in the "expected" way - you get a readiness
		1737	notification as soon as the kernel knows whether and how much data is
		1738	there, and in the case of open files, that's always the case, so you
		1739	always get a readiness notification instantly, and your read (or possibly
		1740	write) will still block on the disk I/O.
		1741
		1742	Another way to view it is that in the case of sockets, pipes, character
		1743	devices and so on, there is another party (the sender) that delivers data
		1744	on its own, but in the case of files, there is no such thing: the disk
		1745	will not send data on its own, simply because it doesn't know what you
		1746	wish to read - you would first have to request some data.
		1747
		1748	Since files are typically not-so-well supported by advanced notification
		1749	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1750	to files, even though you should not use it. The reason for this is
		1751	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1752	usually a tty, often a pipe, but also sometimes files or special devices
		1753	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1754	F</dev/urandom>), and even though the file might better be served with
		1755	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1756	it "just works" instead of freezing.
		1757
		1758	So avoid file descriptors pointing to files when you know it (e.g. use
		1759	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1760	when you rarely read from a file instead of from a socket, and want to
		1761	reuse the same code path.
		1762
		1763	=head3 The special problem of fork
		1764
		1765	Some backends (epoll, kqueue, linuxaio, iouring) do not support C<fork ()>
		1766	at all or exhibit useless behaviour. Libev fully supports fork, but needs
		1767	to be told about it in the child if you want to continue to use it in the
		1768	child.
		1769
		1770	To support fork in your child processes, you have to call C<ev_loop_fork
		1771	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
		1772	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1773
		1774	=head3 The special problem of SIGPIPE
		1775
		1776	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1777	when writing to a pipe whose other end has been closed, your program gets
		1778	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1779	this is sensible behaviour, for daemons, this is usually undesirable.
		1780
		1781	So when you encounter spurious, unexplained daemon exits, make sure you
		1782	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1783	somewhere, as that would have given you a big clue).
		1784
		1785	=head3 The special problem of accept()ing when you can't
		1786
		1787	Many implementations of the POSIX C<accept> function (for example,
		1788	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1789	connection from the pending queue in all error cases.
		1790
		1791	For example, larger servers often run out of file descriptors (because
		1792	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1793	rejecting the connection, leading to libev signalling readiness on
		1794	the next iteration again (the connection still exists after all), and
		1795	typically causing the program to loop at 100% CPU usage.
		1796
		1797	Unfortunately, the set of errors that cause this issue differs between
		1798	operating systems, there is usually little the app can do to remedy the
		1799	situation, and no known thread-safe method of removing the connection to
		1800	cope with overload is known (to me).
		1801
		1802	One of the easiest ways to handle this situation is to just ignore it
		1803	- when the program encounters an overload, it will just loop until the
		1804	situation is over. While this is a form of busy waiting, no OS offers an
		1805	event-based way to handle this situation, so it's the best one can do.
		1806
		1807	A better way to handle the situation is to log any errors other than
		1808	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1809	messages, and continue as usual, which at least gives the user an idea of
		1810	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1811	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1812	usage.
		1813
		1814	If your program is single-threaded, then you could also keep a dummy file
		1815	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1816	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1817	close that fd, and create a new dummy fd. This will gracefully refuse
		1818	clients under typical overload conditions.
		1819
		1820	The last way to handle it is to simply log the error and C<exit>, as
		1821	is often done with C<malloc> failures, but this results in an easy
		1822	opportunity for a DoS attack.
		1823
		1824	=head3 Watcher-Specific Functions
736		1825
737	=over 4	1826	=over 4
738		1827
739	=item ev_io_init (ev_io *, callback, int fd, int events)	1828	=item ev_io_init (ev_io *, callback, int fd, int events)
740		1829
741	=item ev_io_set (ev_io *, int fd, int events)	1830	=item ev_io_set (ev_io *, int fd, int events)
742		1831
743	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to	1832	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
744	rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or	1833	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
745	C<EV_READ | EV_WRITE> to receive the given events.	1834	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1835
		1836	=item ev_io_modify (ev_io *, int events)
		1837
		1838	Similar to C<ev_io_set>, but only changes the event mask. Using this might
		1839	be faster with some backends, as libev can assume that the C<fd> still
		1840	refers to the same underlying file description, something it cannot do
		1841	when using C<ev_io_set>.
		1842
		1843	=item int fd [no-modify]
		1844
		1845	The file descriptor being watched. While it can be read at any time, you
		1846	must not modify this member even when the watcher is stopped - always use
		1847	C<ev_io_set> for that.
		1848
		1849	=item int events [no-modify]
		1850
		1851	The set of events being watched, among other flags. This field is a
		1852	bit set - to test for C<EV_READ>, use C<< w->events & EV_READ >>, and
		1853	similarly for C<EV_WRITE>.
		1854
		1855	As with C<fd>, you must not modify this member even when the watcher is
		1856	stopped, always use C<ev_io_set> or C<ev_io_modify> for that.
746		1857
747	=back	1858	=back
748		1859
		1860	=head3 Examples
		1861
749	Example: call C<stdin_readable_cb> when STDIN_FILENO has become, well	1862	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
750	readable, but only once. Since it is likely line-buffered, you could	1863	readable, but only once. Since it is likely line-buffered, you could
751	attempt to read a whole line in the callback:	1864	attempt to read a whole line in the callback.
752		1865
753	static void	1866	static void
754	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1867	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
755	{	1868	{
756	ev_io_stop (loop, w);	1869	ev_io_stop (loop, w);
757	.. read from stdin here (or from w->fd) and haqndle any I/O errors	1870	.. read from stdin here (or from w->fd) and handle any I/O errors
758	}	1871	}
759		1872
760	...	1873	...
761	struct ev_loop *loop = ev_default_init (0);	1874	struct ev_loop *loop = ev_default_init (0);
762	struct ev_io stdin_readable;	1875	ev_io stdin_readable;
763	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);	1876	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
764	ev_io_start (loop, &stdin_readable);	1877	ev_io_start (loop, &stdin_readable);
765	ev_loop (loop, 0);	1878	ev_run (loop, 0);
766		1879
767		1880
768	=head2 C<ev_timer> - relative and optionally repeating timeouts	1881	=head2 C<ev_timer> - relative and optionally repeating timeouts
769		1882
770	Timer watchers are simple relative timers that generate an event after a	1883	Timer watchers are simple relative timers that generate an event after a
771	given time, and optionally repeating in regular intervals after that.	1884	given time, and optionally repeating in regular intervals after that.
772		1885
773	The timers are based on real time, that is, if you register an event that	1886	The timers are based on real time, that is, if you register an event that
774	times out after an hour and you reset your system clock to last years	1887	times out after an hour and you reset your system clock to January last
775	time, it will still time out after (roughly) and hour. "Roughly" because	1888	year, it will still time out after (roughly) one hour. "Roughly" because
776	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1889	detecting time jumps is hard, and some inaccuracies are unavoidable (the
777	monotonic clock option helps a lot here).	1890	monotonic clock option helps a lot here).
		1891
		1892	The callback is guaranteed to be invoked only I<after> its timeout has
		1893	passed (not I<at>, so on systems with very low-resolution clocks this
		1894	might introduce a small delay, see "the special problem of being too
		1895	early", below). If multiple timers become ready during the same loop
		1896	iteration then the ones with earlier time-out values are invoked before
		1897	ones of the same priority with later time-out values (but this is no
		1898	longer true when a callback calls C<ev_run> recursively).
		1899
		1900	=head3 Be smart about timeouts
		1901
		1902	Many real-world problems involve some kind of timeout, usually for error
		1903	recovery. A typical example is an HTTP request - if the other side hangs,
		1904	you want to raise some error after a while.
		1905
		1906	What follows are some ways to handle this problem, from obvious and
		1907	inefficient to smart and efficient.
		1908
		1909	In the following, a 60 second activity timeout is assumed - a timeout that
		1910	gets reset to 60 seconds each time there is activity (e.g. each time some
		1911	data or other life sign was received).
		1912
		1913	=over 4
		1914
		1915	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1916
		1917	This is the most obvious, but not the most simple way: In the beginning,
		1918	start the watcher:
		1919
		1920	ev_timer_init (timer, callback, 60., 0.);
		1921	ev_timer_start (loop, timer);
		1922
		1923	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1924	and start it again:
		1925
		1926	ev_timer_stop (loop, timer);
		1927	ev_timer_set (timer, 60., 0.);
		1928	ev_timer_start (loop, timer);
		1929
		1930	This is relatively simple to implement, but means that each time there is
		1931	some activity, libev will first have to remove the timer from its internal
		1932	data structure and then add it again. Libev tries to be fast, but it's
		1933	still not a constant-time operation.
		1934
		1935	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1936
		1937	This is the easiest way, and involves using C<ev_timer_again> instead of
		1938	C<ev_timer_start>.
		1939
		1940	To implement this, configure an C<ev_timer> with a C<repeat> value
		1941	of C<60> and then call C<ev_timer_again> at start and each time you
		1942	successfully read or write some data. If you go into an idle state where
		1943	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1944	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1945
		1946	That means you can ignore both the C<ev_timer_start> function and the
		1947	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1948	member and C<ev_timer_again>.
		1949
		1950	At start:
		1951
		1952	ev_init (timer, callback);
		1953	timer->repeat = 60.;
		1954	ev_timer_again (loop, timer);
		1955
		1956	Each time there is some activity:
		1957
		1958	ev_timer_again (loop, timer);
		1959
		1960	It is even possible to change the time-out on the fly, regardless of
		1961	whether the watcher is active or not:
		1962
		1963	timer->repeat = 30.;
		1964	ev_timer_again (loop, timer);
		1965
		1966	This is slightly more efficient then stopping/starting the timer each time
		1967	you want to modify its timeout value, as libev does not have to completely
		1968	remove and re-insert the timer from/into its internal data structure.
		1969
		1970	It is, however, even simpler than the "obvious" way to do it.
		1971
		1972	=item 3. Let the timer time out, but then re-arm it as required.
		1973
		1974	This method is more tricky, but usually most efficient: Most timeouts are
		1975	relatively long compared to the intervals between other activity - in
		1976	our example, within 60 seconds, there are usually many I/O events with
		1977	associated activity resets.
		1978
		1979	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1980	but remember the time of last activity, and check for a real timeout only
		1981	within the callback:
		1982
		1983	ev_tstamp timeout = 60.;
		1984	ev_tstamp last_activity; // time of last activity
		1985	ev_timer timer;
		1986
		1987	static void
		1988	callback (EV_P_ ev_timer *w, int revents)
		1989	{
		1990	// calculate when the timeout would happen
		1991	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
		1992
		1993	// if negative, it means we the timeout already occurred
		1994	if (after < 0.)
		1995	{
		1996	// timeout occurred, take action
		1997	}
		1998	else
		1999	{
		2000	// callback was invoked, but there was some recent
		2001	// activity. simply restart the timer to time out
		2002	// after "after" seconds, which is the earliest time
		2003	// the timeout can occur.
		2004	ev_timer_set (w, after, 0.);
		2005	ev_timer_start (EV_A_ w);
		2006	}
		2007	}
		2008
		2009	To summarise the callback: first calculate in how many seconds the
		2010	timeout will occur (by calculating the absolute time when it would occur,
		2011	C<last_activity + timeout>, and subtracting the current time, C<ev_now
		2012	(EV_A)> from that).
		2013
		2014	If this value is negative, then we are already past the timeout, i.e. we
		2015	timed out, and need to do whatever is needed in this case.
		2016
		2017	Otherwise, we now the earliest time at which the timeout would trigger,
		2018	and simply start the timer with this timeout value.
		2019
		2020	In other words, each time the callback is invoked it will check whether
		2021	the timeout occurred. If not, it will simply reschedule itself to check
		2022	again at the earliest time it could time out. Rinse. Repeat.
		2023
		2024	This scheme causes more callback invocations (about one every 60 seconds
		2025	minus half the average time between activity), but virtually no calls to
		2026	libev to change the timeout.
		2027
		2028	To start the machinery, simply initialise the watcher and set
		2029	C<last_activity> to the current time (meaning there was some activity just
		2030	now), then call the callback, which will "do the right thing" and start
		2031	the timer:
		2032
		2033	last_activity = ev_now (EV_A);
		2034	ev_init (&timer, callback);
		2035	callback (EV_A_ &timer, 0);
		2036
		2037	When there is some activity, simply store the current time in
		2038	C<last_activity>, no libev calls at all:
		2039
		2040	if (activity detected)
		2041	last_activity = ev_now (EV_A);
		2042
		2043	When your timeout value changes, then the timeout can be changed by simply
		2044	providing a new value, stopping the timer and calling the callback, which
		2045	will again do the right thing (for example, time out immediately :).
		2046
		2047	timeout = new_value;
		2048	ev_timer_stop (EV_A_ &timer);
		2049	callback (EV_A_ &timer, 0);
		2050
		2051	This technique is slightly more complex, but in most cases where the
		2052	time-out is unlikely to be triggered, much more efficient.
		2053
		2054	=item 4. Wee, just use a double-linked list for your timeouts.
		2055
		2056	If there is not one request, but many thousands (millions...), all
		2057	employing some kind of timeout with the same timeout value, then one can
		2058	do even better:
		2059
		2060	When starting the timeout, calculate the timeout value and put the timeout
		2061	at the I<end> of the list.
		2062
		2063	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		2064	the list is expected to fire (for example, using the technique #3).
		2065
		2066	When there is some activity, remove the timer from the list, recalculate
		2067	the timeout, append it to the end of the list again, and make sure to
		2068	update the C<ev_timer> if it was taken from the beginning of the list.
		2069
		2070	This way, one can manage an unlimited number of timeouts in O(1) time for
		2071	starting, stopping and updating the timers, at the expense of a major
		2072	complication, and having to use a constant timeout. The constant timeout
		2073	ensures that the list stays sorted.
		2074
		2075	=back
		2076
		2077	So which method the best?
		2078
		2079	Method #2 is a simple no-brain-required solution that is adequate in most
		2080	situations. Method #3 requires a bit more thinking, but handles many cases
		2081	better, and isn't very complicated either. In most case, choosing either
		2082	one is fine, with #3 being better in typical situations.
		2083
		2084	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		2085	rather complicated, but extremely efficient, something that really pays
		2086	off after the first million or so of active timers, i.e. it's usually
		2087	overkill :)
		2088
		2089	=head3 The special problem of being too early
		2090
		2091	If you ask a timer to call your callback after three seconds, then
		2092	you expect it to be invoked after three seconds - but of course, this
		2093	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		2094	guaranteed to any precision by libev - imagine somebody suspending the
		2095	process with a STOP signal for a few hours for example.
		2096
		2097	So, libev tries to invoke your callback as soon as possible I<after> the
		2098	delay has occurred, but cannot guarantee this.
		2099
		2100	A less obvious failure mode is calling your callback too early: many event
		2101	loops compare timestamps with a "elapsed delay >= requested delay", but
		2102	this can cause your callback to be invoked much earlier than you would
		2103	expect.
		2104
		2105	To see why, imagine a system with a clock that only offers full second
		2106	resolution (think windows if you can't come up with a broken enough OS
		2107	yourself). If you schedule a one-second timer at the time 500.9, then the
		2108	event loop will schedule your timeout to elapse at a system time of 500
		2109	(500.9 truncated to the resolution) + 1, or 501.
		2110
		2111	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2112	501" and invoke the callback 0.1s after it was started, even though a
		2113	one-second delay was requested - this is being "too early", despite best
		2114	intentions.
		2115
		2116	This is the reason why libev will never invoke the callback if the elapsed
		2117	delay equals the requested delay, but only when the elapsed delay is
		2118	larger than the requested delay. In the example above, libev would only invoke
		2119	the callback at system time 502, or 1.1s after the timer was started.
		2120
		2121	So, while libev cannot guarantee that your callback will be invoked
		2122	exactly when requested, it I<can> and I<does> guarantee that the requested
		2123	delay has actually elapsed, or in other words, it always errs on the "too
		2124	late" side of things.
		2125
		2126	=head3 The special problem of time updates
		2127
		2128	Establishing the current time is a costly operation (it usually takes
		2129	at least one system call): EV therefore updates its idea of the current
		2130	time only before and after C<ev_run> collects new events, which causes a
		2131	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		2132	lots of events in one iteration.
778		2133
779	The relative timeouts are calculated relative to the C<ev_now ()>	2134	The relative timeouts are calculated relative to the C<ev_now ()>
780	time. This is usually the right thing as this timestamp refers to the time	2135	time. This is usually the right thing as this timestamp refers to the time
781	of the event triggering whatever timeout you are modifying/starting. If	2136	of the event triggering whatever timeout you are modifying/starting. If
782	you suspect event processing to be delayed and you I<need> to base the timeout	2137	you suspect event processing to be delayed and you I<need> to base the
783	on the current time, use something like this to adjust for this:	2138	timeout on the current time, use something like the following to adjust
		2139	for it:
784		2140
785	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2141	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
786		2142
787	The callback is guarenteed to be invoked only when its timeout has passed,	2143	If the event loop is suspended for a long time, you can also force an
788	but if multiple timers become ready during the same loop iteration then	2144	update of the time returned by C<ev_now ()> by calling C<ev_now_update
789	order of execution is undefined.	2145	()>, although that will push the event time of all outstanding events
		2146	further into the future.
		2147
		2148	=head3 The special problem of unsynchronised clocks
		2149
		2150	Modern systems have a variety of clocks - libev itself uses the normal
		2151	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2152	jumps).
		2153
		2154	Neither of these clocks is synchronised with each other or any other clock
		2155	on the system, so C<ev_time ()> might return a considerably different time
		2156	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2157	a call to C<gettimeofday> might return a second count that is one higher
		2158	than a directly following call to C<time>.
		2159
		2160	The moral of this is to only compare libev-related timestamps with
		2161	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2162	a second or so.
		2163
		2164	One more problem arises due to this lack of synchronisation: if libev uses
		2165	the system monotonic clock and you compare timestamps from C<ev_time>
		2166	or C<ev_now> from when you started your timer and when your callback is
		2167	invoked, you will find that sometimes the callback is a bit "early".
		2168
		2169	This is because C<ev_timer>s work in real time, not wall clock time, so
		2170	libev makes sure your callback is not invoked before the delay happened,
		2171	I<measured according to the real time>, not the system clock.
		2172
		2173	If your timeouts are based on a physical timescale (e.g. "time out this
		2174	connection after 100 seconds") then this shouldn't bother you as it is
		2175	exactly the right behaviour.
		2176
		2177	If you want to compare wall clock/system timestamps to your timers, then
		2178	you need to use C<ev_periodic>s, as these are based on the wall clock
		2179	time, where your comparisons will always generate correct results.
		2180
		2181	=head3 The special problems of suspended animation
		2182
		2183	When you leave the server world it is quite customary to hit machines that
		2184	can suspend/hibernate - what happens to the clocks during such a suspend?
		2185
		2186	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		2187	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2188	to run until the system is suspended, but they will not advance while the
		2189	system is suspended. That means, on resume, it will be as if the program
		2190	was frozen for a few seconds, but the suspend time will not be counted
		2191	towards C<ev_timer> when a monotonic clock source is used. The real time
		2192	clock advanced as expected, but if it is used as sole clocksource, then a
		2193	long suspend would be detected as a time jump by libev, and timers would
		2194	be adjusted accordingly.
		2195
		2196	I would not be surprised to see different behaviour in different between
		2197	operating systems, OS versions or even different hardware.
		2198
		2199	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2200	time jump in the monotonic clocks and the realtime clock. If the program
		2201	is suspended for a very long time, and monotonic clock sources are in use,
		2202	then you can expect C<ev_timer>s to expire as the full suspension time
		2203	will be counted towards the timers. When no monotonic clock source is in
		2204	use, then libev will again assume a timejump and adjust accordingly.
		2205
		2206	It might be beneficial for this latter case to call C<ev_suspend>
		2207	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2208	deterministic behaviour in this case (you can do nothing against
		2209	C<SIGSTOP>).
		2210
		2211	=head3 Watcher-Specific Functions and Data Members
790		2212
791	=over 4	2213	=over 4
792		2214
793	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2215	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
794		2216
795	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2217	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
796		2218
797	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2219	Configure the timer to trigger after C<after> seconds (fractional and
798	C<0.>, then it will automatically be stopped. If it is positive, then the	2220	negative values are supported). If C<repeat> is C<0.>, then it will
		2221	automatically be stopped once the timeout is reached. If it is positive,
799	timer will automatically be configured to trigger again C<repeat> seconds	2222	then the timer will automatically be configured to trigger again C<repeat>
800	later, again, and again, until stopped manually.	2223	seconds later, again, and again, until stopped manually.
801		2224
802	The timer itself will do a best-effort at avoiding drift, that is, if you	2225	The timer itself will do a best-effort at avoiding drift, that is, if
803	configure a timer to trigger every 10 seconds, then it will trigger at	2226	you configure a timer to trigger every 10 seconds, then it will normally
804	exactly 10 second intervals. If, however, your program cannot keep up with	2227	trigger at exactly 10 second intervals. If, however, your program cannot
805	the timer (because it takes longer than those 10 seconds to do stuff) the	2228	keep up with the timer (because it takes longer than those 10 seconds to
806	timer will not fire more than once per event loop iteration.	2229	do stuff) the timer will not fire more than once per event loop iteration.
807		2230
808	=item ev_timer_again (loop)	2231	=item ev_timer_again (loop, ev_timer *)
809		2232
810	This will act as if the timer timed out and restart it again if it is	2233	This will act as if the timer timed out, and restarts it again if it is
811	repeating. The exact semantics are:	2234	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2235	timeout to the C<repeat> value and calling C<ev_timer_start>.
812		2236
		2237	The exact semantics are as in the following rules, all of which will be
		2238	applied to the watcher:
		2239
		2240	=over 4
		2241
		2242	=item If the timer is pending, the pending status is always cleared.
		2243
813	If the timer is started but nonrepeating, stop it.	2244	=item If the timer is started but non-repeating, stop it (as if it timed
		2245	out, without invoking it).
814		2246
815	If the timer is repeating, either start it if necessary (with the repeat	2247	=item If the timer is repeating, make the C<repeat> value the new timeout
816	value), or reset the running timer to the repeat value.	2248	and start the timer, if necessary.
817
818	This sounds a bit complicated, but here is a useful and typical
819	example: Imagine you have a tcp connection and you want a so-called idle
820	timeout, that is, you want to be called when there have been, say, 60
821	seconds of inactivity on the socket. The easiest way to do this is to
822	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each
823	time you successfully read or write some data. If you go into an idle
824	state where you do not expect data to travel on the socket, you can stop
825	the timer, and again will automatically restart it if need be.
826		2249
827	=back	2250	=back
828		2251
		2252	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
		2253	usage example.
		2254
		2255	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
		2256
		2257	Returns the remaining time until a timer fires. If the timer is active,
		2258	then this time is relative to the current event loop time, otherwise it's
		2259	the timeout value currently configured.
		2260
		2261	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2262	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2263	will return C<4>. When the timer expires and is restarted, it will return
		2264	roughly C<7> (likely slightly less as callback invocation takes some time,
		2265	too), and so on.
		2266
		2267	=item ev_tstamp repeat [read-write]
		2268
		2269	The current C<repeat> value. Will be used each time the watcher times out
		2270	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2271	which is also when any modifications are taken into account.
		2272
		2273	=back
		2274
		2275	=head3 Examples
		2276
829	Example: create a timer that fires after 60 seconds.	2277	Example: Create a timer that fires after 60 seconds.
830		2278
831	static void	2279	static void
832	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2280	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
833	{	2281	{
834	.. one minute over, w is actually stopped right here	2282	.. one minute over, w is actually stopped right here
835	}	2283	}
836		2284
837	struct ev_timer mytimer;	2285	ev_timer mytimer;
838	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);	2286	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
839	ev_timer_start (loop, &mytimer);	2287	ev_timer_start (loop, &mytimer);
840		2288
841	Example: create a timeout timer that times out after 10 seconds of	2289	Example: Create a timeout timer that times out after 10 seconds of
842	inactivity.	2290	inactivity.
843		2291
844	static void	2292	static void
845	timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2293	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
846	{	2294	{
847	.. ten seconds without any activity	2295	.. ten seconds without any activity
848	}	2296	}
849		2297
850	struct ev_timer mytimer;	2298	ev_timer mytimer;
851	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */	2299	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
852	ev_timer_again (&mytimer); /* start timer */	2300	ev_timer_again (&mytimer); /* start timer */
853	ev_loop (loop, 0);	2301	ev_run (loop, 0);
854		2302
855	// and in some piece of code that gets executed on any "activity":	2303	// and in some piece of code that gets executed on any "activity":
856	// reset the timeout to start ticking again at 10 seconds	2304	// reset the timeout to start ticking again at 10 seconds
857	ev_timer_again (&mytimer);	2305	ev_timer_again (&mytimer);
858		2306
859		2307
860	=head2 C<ev_periodic> - to cron or not to cron?	2308	=head2 C<ev_periodic> - to cron or not to cron?
861		2309
862	Periodic watchers are also timers of a kind, but they are very versatile	2310	Periodic watchers are also timers of a kind, but they are very versatile
863	(and unfortunately a bit complex).	2311	(and unfortunately a bit complex).
864		2312
865	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2313	Unlike C<ev_timer>, periodic watchers are not based on real time (or
866	but on wallclock time (absolute time). You can tell a periodic watcher	2314	relative time, the physical time that passes) but on wall clock time
867	to trigger "at" some specific point in time. For example, if you tell a	2315	(absolute time, the thing you can read on your calendar or clock). The
868	periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()	2316	difference is that wall clock time can run faster or slower than real
869	+ 10.>) and then reset your system clock to the last year, then it will	2317	time, and time jumps are not uncommon (e.g. when you adjust your
870	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2318	wrist-watch).
871	roughly 10 seconds later and of course not if you reset your system time
872	again).
873		2319
874	They can also be used to implement vastly more complex timers, such as	2320	You can tell a periodic watcher to trigger after some specific point
		2321	in time: for example, if you tell a periodic watcher to trigger "in 10
		2322	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2323	not a delay) and then reset your system clock to January of the previous
		2324	year, then it will take a year or more to trigger the event (unlike an
		2325	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2326	it, as it uses a relative timeout).
		2327
		2328	C<ev_periodic> watchers can also be used to implement vastly more complex
875	triggering an event on eahc midnight, local time.	2329	timers, such as triggering an event on each "midnight, local time", or
		2330	other complicated rules. This cannot easily be done with C<ev_timer>
		2331	watchers, as those cannot react to time jumps.
876		2332
877	As with timers, the callback is guarenteed to be invoked only when the	2333	As with timers, the callback is guaranteed to be invoked only when the
878	time (C<at>) has been passed, but if multiple periodic timers become ready	2334	point in time where it is supposed to trigger has passed. If multiple
879	during the same loop iteration then order of execution is undefined.	2335	timers become ready during the same loop iteration then the ones with
		2336	earlier time-out values are invoked before ones with later time-out values
		2337	(but this is no longer true when a callback calls C<ev_run> recursively).
		2338
		2339	=head3 Watcher-Specific Functions and Data Members
880		2340
881	=over 4	2341	=over 4
882		2342
883	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2343	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
884		2344
885	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2345	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
886		2346
887	Lots of arguments, lets sort it out... There are basically three modes of	2347	Lots of arguments, let's sort it out... There are basically three modes of
888	operation, and we will explain them from simplest to complex:	2348	operation, and we will explain them from simplest to most complex:
889		2349
890	=over 4	2350	=over 4
891		2351
892	=item * absolute timer (interval = reschedule_cb = 0)	2352	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
893		2353
894	In this configuration the watcher triggers an event at the wallclock time	2354	In this configuration the watcher triggers an event after the wall clock
895	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2355	time C<offset> has passed. It will not repeat and will not adjust when a
896	that is, if it is to be run at January 1st 2011 then it will run when the	2356	time jump occurs, that is, if it is to be run at January 1st 2011 then it
897	system time reaches or surpasses this time.	2357	will be stopped and invoked when the system clock reaches or surpasses
		2358	this point in time.
898		2359
899	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2360	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
900		2361
901	In this mode the watcher will always be scheduled to time out at the next	2362	In this mode the watcher will always be scheduled to time out at the next
902	C<at + N * interval> time (for some integer N) and then repeat, regardless	2363	C<offset + N * interval> time (for some integer N, which can also be
903	of any time jumps.	2364	negative) and then repeat, regardless of any time jumps. The C<offset>
		2365	argument is merely an offset into the C<interval> periods.
904		2366
905	This can be used to create timers that do not drift with respect to system	2367	This can be used to create timers that do not drift with respect to the
906	time:	2368	system clock, for example, here is an C<ev_periodic> that triggers each
		2369	hour, on the hour (with respect to UTC):
907		2370
908	ev_periodic_set (&periodic, 0., 3600., 0);	2371	ev_periodic_set (&periodic, 0., 3600., 0);
909		2372
910	This doesn't mean there will always be 3600 seconds in between triggers,	2373	This doesn't mean there will always be 3600 seconds in between triggers,
911	but only that the the callback will be called when the system time shows a	2374	but only that the callback will be called when the system time shows a
912	full hour (UTC), or more correctly, when the system time is evenly divisible	2375	full hour (UTC), or more correctly, when the system time is evenly divisible
913	by 3600.	2376	by 3600.
914		2377
915	Another way to think about it (for the mathematically inclined) is that	2378	Another way to think about it (for the mathematically inclined) is that
916	C<ev_periodic> will try to run the callback in this mode at the next possible	2379	C<ev_periodic> will try to run the callback in this mode at the next possible
917	time where C<time = at (mod interval)>, regardless of any time jumps.	2380	time where C<time = offset (mod interval)>, regardless of any time jumps.
918		2381
		2382	The C<interval> I<MUST> be positive, and for numerical stability, the
		2383	interval value should be higher than C<1/8192> (which is around 100
		2384	microseconds) and C<offset> should be higher than C<0> and should have
		2385	at most a similar magnitude as the current time (say, within a factor of
		2386	ten). Typical values for offset are, in fact, C<0> or something between
		2387	C<0> and C<interval>, which is also the recommended range.
		2388
		2389	Note also that there is an upper limit to how often a timer can fire (CPU
		2390	speed for example), so if C<interval> is very small then timing stability
		2391	will of course deteriorate. Libev itself tries to be exact to be about one
		2392	millisecond (if the OS supports it and the machine is fast enough).
		2393
919	=item * manual reschedule mode (reschedule_cb = callback)	2394	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
920		2395
921	In this mode the values for C<interval> and C<at> are both being	2396	In this mode the values for C<interval> and C<offset> are both being
922	ignored. Instead, each time the periodic watcher gets scheduled, the	2397	ignored. Instead, each time the periodic watcher gets scheduled, the
923	reschedule callback will be called with the watcher as first, and the	2398	reschedule callback will be called with the watcher as first, and the
924	current time as second argument.	2399	current time as second argument.
925		2400
926	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2401	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
927	ever, or make any event loop modifications>. If you need to stop it,	2402	or make ANY other event loop modifications whatsoever, unless explicitly
928	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2403	allowed by documentation here>.
929	starting a prepare watcher).
930		2404
		2405	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2406	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2407	only event loop modification you are allowed to do).
		2408
931	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2409	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
932	ev_tstamp now)>, e.g.:	2410	*w, ev_tstamp now)>, e.g.:
933		2411
		2412	static ev_tstamp
934	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2413	my_rescheduler (ev_periodic *w, ev_tstamp now)
935	{	2414	{
936	return now + 60.;	2415	return now + 60.;
937	}	2416	}
938		2417
939	It must return the next time to trigger, based on the passed time value	2418	It must return the next time to trigger, based on the passed time value
940	(that is, the lowest time value larger than to the second argument). It	2419	(that is, the lowest time value larger than to the second argument). It
941	will usually be called just before the callback will be triggered, but	2420	will usually be called just before the callback will be triggered, but
942	might be called at other times, too.	2421	might be called at other times, too.
943		2422
944	NOTE: I<< This callback must always return a time that is later than the	2423	NOTE: I<< This callback must always return a time that is higher than or
945	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2424	equal to the passed C<now> value >>.
946		2425
947	This can be used to create very complex timers, such as a timer that	2426	This can be used to create very complex timers, such as a timer that
948	triggers on each midnight, local time. To do this, you would calculate the	2427	triggers on "next midnight, local time". To do this, you would calculate
949	next midnight after C<now> and return the timestamp value for this. How	2428	the next midnight after C<now> and return the timestamp value for
950	you do this is, again, up to you (but it is not trivial, which is the main	2429	this. Here is a (completely untested, no error checking) example on how to
951	reason I omitted it as an example).	2430	do this:
		2431
		2432	#include <time.h>
		2433
		2434	static ev_tstamp
		2435	my_rescheduler (ev_periodic *w, ev_tstamp now)
		2436	{
		2437	time_t tnow = (time_t)now;
		2438	struct tm tm;
		2439	localtime_r (&tnow, &tm);
		2440
		2441	tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
		2442	++tm.tm_mday; // midnight next day
		2443
		2444	return mktime (&tm);
		2445	}
		2446
		2447	Note: this code might run into trouble on days that have more then two
		2448	midnights (beginning and end).
952		2449
953	=back	2450	=back
954		2451
955	=item ev_periodic_again (loop, ev_periodic *)	2452	=item ev_periodic_again (loop, ev_periodic *)
956		2453
957	Simply stops and restarts the periodic watcher again. This is only useful	2454	Simply stops and restarts the periodic watcher again. This is only useful
958	when you changed some parameters or the reschedule callback would return	2455	when you changed some parameters or the reschedule callback would return
959	a different time than the last time it was called (e.g. in a crond like	2456	a different time than the last time it was called (e.g. in a crond like
960	program when the crontabs have changed).	2457	program when the crontabs have changed).
961		2458
		2459	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2460
		2461	When active, returns the absolute time that the watcher is supposed
		2462	to trigger next. This is not the same as the C<offset> argument to
		2463	C<ev_periodic_set>, but indeed works even in interval and manual
		2464	rescheduling modes.
		2465
		2466	=item ev_tstamp offset [read-write]
		2467
		2468	When repeating, this contains the offset value, otherwise this is the
		2469	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2470	although libev might modify this value for better numerical stability).
		2471
		2472	Can be modified any time, but changes only take effect when the periodic
		2473	timer fires or C<ev_periodic_again> is being called.
		2474
		2475	=item ev_tstamp interval [read-write]
		2476
		2477	The current interval value. Can be modified any time, but changes only
		2478	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2479	called.
		2480
		2481	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2482
		2483	The current reschedule callback, or C<0>, if this functionality is
		2484	switched off. Can be changed any time, but changes only take effect when
		2485	the periodic timer fires or C<ev_periodic_again> is being called.
		2486
962	=back	2487	=back
963		2488
		2489	=head3 Examples
		2490
964	Example: call a callback every hour, or, more precisely, whenever the	2491	Example: Call a callback every hour, or, more precisely, whenever the
965	system clock is divisible by 3600. The callback invocation times have	2492	system time is divisible by 3600. The callback invocation times have
966	potentially a lot of jittering, but good long-term stability.	2493	potentially a lot of jitter, but good long-term stability.
967		2494
968	static void	2495	static void
969	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)	2496	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
970	{	2497	{
971	... its now a full hour (UTC, or TAI or whatever your clock follows)	2498	... its now a full hour (UTC, or TAI or whatever your clock follows)
972	}	2499	}
973		2500
974	struct ev_periodic hourly_tick;	2501	ev_periodic hourly_tick;
975	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);	2502	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
976	ev_periodic_start (loop, &hourly_tick);	2503	ev_periodic_start (loop, &hourly_tick);
977		2504
978	Example: the same as above, but use a reschedule callback to do it:	2505	Example: The same as above, but use a reschedule callback to do it:
979		2506
980	#include <math.h>	2507	#include <math.h>
981		2508
982	static ev_tstamp	2509	static ev_tstamp
983	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)	2510	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
984	{	2511	{
985	return fmod (now, 3600.) + 3600.;	2512	return now + (3600. - fmod (now, 3600.));
986	}	2513	}
987		2514
988	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);	2515	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
989		2516
990	Example: call a callback every hour, starting now:	2517	Example: Call a callback every hour, starting now:
991		2518
992	struct ev_periodic hourly_tick;	2519	ev_periodic hourly_tick;
993	ev_periodic_init (&hourly_tick, clock_cb,	2520	ev_periodic_init (&hourly_tick, clock_cb,
994	fmod (ev_now (loop), 3600.), 3600., 0);	2521	fmod (ev_now (loop), 3600.), 3600., 0);
995	ev_periodic_start (loop, &hourly_tick);	2522	ev_periodic_start (loop, &hourly_tick);
996		2523
997		2524
998	=head2 C<ev_signal> - signal me when a signal gets signalled!	2525	=head2 C<ev_signal> - signal me when a signal gets signalled!
999		2526
1000	Signal watchers will trigger an event when the process receives a specific	2527	Signal watchers will trigger an event when the process receives a specific
1001	signal one or more times. Even though signals are very asynchronous, libev	2528	signal one or more times. Even though signals are very asynchronous, libev
1002	will try it's best to deliver signals synchronously, i.e. as part of the	2529	will try its best to deliver signals synchronously, i.e. as part of the
1003	normal event processing, like any other event.	2530	normal event processing, like any other event.
1004		2531
		2532	If you want signals to be delivered truly asynchronously, just use
		2533	C<sigaction> as you would do without libev and forget about sharing
		2534	the signal. You can even use C<ev_async> from a signal handler to
		2535	synchronously wake up an event loop.
		2536
1005	You can configure as many watchers as you like per signal. Only when the	2537	You can configure as many watchers as you like for the same signal, but
1006	first watcher gets started will libev actually register a signal watcher	2538	only within the same loop, i.e. you can watch for C<SIGINT> in your
1007	with the kernel (thus it coexists with your own signal handlers as long	2539	default loop and for C<SIGIO> in another loop, but you cannot watch for
1008	as you don't register any with libev). Similarly, when the last signal	2540	C<SIGINT> in both the default loop and another loop at the same time. At
1009	watcher for a signal is stopped libev will reset the signal handler to	2541	the moment, C<SIGCHLD> is permanently tied to the default loop.
1010	SIG_DFL (regardless of what it was set to before).	2542
		2543	Only after the first watcher for a signal is started will libev actually
		2544	register something with the kernel. It thus coexists with your own signal
		2545	handlers as long as you don't register any with libev for the same signal.
		2546
		2547	If possible and supported, libev will install its handlers with
		2548	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2549	not be unduly interrupted. If you have a problem with system calls getting
		2550	interrupted by signals you can block all signals in an C<ev_check> watcher
		2551	and unblock them in an C<ev_prepare> watcher.
		2552
		2553	=head3 The special problem of inheritance over fork/execve/pthread_create
		2554
		2555	Both the signal mask (C<sigprocmask>) and the signal disposition
		2556	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2557	stopping it again), that is, libev might or might not block the signal,
		2558	and might or might not set or restore the installed signal handler (but
		2559	see C<EVFLAG_NOSIGMASK>).
		2560
		2561	While this does not matter for the signal disposition (libev never
		2562	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2563	C<execve>), this matters for the signal mask: many programs do not expect
		2564	certain signals to be blocked.
		2565
		2566	This means that before calling C<exec> (from the child) you should reset
		2567	the signal mask to whatever "default" you expect (all clear is a good
		2568	choice usually).
		2569
		2570	The simplest way to ensure that the signal mask is reset in the child is
		2571	to install a fork handler with C<pthread_atfork> that resets it. That will
		2572	catch fork calls done by libraries (such as the libc) as well.
		2573
		2574	In current versions of libev, the signal will not be blocked indefinitely
		2575	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2576	the window of opportunity for problems, it will not go away, as libev
		2577	I<has> to modify the signal mask, at least temporarily.
		2578
		2579	So I can't stress this enough: I<If you do not reset your signal mask when
		2580	you expect it to be empty, you have a race condition in your code>. This
		2581	is not a libev-specific thing, this is true for most event libraries.
		2582
		2583	=head3 The special problem of threads signal handling
		2584
		2585	POSIX threads has problematic signal handling semantics, specifically,
		2586	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2587	threads in a process block signals, which is hard to achieve.
		2588
		2589	When you want to use sigwait (or mix libev signal handling with your own
		2590	for the same signals), you can tackle this problem by globally blocking
		2591	all signals before creating any threads (or creating them with a fully set
		2592	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2593	loops. Then designate one thread as "signal receiver thread" which handles
		2594	these signals. You can pass on any signals that libev might be interested
		2595	in by calling C<ev_feed_signal>.
		2596
		2597	=head3 Watcher-Specific Functions and Data Members
1011		2598
1012	=over 4	2599	=over 4
1013		2600
1014	=item ev_signal_init (ev_signal *, callback, int signum)	2601	=item ev_signal_init (ev_signal *, callback, int signum)
1015		2602
1016	=item ev_signal_set (ev_signal *, int signum)	2603	=item ev_signal_set (ev_signal *, int signum)
1017		2604
1018	Configures the watcher to trigger on the given signal number (usually one	2605	Configures the watcher to trigger on the given signal number (usually one
1019	of the C<SIGxxx> constants).	2606	of the C<SIGxxx> constants).
1020		2607
		2608	=item int signum [read-only]
		2609
		2610	The signal the watcher watches out for.
		2611
1021	=back	2612	=back
1022		2613
		2614	=head3 Examples
		2615
		2616	Example: Try to exit cleanly on SIGINT.
		2617
		2618	static void
		2619	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2620	{
		2621	ev_break (loop, EVBREAK_ALL);
		2622	}
		2623
		2624	ev_signal signal_watcher;
		2625	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2626	ev_signal_start (loop, &signal_watcher);
		2627
1023		2628
1024	=head2 C<ev_child> - watch out for process status changes	2629	=head2 C<ev_child> - watch out for process status changes
1025		2630
1026	Child watchers trigger when your process receives a SIGCHLD in response to	2631	Child watchers trigger when your process receives a SIGCHLD in response to
1027	some child status changes (most typically when a child of yours dies).	2632	some child status changes (most typically when a child of yours dies or
		2633	exits). It is permissible to install a child watcher I<after> the child
		2634	has been forked (which implies it might have already exited), as long
		2635	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2636	forking and then immediately registering a watcher for the child is fine,
		2637	but forking and registering a watcher a few event loop iterations later or
		2638	in the next callback invocation is not.
		2639
		2640	Only the default event loop is capable of handling signals, and therefore
		2641	you can only register child watchers in the default event loop.
		2642
		2643	Due to some design glitches inside libev, child watchers will always be
		2644	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2645	libev)
		2646
		2647	=head3 Process Interaction
		2648
		2649	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2650	initialised. This is necessary to guarantee proper behaviour even if the
		2651	first child watcher is started after the child exits. The occurrence
		2652	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2653	synchronously as part of the event loop processing. Libev always reaps all
		2654	children, even ones not watched.
		2655
		2656	=head3 Overriding the Built-In Processing
		2657
		2658	Libev offers no special support for overriding the built-in child
		2659	processing, but if your application collides with libev's default child
		2660	handler, you can override it easily by installing your own handler for
		2661	C<SIGCHLD> after initialising the default loop, and making sure the
		2662	default loop never gets destroyed. You are encouraged, however, to use an
		2663	event-based approach to child reaping and thus use libev's support for
		2664	that, so other libev users can use C<ev_child> watchers freely.
		2665
		2666	=head3 Stopping the Child Watcher
		2667
		2668	Currently, the child watcher never gets stopped, even when the
		2669	child terminates, so normally one needs to stop the watcher in the
		2670	callback. Future versions of libev might stop the watcher automatically
		2671	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2672	problem).
		2673
		2674	=head3 Watcher-Specific Functions and Data Members
1028		2675
1029	=over 4	2676	=over 4
1030		2677
1031	=item ev_child_init (ev_child *, callback, int pid)	2678	=item ev_child_init (ev_child *, callback, int pid, int trace)
1032		2679
1033	=item ev_child_set (ev_child *, int pid)	2680	=item ev_child_set (ev_child *, int pid, int trace)
1034		2681
1035	Configures the watcher to wait for status changes of process C<pid> (or	2682	Configures the watcher to wait for status changes of process C<pid> (or
1036	I<any> process if C<pid> is specified as C<0>). The callback can look	2683	I<any> process if C<pid> is specified as C<0>). The callback can look
1037	at the C<rstatus> member of the C<ev_child> watcher structure to see	2684	at the C<rstatus> member of the C<ev_child> watcher structure to see
1038	the status word (use the macros from C<sys/wait.h> and see your systems	2685	the status word (use the macros from C<sys/wait.h> and see your systems
1039	C<waitpid> documentation). The C<rpid> member contains the pid of the	2686	C<waitpid> documentation). The C<rpid> member contains the pid of the
1040	process causing the status change.	2687	process causing the status change. C<trace> must be either C<0> (only
		2688	activate the watcher when the process terminates) or C<1> (additionally
		2689	activate the watcher when the process is stopped or continued).
		2690
		2691	=item int pid [read-only]
		2692
		2693	The process id this watcher watches out for, or C<0>, meaning any process id.
		2694
		2695	=item int rpid [read-write]
		2696
		2697	The process id that detected a status change.
		2698
		2699	=item int rstatus [read-write]
		2700
		2701	The process exit/trace status caused by C<rpid> (see your systems
		2702	C<waitpid> and C<sys/wait.h> documentation for details).
1041		2703
1042	=back	2704	=back
1043		2705
1044	Example: try to exit cleanly on SIGINT and SIGTERM.	2706	=head3 Examples
1045		2707
		2708	Example: C<fork()> a new process and install a child handler to wait for
		2709	its completion.
		2710
		2711	ev_child cw;
		2712
1046	static void	2713	static void
1047	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)	2714	child_cb (EV_P_ ev_child *w, int revents)
1048	{	2715	{
1049	ev_unloop (loop, EVUNLOOP_ALL);	2716	ev_child_stop (EV_A_ w);
		2717	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1050	}	2718	}
1051		2719
1052	struct ev_signal signal_watcher;	2720	pid_t pid = fork ();
1053	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	2721
1054	ev_signal_start (loop, &sigint_cb);	2722	if (pid < 0)
		2723	// error
		2724	else if (pid == 0)
		2725	{
		2726	// the forked child executes here
		2727	exit (1);
		2728	}
		2729	else
		2730	{
		2731	ev_child_init (&cw, child_cb, pid, 0);
		2732	ev_child_start (EV_DEFAULT_ &cw);
		2733	}
		2734
		2735
		2736	=head2 C<ev_stat> - did the file attributes just change?
		2737
		2738	This watches a file system path for attribute changes. That is, it calls
		2739	C<stat> on that path in regular intervals (or when the OS says it changed)
		2740	and sees if it changed compared to the last time, invoking the callback
		2741	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2742	happen after the watcher has been started will be reported.
		2743
		2744	The path does not need to exist: changing from "path exists" to "path does
		2745	not exist" is a status change like any other. The condition "path does not
		2746	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2747	C<st_nlink> field being zero (which is otherwise always forced to be at
		2748	least one) and all the other fields of the stat buffer having unspecified
		2749	contents.
		2750
		2751	The path I<must not> end in a slash or contain special components such as
		2752	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2753	your working directory changes, then the behaviour is undefined.
		2754
		2755	Since there is no portable change notification interface available, the
		2756	portable implementation simply calls C<stat(2)> regularly on the path
		2757	to see if it changed somehow. You can specify a recommended polling
		2758	interval for this case. If you specify a polling interval of C<0> (highly
		2759	recommended!) then a I<suitable, unspecified default> value will be used
		2760	(which you can expect to be around five seconds, although this might
		2761	change dynamically). Libev will also impose a minimum interval which is
		2762	currently around C<0.1>, but that's usually overkill.
		2763
		2764	This watcher type is not meant for massive numbers of stat watchers,
		2765	as even with OS-supported change notifications, this can be
		2766	resource-intensive.
		2767
		2768	At the time of this writing, the only OS-specific interface implemented
		2769	is the Linux inotify interface (implementing kqueue support is left as an
		2770	exercise for the reader. Note, however, that the author sees no way of
		2771	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2772
		2773	=head3 ABI Issues (Largefile Support)
		2774
		2775	Libev by default (unless the user overrides this) uses the default
		2776	compilation environment, which means that on systems with large file
		2777	support disabled by default, you get the 32 bit version of the stat
		2778	structure. When using the library from programs that change the ABI to
		2779	use 64 bit file offsets the programs will fail. In that case you have to
		2780	compile libev with the same flags to get binary compatibility. This is
		2781	obviously the case with any flags that change the ABI, but the problem is
		2782	most noticeably displayed with ev_stat and large file support.
		2783
		2784	The solution for this is to lobby your distribution maker to make large
		2785	file interfaces available by default (as e.g. FreeBSD does) and not
		2786	optional. Libev cannot simply switch on large file support because it has
		2787	to exchange stat structures with application programs compiled using the
		2788	default compilation environment.
		2789
		2790	=head3 Inotify and Kqueue
		2791
		2792	When C<inotify (7)> support has been compiled into libev and present at
		2793	runtime, it will be used to speed up change detection where possible. The
		2794	inotify descriptor will be created lazily when the first C<ev_stat>
		2795	watcher is being started.
		2796
		2797	Inotify presence does not change the semantics of C<ev_stat> watchers
		2798	except that changes might be detected earlier, and in some cases, to avoid
		2799	making regular C<stat> calls. Even in the presence of inotify support
		2800	there are many cases where libev has to resort to regular C<stat> polling,
		2801	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2802	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2803	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2804	xfs are fully working) libev usually gets away without polling.
		2805
		2806	There is no support for kqueue, as apparently it cannot be used to
		2807	implement this functionality, due to the requirement of having a file
		2808	descriptor open on the object at all times, and detecting renames, unlinks
		2809	etc. is difficult.
		2810
		2811	=head3 C<stat ()> is a synchronous operation
		2812
		2813	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2814	the process. The exception are C<ev_stat> watchers - those call C<stat
		2815	()>, which is a synchronous operation.
		2816
		2817	For local paths, this usually doesn't matter: unless the system is very
		2818	busy or the intervals between stat's are large, a stat call will be fast,
		2819	as the path data is usually in memory already (except when starting the
		2820	watcher).
		2821
		2822	For networked file systems, calling C<stat ()> can block an indefinite
		2823	time due to network issues, and even under good conditions, a stat call
		2824	often takes multiple milliseconds.
		2825
		2826	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2827	paths, although this is fully supported by libev.
		2828
		2829	=head3 The special problem of stat time resolution
		2830
		2831	The C<stat ()> system call only supports full-second resolution portably,
		2832	and even on systems where the resolution is higher, most file systems
		2833	still only support whole seconds.
		2834
		2835	That means that, if the time is the only thing that changes, you can
		2836	easily miss updates: on the first update, C<ev_stat> detects a change and
		2837	calls your callback, which does something. When there is another update
		2838	within the same second, C<ev_stat> will be unable to detect unless the
		2839	stat data does change in other ways (e.g. file size).
		2840
		2841	The solution to this is to delay acting on a change for slightly more
		2842	than a second (or till slightly after the next full second boundary), using
		2843	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2844	ev_timer_again (loop, w)>).
		2845
		2846	The C<.02> offset is added to work around small timing inconsistencies
		2847	of some operating systems (where the second counter of the current time
		2848	might be be delayed. One such system is the Linux kernel, where a call to
		2849	C<gettimeofday> might return a timestamp with a full second later than
		2850	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2851	update file times then there will be a small window where the kernel uses
		2852	the previous second to update file times but libev might already execute
		2853	the timer callback).
		2854
		2855	=head3 Watcher-Specific Functions and Data Members
		2856
		2857	=over 4
		2858
		2859	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2860
		2861	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2862
		2863	Configures the watcher to wait for status changes of the given
		2864	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2865	be detected and should normally be specified as C<0> to let libev choose
		2866	a suitable value. The memory pointed to by C<path> must point to the same
		2867	path for as long as the watcher is active.
		2868
		2869	The callback will receive an C<EV_STAT> event when a change was detected,
		2870	relative to the attributes at the time the watcher was started (or the
		2871	last change was detected).
		2872
		2873	=item ev_stat_stat (loop, ev_stat *)
		2874
		2875	Updates the stat buffer immediately with new values. If you change the
		2876	watched path in your callback, you could call this function to avoid
		2877	detecting this change (while introducing a race condition if you are not
		2878	the only one changing the path). Can also be useful simply to find out the
		2879	new values.
		2880
		2881	=item ev_statdata attr [read-only]
		2882
		2883	The most-recently detected attributes of the file. Although the type is
		2884	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2885	suitable for your system, but you can only rely on the POSIX-standardised
		2886	members to be present. If the C<st_nlink> member is C<0>, then there was
		2887	some error while C<stat>ing the file.
		2888
		2889	=item ev_statdata prev [read-only]
		2890
		2891	The previous attributes of the file. The callback gets invoked whenever
		2892	C<prev> != C<attr>, or, more precisely, one or more of these members
		2893	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2894	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2895
		2896	=item ev_tstamp interval [read-only]
		2897
		2898	The specified interval.
		2899
		2900	=item const char *path [read-only]
		2901
		2902	The file system path that is being watched.
		2903
		2904	=back
		2905
		2906	=head3 Examples
		2907
		2908	Example: Watch C</etc/passwd> for attribute changes.
		2909
		2910	static void
		2911	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2912	{
		2913	/* /etc/passwd changed in some way */
		2914	if (w->attr.st_nlink)
		2915	{
		2916	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2917	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2918	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2919	}
		2920	else
		2921	/* you shalt not abuse printf for puts */
		2922	puts ("wow, /etc/passwd is not there, expect problems. "
		2923	"if this is windows, they already arrived\n");
		2924	}
		2925
		2926	...
		2927	ev_stat passwd;
		2928
		2929	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2930	ev_stat_start (loop, &passwd);
		2931
		2932	Example: Like above, but additionally use a one-second delay so we do not
		2933	miss updates (however, frequent updates will delay processing, too, so
		2934	one might do the work both on C<ev_stat> callback invocation I<and> on
		2935	C<ev_timer> callback invocation).
		2936
		2937	static ev_stat passwd;
		2938	static ev_timer timer;
		2939
		2940	static void
		2941	timer_cb (EV_P_ ev_timer *w, int revents)
		2942	{
		2943	ev_timer_stop (EV_A_ w);
		2944
		2945	/* now it's one second after the most recent passwd change */
		2946	}
		2947
		2948	static void
		2949	stat_cb (EV_P_ ev_stat *w, int revents)
		2950	{
		2951	/* reset the one-second timer */
		2952	ev_timer_again (EV_A_ &timer);
		2953	}
		2954
		2955	...
		2956	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2957	ev_stat_start (loop, &passwd);
		2958	ev_timer_init (&timer, timer_cb, 0., 1.02);
1055		2959
1056		2960
1057	=head2 C<ev_idle> - when you've got nothing better to do...	2961	=head2 C<ev_idle> - when you've got nothing better to do...
1058		2962
1059	Idle watchers trigger events when there are no other events are pending	2963	Idle watchers trigger events when no other events of the same or higher
1060	(prepare, check and other idle watchers do not count). That is, as long	2964	priority are pending (prepare, check and other idle watchers do not count
1061	as your process is busy handling sockets or timeouts (or even signals,	2965	as receiving "events").
1062	imagine) it will not be triggered. But when your process is idle all idle	2966
1063	watchers are being called again and again, once per event loop iteration -	2967	That is, as long as your process is busy handling sockets or timeouts
		2968	(or even signals, imagine) of the same or higher priority it will not be
		2969	triggered. But when your process is idle (or only lower-priority watchers
		2970	are pending), the idle watchers are being called once per event loop
1064	until stopped, that is, or your process receives more events and becomes	2971	iteration - until stopped, that is, or your process receives more events
1065	busy.	2972	and becomes busy again with higher priority stuff.
1066		2973
1067	The most noteworthy effect is that as long as any idle watchers are	2974	The most noteworthy effect is that as long as any idle watchers are
1068	active, the process will not block when waiting for new events.	2975	active, the process will not block when waiting for new events.
1069		2976
1070	Apart from keeping your process non-blocking (which is a useful	2977	Apart from keeping your process non-blocking (which is a useful
1071	effect on its own sometimes), idle watchers are a good place to do	2978	effect on its own sometimes), idle watchers are a good place to do
1072	"pseudo-background processing", or delay processing stuff to after the	2979	"pseudo-background processing", or delay processing stuff to after the
1073	event loop has handled all outstanding events.	2980	event loop has handled all outstanding events.
1074		2981
		2982	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2983
		2984	As long as there is at least one active idle watcher, libev will never
		2985	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2986	For this to work, the idle watcher doesn't need to be invoked at all - the
		2987	lowest priority will do.
		2988
		2989	This mode of operation can be useful together with an C<ev_check> watcher,
		2990	to do something on each event loop iteration - for example to balance load
		2991	between different connections.
		2992
		2993	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2994	example.
		2995
		2996	=head3 Watcher-Specific Functions and Data Members
		2997
1075	=over 4	2998	=over 4
1076		2999
1077	=item ev_idle_init (ev_signal *, callback)	3000	=item ev_idle_init (ev_idle *, callback)
1078		3001
1079	Initialises and configures the idle watcher - it has no parameters of any	3002	Initialises and configures the idle watcher - it has no parameters of any
1080	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	3003	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1081	believe me.	3004	believe me.
1082		3005
1083	=back	3006	=back
1084		3007
		3008	=head3 Examples
		3009
1085	Example: dynamically allocate an C<ev_idle>, start it, and in the	3010	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1086	callback, free it. Alos, use no error checking, as usual.	3011	callback, free it. Also, use no error checking, as usual.
1087		3012
1088	static void	3013	static void
1089	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)	3014	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
1090	{	3015	{
		3016	// stop the watcher
		3017	ev_idle_stop (loop, w);
		3018
		3019	// now we can free it
1091	free (w);	3020	free (w);
		3021
1092	// now do something you wanted to do when the program has	3022	// now do something you wanted to do when the program has
1093	// no longer asnything immediate to do.	3023	// no longer anything immediate to do.
1094	}	3024	}
1095		3025
1096	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));	3026	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
1097	ev_idle_init (idle_watcher, idle_cb);	3027	ev_idle_init (idle_watcher, idle_cb);
1098	ev_idle_start (loop, idle_cb);	3028	ev_idle_start (loop, idle_watcher);
1099		3029
1100		3030
1101	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	3031	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1102		3032
1103	Prepare and check watchers are usually (but not always) used in tandem:	3033	Prepare and check watchers are often (but not always) used in pairs:
1104	prepare watchers get invoked before the process blocks and check watchers	3034	prepare watchers get invoked before the process blocks and check watchers
1105	afterwards.	3035	afterwards.
1106		3036
		3037	You I<must not> call C<ev_run> (or similar functions that enter the
		3038	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
		3039	C<ev_check> watchers. Other loops than the current one are fine,
		3040	however. The rationale behind this is that you do not need to check
		3041	for recursion in those watchers, i.e. the sequence will always be
		3042	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
		3043	kind they will always be called in pairs bracketing the blocking call.
		3044
1107	Their main purpose is to integrate other event mechanisms into libev and	3045	Their main purpose is to integrate other event mechanisms into libev and
1108	their use is somewhat advanced. This could be used, for example, to track	3046	their use is somewhat advanced. They could be used, for example, to track
1109	variable changes, implement your own watchers, integrate net-snmp or a	3047	variable changes, implement your own watchers, integrate net-snmp or a
1110	coroutine library and lots more.	3048	coroutine library and lots more. They are also occasionally useful if
		3049	you cache some data and want to flush it before blocking (for example,
		3050	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		3051	watcher).
1111		3052
1112	This is done by examining in each prepare call which file descriptors need	3053	This is done by examining in each prepare call which file descriptors
1113	to be watched by the other library, registering C<ev_io> watchers for	3054	need to be watched by the other library, registering C<ev_io> watchers
1114	them and starting an C<ev_timer> watcher for any timeouts (many libraries	3055	for them and starting an C<ev_timer> watcher for any timeouts (many
1115	provide just this functionality). Then, in the check watcher you check for	3056	libraries provide exactly this functionality). Then, in the check watcher,
1116	any events that occured (by checking the pending status of all watchers	3057	you check for any events that occurred (by checking the pending status
1117	and stopping them) and call back into the library. The I/O and timer	3058	of all watchers and stopping them) and call back into the library. The
1118	callbacks will never actually be called (but must be valid nevertheless,	3059	I/O and timer callbacks will never actually be called (but must be valid
1119	because you never know, you know?).	3060	nevertheless, because you never know, you know?).
1120		3061
1121	As another example, the Perl Coro module uses these hooks to integrate	3062	As another example, the Perl Coro module uses these hooks to integrate
1122	coroutines into libev programs, by yielding to other active coroutines	3063	coroutines into libev programs, by yielding to other active coroutines
1123	during each prepare and only letting the process block if no coroutines	3064	during each prepare and only letting the process block if no coroutines
1124	are ready to run (it's actually more complicated: it only runs coroutines	3065	are ready to run (it's actually more complicated: it only runs coroutines
1125	with priority higher than or equal to the event loop and one coroutine	3066	with priority higher than or equal to the event loop and one coroutine
1126	of lower priority, but only once, using idle watchers to keep the event	3067	of lower priority, but only once, using idle watchers to keep the event
1127	loop from blocking if lower-priority coroutines are active, thus mapping	3068	loop from blocking if lower-priority coroutines are active, thus mapping
1128	low-priority coroutines to idle/background tasks).	3069	low-priority coroutines to idle/background tasks).
1129		3070
		3071	When used for this purpose, it is recommended to give C<ev_check> watchers
		3072	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
		3073	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		3074	watchers).
		3075
		3076	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		3077	activate ("feed") events into libev. While libev fully supports this, they
		3078	might get executed before other C<ev_check> watchers did their job. As
		3079	C<ev_check> watchers are often used to embed other (non-libev) event
		3080	loops those other event loops might be in an unusable state until their
		3081	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		3082	others).
		3083
		3084	=head3 Abusing an C<ev_check> watcher for its side-effect
		3085
		3086	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		3087	useful because they are called once per event loop iteration. For
		3088	example, if you want to handle a large number of connections fairly, you
		3089	normally only do a bit of work for each active connection, and if there
		3090	is more work to do, you wait for the next event loop iteration, so other
		3091	connections have a chance of making progress.
		3092
		3093	Using an C<ev_check> watcher is almost enough: it will be called on the
		3094	next event loop iteration. However, that isn't as soon as possible -
		3095	without external events, your C<ev_check> watcher will not be invoked.
		3096
		3097	This is where C<ev_idle> watchers come in handy - all you need is a
		3098	single global idle watcher that is active as long as you have one active
		3099	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		3100	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		3101	invoked. Neither watcher alone can do that.
		3102
		3103	=head3 Watcher-Specific Functions and Data Members
		3104
1130	=over 4	3105	=over 4
1131		3106
1132	=item ev_prepare_init (ev_prepare *, callback)	3107	=item ev_prepare_init (ev_prepare *, callback)
1133		3108
1134	=item ev_check_init (ev_check *, callback)	3109	=item ev_check_init (ev_check *, callback)
1135		3110
1136	Initialises and configures the prepare or check watcher - they have no	3111	Initialises and configures the prepare or check watcher - they have no
1137	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	3112	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1138	macros, but using them is utterly, utterly and completely pointless.	3113	macros, but using them is utterly, utterly, utterly and completely
		3114	pointless.
1139		3115
1140	=back	3116	=back
1141		3117
1142	Example: *TODO*.	3118	=head3 Examples
		3119
		3120	There are a number of principal ways to embed other event loops or modules
		3121	into libev. Here are some ideas on how to include libadns into libev
		3122	(there is a Perl module named C<EV::ADNS> that does this, which you could
		3123	use as a working example. Another Perl module named C<EV::Glib> embeds a
		3124	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		3125	Glib event loop).
		3126
		3127	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		3128	and in a check watcher, destroy them and call into libadns. What follows
		3129	is pseudo-code only of course. This requires you to either use a low
		3130	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		3131	the callbacks for the IO/timeout watchers might not have been called yet.
		3132
		3133	static ev_io iow [nfd];
		3134	static ev_timer tw;
		3135
		3136	static void
		3137	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		3138	{
		3139	}
		3140
		3141	// create io watchers for each fd and a timer before blocking
		3142	static void
		3143	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		3144	{
		3145	int timeout = 3600000;
		3146	struct pollfd fds [nfd];
		3147	// actual code will need to loop here and realloc etc.
		3148	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		3149
		3150	/* the callback is illegal, but won't be called as we stop during check */
		3151	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		3152	ev_timer_start (loop, &tw);
		3153
		3154	// create one ev_io per pollfd
		3155	for (int i = 0; i < nfd; ++i)
		3156	{
		3157	ev_io_init (iow + i, io_cb, fds [i].fd,
		3158	((fds [i].events & POLLIN ? EV_READ : 0)
		3159	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		3160
		3161	fds [i].revents = 0;
		3162	ev_io_start (loop, iow + i);
		3163	}
		3164	}
		3165
		3166	// stop all watchers after blocking
		3167	static void
		3168	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		3169	{
		3170	ev_timer_stop (loop, &tw);
		3171
		3172	for (int i = 0; i < nfd; ++i)
		3173	{
		3174	// set the relevant poll flags
		3175	// could also call adns_processreadable etc. here
		3176	struct pollfd *fd = fds + i;
		3177	int revents = ev_clear_pending (iow + i);
		3178	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		3179	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		3180
		3181	// now stop the watcher
		3182	ev_io_stop (loop, iow + i);
		3183	}
		3184
		3185	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		3186	}
		3187
		3188	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		3189	in the prepare watcher and would dispose of the check watcher.
		3190
		3191	Method 3: If the module to be embedded supports explicit event
		3192	notification (libadns does), you can also make use of the actual watcher
		3193	callbacks, and only destroy/create the watchers in the prepare watcher.
		3194
		3195	static void
		3196	timer_cb (EV_P_ ev_timer *w, int revents)
		3197	{
		3198	adns_state ads = (adns_state)w->data;
		3199	update_now (EV_A);
		3200
		3201	adns_processtimeouts (ads, &tv_now);
		3202	}
		3203
		3204	static void
		3205	io_cb (EV_P_ ev_io *w, int revents)
		3206	{
		3207	adns_state ads = (adns_state)w->data;
		3208	update_now (EV_A);
		3209
		3210	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		3211	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		3212	}
		3213
		3214	// do not ever call adns_afterpoll
		3215
		3216	Method 4: Do not use a prepare or check watcher because the module you
		3217	want to embed is not flexible enough to support it. Instead, you can
		3218	override their poll function. The drawback with this solution is that the
		3219	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		3220	this approach, effectively embedding EV as a client into the horrible
		3221	libglib event loop.
		3222
		3223	static gint
		3224	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		3225	{
		3226	int got_events = 0;
		3227
		3228	for (n = 0; n < nfds; ++n)
		3229	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		3230
		3231	if (timeout >= 0)
		3232	// create/start timer
		3233
		3234	// poll
		3235	ev_run (EV_A_ 0);
		3236
		3237	// stop timer again
		3238	if (timeout >= 0)
		3239	ev_timer_stop (EV_A_ &to);
		3240
		3241	// stop io watchers again - their callbacks should have set
		3242	for (n = 0; n < nfds; ++n)
		3243	ev_io_stop (EV_A_ iow [n]);
		3244
		3245	return got_events;
		3246	}
1143		3247
1144		3248
1145	=head2 C<ev_embed> - when one backend isn't enough...	3249	=head2 C<ev_embed> - when one backend isn't enough...
1146		3250
1147	This is a rather advanced watcher type that lets you embed one event loop	3251	This is a rather advanced watcher type that lets you embed one event loop
…		…
1153	prioritise I/O.	3257	prioritise I/O.
1154		3258
1155	As an example for a bug workaround, the kqueue backend might only support	3259	As an example for a bug workaround, the kqueue backend might only support
1156	sockets on some platform, so it is unusable as generic backend, but you	3260	sockets on some platform, so it is unusable as generic backend, but you
1157	still want to make use of it because you have many sockets and it scales	3261	still want to make use of it because you have many sockets and it scales
1158	so nicely. In this case, you would create a kqueue-based loop and embed it	3262	so nicely. In this case, you would create a kqueue-based loop and embed
1159	into your default loop (which might use e.g. poll). Overall operation will	3263	it into your default loop (which might use e.g. poll). Overall operation
1160	be a bit slower because first libev has to poll and then call kevent, but	3264	will be a bit slower because first libev has to call C<poll> and then
1161	at least you can use both at what they are best.	3265	C<kevent>, but at least you can use both mechanisms for what they are
		3266	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
1162		3267
1163	As for prioritising I/O: rarely you have the case where some fds have	3268	As for prioritising I/O: under rare circumstances you have the case where
1164	to be watched and handled very quickly (with low latency), and even	3269	some fds have to be watched and handled very quickly (with low latency),
1165	priorities and idle watchers might have too much overhead. In this case	3270	and even priorities and idle watchers might have too much overhead. In
1166	you would put all the high priority stuff in one loop and all the rest in	3271	this case you would put all the high priority stuff in one loop and all
1167	a second one, and embed the second one in the first.	3272	the rest in a second one, and embed the second one in the first.
1168		3273
1169	As long as the watcher is active, the callback will be invoked every time	3274	As long as the watcher is active, the callback will be invoked every
1170	there might be events pending in the embedded loop. The callback must then	3275	time there might be events pending in the embedded loop. The callback
1171	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	3276	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
1172	their callbacks (you could also start an idle watcher to give the embedded	3277	sweep and invoke their callbacks (the callback doesn't need to invoke the
1173	loop strictly lower priority for example). You can also set the callback	3278	C<ev_embed_sweep> function directly, it could also start an idle watcher
1174	to C<0>, in which case the embed watcher will automatically execute the	3279	to give the embedded loop strictly lower priority for example).
1175	embedded loop sweep.
1176		3280
1177	As long as the watcher is started it will automatically handle events. The	3281	You can also set the callback to C<0>, in which case the embed watcher
1178	callback will be invoked whenever some events have been handled. You can	3282	will automatically execute the embedded loop sweep whenever necessary.
1179	set the callback to C<0> to avoid having to specify one if you are not
1180	interested in that.
1181		3283
1182	Also, there have not currently been made special provisions for forking:	3284	Fork detection will be handled transparently while the C<ev_embed> watcher
1183	when you fork, you not only have to call C<ev_loop_fork> on both loops,	3285	is active, i.e., the embedded loop will automatically be forked when the
1184	but you will also have to stop and restart any C<ev_embed> watchers	3286	embedding loop forks. In other cases, the user is responsible for calling
1185	yourself.	3287	C<ev_loop_fork> on the embedded loop.
1186		3288
1187	Unfortunately, not all backends are embeddable, only the ones returned by	3289	Unfortunately, not all backends are embeddable: only the ones returned by
1188	C<ev_embeddable_backends> are, which, unfortunately, does not include any	3290	C<ev_embeddable_backends> are, which, unfortunately, does not include any
1189	portable one.	3291	portable one.
1190		3292
1191	So when you want to use this feature you will always have to be prepared	3293	So when you want to use this feature you will always have to be prepared
1192	that you cannot get an embeddable loop. The recommended way to get around	3294	that you cannot get an embeddable loop. The recommended way to get around
1193	this is to have a separate variables for your embeddable loop, try to	3295	this is to have a separate variables for your embeddable loop, try to
1194	create it, and if that fails, use the normal loop for everything:	3296	create it, and if that fails, use the normal loop for everything.
1195		3297
1196	struct ev_loop *loop_hi = ev_default_init (0);	3298	=head3 C<ev_embed> and fork
1197	struct ev_loop *loop_lo = 0;
1198	struct ev_embed embed;
1199
1200	// see if there is a chance of getting one that works
1201	// (remember that a flags value of 0 means autodetection)
1202	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1203	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1204	: 0;
1205		3299
1206	// if we got one, then embed it, otherwise default to loop_hi	3300	While the C<ev_embed> watcher is running, forks in the embedding loop will
1207	if (loop_lo)	3301	automatically be applied to the embedded loop as well, so no special
1208	{	3302	fork handling is required in that case. When the watcher is not running,
1209	ev_embed_init (&embed, 0, loop_lo);	3303	however, it is still the task of the libev user to call C<ev_loop_fork ()>
1210	ev_embed_start (loop_hi, &embed);	3304	as applicable.
1211	}	3305
1212	else	3306	=head3 Watcher-Specific Functions and Data Members
1213	loop_lo = loop_hi;
1214		3307
1215	=over 4	3308	=over 4
1216		3309
1217	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)	3310	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
1218		3311
1219	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)	3312	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
1220		3313
1221	Configures the watcher to embed the given loop, which must be	3314	Configures the watcher to embed the given loop, which must be
1222	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be	3315	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1223	invoked automatically, otherwise it is the responsibility of the callback	3316	invoked automatically, otherwise it is the responsibility of the callback
1224	to invoke it (it will continue to be called until the sweep has been done,	3317	to invoke it (it will continue to be called until the sweep has been done,
1225	if you do not want thta, you need to temporarily stop the embed watcher).	3318	if you do not want that, you need to temporarily stop the embed watcher).
1226		3319
1227	=item ev_embed_sweep (loop, ev_embed *)	3320	=item ev_embed_sweep (loop, ev_embed *)
1228		3321
1229	Make a single, non-blocking sweep over the embedded loop. This works	3322	Make a single, non-blocking sweep over the embedded loop. This works
1230	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most	3323	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
1231	apropriate way for embedded loops.	3324	appropriate way for embedded loops.
		3325
		3326	=item struct ev_loop *other [read-only]
		3327
		3328	The embedded event loop.
1232		3329
1233	=back	3330	=back
1234		3331
		3332	=head3 Examples
		3333
		3334	Example: Try to get an embeddable event loop and embed it into the default
		3335	event loop. If that is not possible, use the default loop. The default
		3336	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3337	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3338	used).
		3339
		3340	struct ev_loop *loop_hi = ev_default_init (0);
		3341	struct ev_loop *loop_lo = 0;
		3342	ev_embed embed;
		3343
		3344	// see if there is a chance of getting one that works
		3345	// (remember that a flags value of 0 means autodetection)
		3346	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3347	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3348	: 0;
		3349
		3350	// if we got one, then embed it, otherwise default to loop_hi
		3351	if (loop_lo)
		3352	{
		3353	ev_embed_init (&embed, 0, loop_lo);
		3354	ev_embed_start (loop_hi, &embed);
		3355	}
		3356	else
		3357	loop_lo = loop_hi;
		3358
		3359	Example: Check if kqueue is available but not recommended and create
		3360	a kqueue backend for use with sockets (which usually work with any
		3361	kqueue implementation). Store the kqueue/socket-only event loop in
		3362	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3363
		3364	struct ev_loop *loop = ev_default_init (0);
		3365	struct ev_loop *loop_socket = 0;
		3366	ev_embed embed;
		3367
		3368	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3369	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3370	{
		3371	ev_embed_init (&embed, 0, loop_socket);
		3372	ev_embed_start (loop, &embed);
		3373	}
		3374
		3375	if (!loop_socket)
		3376	loop_socket = loop;
		3377
		3378	// now use loop_socket for all sockets, and loop for everything else
		3379
		3380
		3381	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3382
		3383	Fork watchers are called when a C<fork ()> was detected (usually because
		3384	whoever is a good citizen cared to tell libev about it by calling
		3385	C<ev_loop_fork>). The invocation is done before the event loop blocks next
		3386	and before C<ev_check> watchers are being called, and only in the child
		3387	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
		3388	and calls it in the wrong process, the fork handlers will be invoked, too,
		3389	of course.
		3390
		3391	=head3 The special problem of life after fork - how is it possible?
		3392
		3393	Most uses of C<fork ()> consist of forking, then some simple calls to set
		3394	up/change the process environment, followed by a call to C<exec()>. This
		3395	sequence should be handled by libev without any problems.
		3396
		3397	This changes when the application actually wants to do event handling
		3398	in the child, or both parent in child, in effect "continuing" after the
		3399	fork.
		3400
		3401	The default mode of operation (for libev, with application help to detect
		3402	forks) is to duplicate all the state in the child, as would be expected
		3403	when I<either> the parent I<or> the child process continues.
		3404
		3405	When both processes want to continue using libev, then this is usually the
		3406	wrong result. In that case, usually one process (typically the parent) is
		3407	supposed to continue with all watchers in place as before, while the other
		3408	process typically wants to start fresh, i.e. without any active watchers.
		3409
		3410	The cleanest and most efficient way to achieve that with libev is to
		3411	simply create a new event loop, which of course will be "empty", and
		3412	use that for new watchers. This has the advantage of not touching more
		3413	memory than necessary, and thus avoiding the copy-on-write, and the
		3414	disadvantage of having to use multiple event loops (which do not support
		3415	signal watchers).
		3416
		3417	When this is not possible, or you want to use the default loop for
		3418	other reasons, then in the process that wants to start "fresh", call
		3419	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3420	Destroying the default loop will "orphan" (not stop) all registered
		3421	watchers, so you have to be careful not to execute code that modifies
		3422	those watchers. Note also that in that case, you have to re-register any
		3423	signal watchers.
		3424
		3425	=head3 Watcher-Specific Functions and Data Members
		3426
		3427	=over 4
		3428
		3429	=item ev_fork_init (ev_fork *, callback)
		3430
		3431	Initialises and configures the fork watcher - it has no parameters of any
		3432	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3433	really.
		3434
		3435	=back
		3436
		3437
		3438	=head2 C<ev_cleanup> - even the best things end
		3439
		3440	Cleanup watchers are called just before the event loop is being destroyed
		3441	by a call to C<ev_loop_destroy>.
		3442
		3443	While there is no guarantee that the event loop gets destroyed, cleanup
		3444	watchers provide a convenient method to install cleanup hooks for your
		3445	program, worker threads and so on - you just to make sure to destroy the
		3446	loop when you want them to be invoked.
		3447
		3448	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3449	all other watchers, they do not keep a reference to the event loop (which
		3450	makes a lot of sense if you think about it). Like all other watchers, you
		3451	can call libev functions in the callback, except C<ev_cleanup_start>.
		3452
		3453	=head3 Watcher-Specific Functions and Data Members
		3454
		3455	=over 4
		3456
		3457	=item ev_cleanup_init (ev_cleanup *, callback)
		3458
		3459	Initialises and configures the cleanup watcher - it has no parameters of
		3460	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3461	pointless, I assure you.
		3462
		3463	=back
		3464
		3465	Example: Register an atexit handler to destroy the default loop, so any
		3466	cleanup functions are called.
		3467
		3468	static void
		3469	program_exits (void)
		3470	{
		3471	ev_loop_destroy (EV_DEFAULT_UC);
		3472	}
		3473
		3474	...
		3475	atexit (program_exits);
		3476
		3477
		3478	=head2 C<ev_async> - how to wake up an event loop
		3479
		3480	In general, you cannot use an C<ev_loop> from multiple threads or other
		3481	asynchronous sources such as signal handlers (as opposed to multiple event
		3482	loops - those are of course safe to use in different threads).
		3483
		3484	Sometimes, however, you need to wake up an event loop you do not control,
		3485	for example because it belongs to another thread. This is what C<ev_async>
		3486	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3487	it by calling C<ev_async_send>, which is thread- and signal safe.
		3488
		3489	This functionality is very similar to C<ev_signal> watchers, as signals,
		3490	too, are asynchronous in nature, and signals, too, will be compressed
		3491	(i.e. the number of callback invocations may be less than the number of
		3492	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
		3493	of "global async watchers" by using a watcher on an otherwise unused
		3494	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3495	even without knowing which loop owns the signal.
		3496
		3497	=head3 Queueing
		3498
		3499	C<ev_async> does not support queueing of data in any way. The reason
		3500	is that the author does not know of a simple (or any) algorithm for a
		3501	multiple-writer-single-reader queue that works in all cases and doesn't
		3502	need elaborate support such as pthreads or unportable memory access
		3503	semantics.
		3504
		3505	That means that if you want to queue data, you have to provide your own
		3506	queue. But at least I can tell you how to implement locking around your
		3507	queue:
		3508
		3509	=over 4
		3510
		3511	=item queueing from a signal handler context
		3512
		3513	To implement race-free queueing, you simply add to the queue in the signal
		3514	handler but you block the signal handler in the watcher callback. Here is
		3515	an example that does that for some fictitious SIGUSR1 handler:
		3516
		3517	static ev_async mysig;
		3518
		3519	static void
		3520	sigusr1_handler (void)
		3521	{
		3522	sometype data;
		3523
		3524	// no locking etc.
		3525	queue_put (data);
		3526	ev_async_send (EV_DEFAULT_ &mysig);
		3527	}
		3528
		3529	static void
		3530	mysig_cb (EV_P_ ev_async *w, int revents)
		3531	{
		3532	sometype data;
		3533	sigset_t block, prev;
		3534
		3535	sigemptyset (&block);
		3536	sigaddset (&block, SIGUSR1);
		3537	sigprocmask (SIG_BLOCK, &block, &prev);
		3538
		3539	while (queue_get (&data))
		3540	process (data);
		3541
		3542	if (sigismember (&prev, SIGUSR1)
		3543	sigprocmask (SIG_UNBLOCK, &block, 0);
		3544	}
		3545
		3546	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3547	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3548	either...).
		3549
		3550	=item queueing from a thread context
		3551
		3552	The strategy for threads is different, as you cannot (easily) block
		3553	threads but you can easily preempt them, so to queue safely you need to
		3554	employ a traditional mutex lock, such as in this pthread example:
		3555
		3556	static ev_async mysig;
		3557	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3558
		3559	static void
		3560	otherthread (void)
		3561	{
		3562	// only need to lock the actual queueing operation
		3563	pthread_mutex_lock (&mymutex);
		3564	queue_put (data);
		3565	pthread_mutex_unlock (&mymutex);
		3566
		3567	ev_async_send (EV_DEFAULT_ &mysig);
		3568	}
		3569
		3570	static void
		3571	mysig_cb (EV_P_ ev_async *w, int revents)
		3572	{
		3573	pthread_mutex_lock (&mymutex);
		3574
		3575	while (queue_get (&data))
		3576	process (data);
		3577
		3578	pthread_mutex_unlock (&mymutex);
		3579	}
		3580
		3581	=back
		3582
		3583
		3584	=head3 Watcher-Specific Functions and Data Members
		3585
		3586	=over 4
		3587
		3588	=item ev_async_init (ev_async *, callback)
		3589
		3590	Initialises and configures the async watcher - it has no parameters of any
		3591	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3592	trust me.
		3593
		3594	=item ev_async_send (loop, ev_async *)
		3595
		3596	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3597	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3598	returns.
		3599
		3600	Unlike C<ev_feed_event>, this call is safe to do from other threads,
		3601	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
		3602	embedding section below on what exactly this means).
		3603
		3604	Note that, as with other watchers in libev, multiple events might get
		3605	compressed into a single callback invocation (another way to look at
		3606	this is that C<ev_async> watchers are level-triggered: they are set on
		3607	C<ev_async_send>, reset when the event loop detects that).
		3608
		3609	This call incurs the overhead of at most one extra system call per event
		3610	loop iteration, if the event loop is blocked, and no syscall at all if
		3611	the event loop (or your program) is processing events. That means that
		3612	repeated calls are basically free (there is no need to avoid calls for
		3613	performance reasons) and that the overhead becomes smaller (typically
		3614	zero) under load.
		3615
		3616	=item bool = ev_async_pending (ev_async *)
		3617
		3618	Returns a non-zero value when C<ev_async_send> has been called on the
		3619	watcher but the event has not yet been processed (or even noted) by the
		3620	event loop.
		3621
		3622	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3623	the loop iterates next and checks for the watcher to have become active,
		3624	it will reset the flag again. C<ev_async_pending> can be used to very
		3625	quickly check whether invoking the loop might be a good idea.
		3626
		3627	Not that this does I<not> check whether the watcher itself is pending,
		3628	only whether it has been requested to make this watcher pending: there
		3629	is a time window between the event loop checking and resetting the async
		3630	notification, and the callback being invoked.
		3631
		3632	=back
		3633
1235		3634
1236	=head1 OTHER FUNCTIONS	3635	=head1 OTHER FUNCTIONS
1237		3636
1238	There are some other functions of possible interest. Described. Here. Now.	3637	There are some other functions of possible interest. Described. Here. Now.
1239		3638
1240	=over 4	3639	=over 4
1241		3640
1242	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3641	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
1243		3642
1244	This function combines a simple timer and an I/O watcher, calls your	3643	This function combines a simple timer and an I/O watcher, calls your
1245	callback on whichever event happens first and automatically stop both	3644	callback on whichever event happens first and automatically stops both
1246	watchers. This is useful if you want to wait for a single event on an fd	3645	watchers. This is useful if you want to wait for a single event on an fd
1247	or timeout without having to allocate/configure/start/stop/free one or	3646	or timeout without having to allocate/configure/start/stop/free one or
1248	more watchers yourself.	3647	more watchers yourself.
1249		3648
1250	If C<fd> is less than 0, then no I/O watcher will be started and events	3649	If C<fd> is less than 0, then no I/O watcher will be started and the
1251	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3650	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
1252	C<events> set will be craeted and started.	3651	the given C<fd> and C<events> set will be created and started.
1253		3652
1254	If C<timeout> is less than 0, then no timeout watcher will be	3653	If C<timeout> is less than 0, then no timeout watcher will be
1255	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3654	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
1256	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3655	repeat = 0) will be started. C<0> is a valid timeout.
1257	dubious value.
1258		3656
1259	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3657	The callback has the type C<void (*cb)(int revents, void *arg)> and is
1260	passed an C<revents> set like normal event callbacks (a combination of	3658	passed an C<revents> set like normal event callbacks (a combination of
1261	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3659	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
1262	value passed to C<ev_once>:	3660	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3661	a timeout and an io event at the same time - you probably should give io
		3662	events precedence.
1263		3663
		3664	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3665
1264	static void stdin_ready (int revents, void *arg)	3666	static void stdin_ready (int revents, void *arg)
		3667	{
		3668	if (revents & EV_READ)
		3669	/* stdin might have data for us, joy! */;
		3670	else if (revents & EV_TIMER)
		3671	/* doh, nothing entered */;
		3672	}
		3673
		3674	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
		3675
		3676	=item ev_feed_fd_event (loop, int fd, int revents)
		3677
		3678	Feed an event on the given fd, as if a file descriptor backend detected
		3679	the given events.
		3680
		3681	=item ev_feed_signal_event (loop, int signum)
		3682
		3683	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3684	which is async-safe.
		3685
		3686	=back
		3687
		3688
		3689	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3690
		3691	This section explains some common idioms that are not immediately
		3692	obvious. Note that examples are sprinkled over the whole manual, and this
		3693	section only contains stuff that wouldn't fit anywhere else.
		3694
		3695	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3696
		3697	Each watcher has, by default, a C<void *data> member that you can read
		3698	or modify at any time: libev will completely ignore it. This can be used
		3699	to associate arbitrary data with your watcher. If you need more data and
		3700	don't want to allocate memory separately and store a pointer to it in that
		3701	data member, you can also "subclass" the watcher type and provide your own
		3702	data:
		3703
		3704	struct my_io
		3705	{
		3706	ev_io io;
		3707	int otherfd;
		3708	void *somedata;
		3709	struct whatever *mostinteresting;
		3710	};
		3711
		3712	...
		3713	struct my_io w;
		3714	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3715
		3716	And since your callback will be called with a pointer to the watcher, you
		3717	can cast it back to your own type:
		3718
		3719	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3720	{
		3721	struct my_io *w = (struct my_io *)w_;
		3722	...
		3723	}
		3724
		3725	More interesting and less C-conformant ways of casting your callback
		3726	function type instead have been omitted.
		3727
		3728	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3729
		3730	Another common scenario is to use some data structure with multiple
		3731	embedded watchers, in effect creating your own watcher that combines
		3732	multiple libev event sources into one "super-watcher":
		3733
		3734	struct my_biggy
		3735	{
		3736	int some_data;
		3737	ev_timer t1;
		3738	ev_timer t2;
		3739	}
		3740
		3741	In this case getting the pointer to C<my_biggy> is a bit more
		3742	complicated: Either you store the address of your C<my_biggy> struct in
		3743	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3744	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3745	real programmers):
		3746
		3747	#include <stddef.h>
		3748
		3749	static void
		3750	t1_cb (EV_P_ ev_timer *w, int revents)
		3751	{
		3752	struct my_biggy big = (struct my_biggy *)
		3753	(((char *)w) - offsetof (struct my_biggy, t1));
		3754	}
		3755
		3756	static void
		3757	t2_cb (EV_P_ ev_timer *w, int revents)
		3758	{
		3759	struct my_biggy big = (struct my_biggy *)
		3760	(((char *)w) - offsetof (struct my_biggy, t2));
		3761	}
		3762
		3763	=head2 AVOIDING FINISHING BEFORE RETURNING
		3764
		3765	Often you have structures like this in event-based programs:
		3766
		3767	callback ()
1265	{	3768	{
1266	if (revents & EV_TIMEOUT)	3769	free (request);
1267	/* doh, nothing entered */;
1268	else if (revents & EV_READ)
1269	/* stdin might have data for us, joy! */;
1270	}	3770	}
1271		3771
1272	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3772	request = start_new_request (..., callback);
1273		3773
1274	=item ev_feed_event (ev_loop *, watcher *, int revents)	3774	The intent is to start some "lengthy" operation. The C<request> could be
		3775	used to cancel the operation, or do other things with it.
1275		3776
1276	Feeds the given event set into the event loop, as if the specified event	3777	It's not uncommon to have code paths in C<start_new_request> that
1277	had happened for the specified watcher (which must be a pointer to an	3778	immediately invoke the callback, for example, to report errors. Or you add
1278	initialised but not necessarily started event watcher).	3779	some caching layer that finds that it can skip the lengthy aspects of the
		3780	operation and simply invoke the callback with the result.
1279		3781
1280	=item ev_feed_fd_event (ev_loop *, int fd, int revents)	3782	The problem here is that this will happen I<before> C<start_new_request>
		3783	has returned, so C<request> is not set.
1281		3784
1282	Feed an event on the given fd, as if a file descriptor backend detected	3785	Even if you pass the request by some safer means to the callback, you
1283	the given events it.	3786	might want to do something to the request after starting it, such as
		3787	canceling it, which probably isn't working so well when the callback has
		3788	already been invoked.
1284		3789
1285	=item ev_feed_signal_event (ev_loop *loop, int signum)	3790	A common way around all these issues is to make sure that
		3791	C<start_new_request> I<always> returns before the callback is invoked. If
		3792	C<start_new_request> immediately knows the result, it can artificially
		3793	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3794	example, or more sneakily, by reusing an existing (stopped) watcher and
		3795	pushing it into the pending queue:
1286		3796
1287	Feed an event as if the given signal occured (C<loop> must be the default	3797	ev_set_cb (watcher, callback);
1288	loop!).	3798	ev_feed_event (EV_A_ watcher, 0);
1289		3799
1290	=back	3800	This way, C<start_new_request> can safely return before the callback is
		3801	invoked, while not delaying callback invocation too much.
		3802
		3803	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3804
		3805	Often (especially in GUI toolkits) there are places where you have
		3806	I<modal> interaction, which is most easily implemented by recursively
		3807	invoking C<ev_run>.
		3808
		3809	This brings the problem of exiting - a callback might want to finish the
		3810	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3811	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3812	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3813	other combination: In these cases, a simple C<ev_break> will not work.
		3814
		3815	The solution is to maintain "break this loop" variable for each C<ev_run>
		3816	invocation, and use a loop around C<ev_run> until the condition is
		3817	triggered, using C<EVRUN_ONCE>:
		3818
		3819	// main loop
		3820	int exit_main_loop = 0;
		3821
		3822	while (!exit_main_loop)
		3823	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3824
		3825	// in a modal watcher
		3826	int exit_nested_loop = 0;
		3827
		3828	while (!exit_nested_loop)
		3829	ev_run (EV_A_ EVRUN_ONCE);
		3830
		3831	To exit from any of these loops, just set the corresponding exit variable:
		3832
		3833	// exit modal loop
		3834	exit_nested_loop = 1;
		3835
		3836	// exit main program, after modal loop is finished
		3837	exit_main_loop = 1;
		3838
		3839	// exit both
		3840	exit_main_loop = exit_nested_loop = 1;
		3841
		3842	=head2 THREAD LOCKING EXAMPLE
		3843
		3844	Here is a fictitious example of how to run an event loop in a different
		3845	thread from where callbacks are being invoked and watchers are
		3846	created/added/removed.
		3847
		3848	For a real-world example, see the C<EV::Loop::Async> perl module,
		3849	which uses exactly this technique (which is suited for many high-level
		3850	languages).
		3851
		3852	The example uses a pthread mutex to protect the loop data, a condition
		3853	variable to wait for callback invocations, an async watcher to notify the
		3854	event loop thread and an unspecified mechanism to wake up the main thread.
		3855
		3856	First, you need to associate some data with the event loop:
		3857
		3858	typedef struct {
		3859	mutex_t lock; /* global loop lock */
		3860	ev_async async_w;
		3861	thread_t tid;
		3862	cond_t invoke_cv;
		3863	} userdata;
		3864
		3865	void prepare_loop (EV_P)
		3866	{
		3867	// for simplicity, we use a static userdata struct.
		3868	static userdata u;
		3869
		3870	ev_async_init (&u->async_w, async_cb);
		3871	ev_async_start (EV_A_ &u->async_w);
		3872
		3873	pthread_mutex_init (&u->lock, 0);
		3874	pthread_cond_init (&u->invoke_cv, 0);
		3875
		3876	// now associate this with the loop
		3877	ev_set_userdata (EV_A_ u);
		3878	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3879	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3880
		3881	// then create the thread running ev_run
		3882	pthread_create (&u->tid, 0, l_run, EV_A);
		3883	}
		3884
		3885	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3886	solely to wake up the event loop so it takes notice of any new watchers
		3887	that might have been added:
		3888
		3889	static void
		3890	async_cb (EV_P_ ev_async *w, int revents)
		3891	{
		3892	// just used for the side effects
		3893	}
		3894
		3895	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3896	protecting the loop data, respectively.
		3897
		3898	static void
		3899	l_release (EV_P)
		3900	{
		3901	userdata *u = ev_userdata (EV_A);
		3902	pthread_mutex_unlock (&u->lock);
		3903	}
		3904
		3905	static void
		3906	l_acquire (EV_P)
		3907	{
		3908	userdata *u = ev_userdata (EV_A);
		3909	pthread_mutex_lock (&u->lock);
		3910	}
		3911
		3912	The event loop thread first acquires the mutex, and then jumps straight
		3913	into C<ev_run>:
		3914
		3915	void *
		3916	l_run (void *thr_arg)
		3917	{
		3918	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3919
		3920	l_acquire (EV_A);
		3921	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3922	ev_run (EV_A_ 0);
		3923	l_release (EV_A);
		3924
		3925	return 0;
		3926	}
		3927
		3928	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3929	signal the main thread via some unspecified mechanism (signals? pipe
		3930	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3931	have been called (in a while loop because a) spurious wakeups are possible
		3932	and b) skipping inter-thread-communication when there are no pending
		3933	watchers is very beneficial):
		3934
		3935	static void
		3936	l_invoke (EV_P)
		3937	{
		3938	userdata *u = ev_userdata (EV_A);
		3939
		3940	while (ev_pending_count (EV_A))
		3941	{
		3942	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3943	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3944	}
		3945	}
		3946
		3947	Now, whenever the main thread gets told to invoke pending watchers, it
		3948	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3949	thread to continue:
		3950
		3951	static void
		3952	real_invoke_pending (EV_P)
		3953	{
		3954	userdata *u = ev_userdata (EV_A);
		3955
		3956	pthread_mutex_lock (&u->lock);
		3957	ev_invoke_pending (EV_A);
		3958	pthread_cond_signal (&u->invoke_cv);
		3959	pthread_mutex_unlock (&u->lock);
		3960	}
		3961
		3962	Whenever you want to start/stop a watcher or do other modifications to an
		3963	event loop, you will now have to lock:
		3964
		3965	ev_timer timeout_watcher;
		3966	userdata *u = ev_userdata (EV_A);
		3967
		3968	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3969
		3970	pthread_mutex_lock (&u->lock);
		3971	ev_timer_start (EV_A_ &timeout_watcher);
		3972	ev_async_send (EV_A_ &u->async_w);
		3973	pthread_mutex_unlock (&u->lock);
		3974
		3975	Note that sending the C<ev_async> watcher is required because otherwise
		3976	an event loop currently blocking in the kernel will have no knowledge
		3977	about the newly added timer. By waking up the loop it will pick up any new
		3978	watchers in the next event loop iteration.
		3979
		3980	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3981
		3982	While the overhead of a callback that e.g. schedules a thread is small, it
		3983	is still an overhead. If you embed libev, and your main usage is with some
		3984	kind of threads or coroutines, you might want to customise libev so that
		3985	doesn't need callbacks anymore.
		3986
		3987	Imagine you have coroutines that you can switch to using a function
		3988	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3989	and that due to some magic, the currently active coroutine is stored in a
		3990	global called C<current_coro>. Then you can build your own "wait for libev
		3991	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3992	the differing C<;> conventions):
		3993
		3994	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3995	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3996
		3997	That means instead of having a C callback function, you store the
		3998	coroutine to switch to in each watcher, and instead of having libev call
		3999	your callback, you instead have it switch to that coroutine.
		4000
		4001	A coroutine might now wait for an event with a function called
		4002	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		4003	matter when, or whether the watcher is active or not when this function is
		4004	called):
		4005
		4006	void
		4007	wait_for_event (ev_watcher *w)
		4008	{
		4009	ev_set_cb (w, current_coro);
		4010	switch_to (libev_coro);
		4011	}
		4012
		4013	That basically suspends the coroutine inside C<wait_for_event> and
		4014	continues the libev coroutine, which, when appropriate, switches back to
		4015	this or any other coroutine.
		4016
		4017	You can do similar tricks if you have, say, threads with an event queue -
		4018	instead of storing a coroutine, you store the queue object and instead of
		4019	switching to a coroutine, you push the watcher onto the queue and notify
		4020	any waiters.
		4021
		4022	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		4023	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		4024
		4025	// my_ev.h
		4026	#define EV_CB_DECLARE(type) struct my_coro *cb;
		4027	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		4028	#include "../libev/ev.h"
		4029
		4030	// my_ev.c
		4031	#define EV_H "my_ev.h"
		4032	#include "../libev/ev.c"
		4033
		4034	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		4035	F<my_ev.c> into your project. When properly specifying include paths, you
		4036	can even use F<ev.h> as header file name directly.
1291		4037
1292		4038
1293	=head1 LIBEVENT EMULATION	4039	=head1 LIBEVENT EMULATION
1294		4040
1295	Libev offers a compatibility emulation layer for libevent. It cannot	4041	Libev offers a compatibility emulation layer for libevent. It cannot
1296	emulate the internals of libevent, so here are some usage hints:	4042	emulate the internals of libevent, so here are some usage hints:
1297		4043
1298	=over 4	4044	=over 4
		4045
		4046	=item * Only the libevent-1.4.1-beta API is being emulated.
		4047
		4048	This was the newest libevent version available when libev was implemented,
		4049	and is still mostly unchanged in 2010.
1299		4050
1300	=item * Use it by including <event.h>, as usual.	4051	=item * Use it by including <event.h>, as usual.
1301		4052
1302	=item * The following members are fully supported: ev_base, ev_callback,	4053	=item * The following members are fully supported: ev_base, ev_callback,
1303	ev_arg, ev_fd, ev_res, ev_events.	4054	ev_arg, ev_fd, ev_res, ev_events.
…		…
1308		4059
1309	=item * Priorities are not currently supported. Initialising priorities	4060	=item * Priorities are not currently supported. Initialising priorities
1310	will fail and all watchers will have the same priority, even though there	4061	will fail and all watchers will have the same priority, even though there
1311	is an ev_pri field.	4062	is an ev_pri field.
1312		4063
		4064	=item * In libevent, the last base created gets the signals, in libev, the
		4065	base that registered the signal gets the signals.
		4066
1313	=item * Other members are not supported.	4067	=item * Other members are not supported.
1314		4068
1315	=item * The libev emulation is I<not> ABI compatible to libevent, you need	4069	=item * The libev emulation is I<not> ABI compatible to libevent, you need
1316	to use the libev header file and library.	4070	to use the libev header file and library.
1317		4071
1318	=back	4072	=back
1319		4073
1320	=head1 C++ SUPPORT	4074	=head1 C++ SUPPORT
1321		4075
		4076	=head2 C API
		4077
		4078	The normal C API should work fine when used from C++: both ev.h and the
		4079	libev sources can be compiled as C++. Therefore, code that uses the C API
		4080	will work fine.
		4081
		4082	Proper exception specifications might have to be added to callbacks passed
		4083	to libev: exceptions may be thrown only from watcher callbacks, all other
		4084	callbacks (allocator, syserr, loop acquire/release and periodic reschedule
		4085	callbacks) must not throw exceptions, and might need a C<noexcept>
		4086	specification. If you have code that needs to be compiled as both C and
		4087	C++ you can use the C<EV_NOEXCEPT> macro for this:
		4088
		4089	static void
		4090	fatal_error (const char *msg) EV_NOEXCEPT
		4091	{
		4092	perror (msg);
		4093	abort ();
		4094	}
		4095
		4096	...
		4097	ev_set_syserr_cb (fatal_error);
		4098
		4099	The only API functions that can currently throw exceptions are C<ev_run>,
		4100	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		4101	because it runs cleanup watchers).
		4102
		4103	Throwing exceptions in watcher callbacks is only supported if libev itself
		4104	is compiled with a C++ compiler or your C and C++ environments allow
		4105	throwing exceptions through C libraries (most do).
		4106
		4107	=head2 C++ API
		4108
1322	Libev comes with some simplistic wrapper classes for C++ that mainly allow	4109	Libev comes with some simplistic wrapper classes for C++ that mainly allow
1323	you to use some convinience methods to start/stop watchers and also change	4110	you to use some convenience methods to start/stop watchers and also change
1324	the callback model to a model using method callbacks on objects.	4111	the callback model to a model using method callbacks on objects.
1325		4112
1326	To use it,	4113	To use it,
1327		4114
1328	#include <ev++.h>	4115	#include <ev++.h>
1329		4116
1330	(it is not installed by default). This automatically includes F<ev.h>	4117	This automatically includes F<ev.h> and puts all of its definitions (many
1331	and puts all of its definitions (many of them macros) into the global	4118	of them macros) into the global namespace. All C++ specific things are
1332	namespace. All C++ specific things are put into the C<ev> namespace.	4119	put into the C<ev> namespace. It should support all the same embedding
		4120	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
1333		4121
1334	It should support all the same embedding options as F<ev.h>, most notably	4122	Care has been taken to keep the overhead low. The only data member the C++
1335	C<EV_MULTIPLICITY>.	4123	classes add (compared to plain C-style watchers) is the event loop pointer
		4124	that the watcher is associated with (or no additional members at all if
		4125	you disable C<EV_MULTIPLICITY> when embedding libev).
		4126
		4127	Currently, functions, static and non-static member functions and classes
		4128	with C<operator ()> can be used as callbacks. Other types should be easy
		4129	to add as long as they only need one additional pointer for context. If
		4130	you need support for other types of functors please contact the author
		4131	(preferably after implementing it).
		4132
		4133	For all this to work, your C++ compiler either has to use the same calling
		4134	conventions as your C compiler (for static member functions), or you have
		4135	to embed libev and compile libev itself as C++.
1336		4136
1337	Here is a list of things available in the C<ev> namespace:	4137	Here is a list of things available in the C<ev> namespace:
1338		4138
1339	=over 4	4139	=over 4
1340		4140
…		…
1350	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.	4150	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
1351		4151
1352	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of	4152	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
1353	the same name in the C<ev> namespace, with the exception of C<ev_signal>	4153	the same name in the C<ev> namespace, with the exception of C<ev_signal>
1354	which is called C<ev::sig> to avoid clashes with the C<signal> macro	4154	which is called C<ev::sig> to avoid clashes with the C<signal> macro
1355	defines by many implementations.	4155	defined by many implementations.
1356		4156
1357	All of those classes have these methods:	4157	All of those classes have these methods:
1358		4158
1359	=over 4	4159	=over 4
1360		4160
1361	=item ev::TYPE::TYPE (object *, object::method *)	4161	=item ev::TYPE::TYPE ()
1362		4162
1363	=item ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)	4163	=item ev::TYPE::TYPE (loop)
1364		4164
1365	=item ev::TYPE::~TYPE	4165	=item ev::TYPE::~TYPE
1366		4166
1367	The constructor takes a pointer to an object and a method pointer to	4167	The constructor (optionally) takes an event loop to associate the watcher
1368	the event handler callback to call in this class. The constructor calls	4168	with. If it is omitted, it will use C<EV_DEFAULT>.
1369	C<ev_init> for you, which means you have to call the C<set> method	4169
1370	before starting it. If you do not specify a loop then the constructor	4170	The constructor calls C<ev_init> for you, which means you have to call the
1371	automatically associates the default loop with this watcher.	4171	C<set> method before starting it.
		4172
		4173	It will not set a callback, however: You have to call the templated C<set>
		4174	method to set a callback before you can start the watcher.
		4175
		4176	(The reason why you have to use a method is a limitation in C++ which does
		4177	not allow explicit template arguments for constructors).
1372		4178
1373	The destructor automatically stops the watcher if it is active.	4179	The destructor automatically stops the watcher if it is active.
1374		4180
		4181	=item w->set<class, &class::method> (object *)
		4182
		4183	This method sets the callback method to call. The method has to have a
		4184	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		4185	first argument and the C<revents> as second. The object must be given as
		4186	parameter and is stored in the C<data> member of the watcher.
		4187
		4188	This method synthesizes efficient thunking code to call your method from
		4189	the C callback that libev requires. If your compiler can inline your
		4190	callback (i.e. it is visible to it at the place of the C<set> call and
		4191	your compiler is good :), then the method will be fully inlined into the
		4192	thunking function, making it as fast as a direct C callback.
		4193
		4194	Example: simple class declaration and watcher initialisation
		4195
		4196	struct myclass
		4197	{
		4198	void io_cb (ev::io &w, int revents) { }
		4199	}
		4200
		4201	myclass obj;
		4202	ev::io iow;
		4203	iow.set <myclass, &myclass::io_cb> (&obj);
		4204
		4205	=item w->set (object *)
		4206
		4207	This is a variation of a method callback - leaving out the method to call
		4208	will default the method to C<operator ()>, which makes it possible to use
		4209	functor objects without having to manually specify the C<operator ()> all
		4210	the time. Incidentally, you can then also leave out the template argument
		4211	list.
		4212
		4213	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		4214	int revents)>.
		4215
		4216	See the method-C<set> above for more details.
		4217
		4218	Example: use a functor object as callback.
		4219
		4220	struct myfunctor
		4221	{
		4222	void operator() (ev::io &w, int revents)
		4223	{
		4224	...
		4225	}
		4226	}
		4227
		4228	myfunctor f;
		4229
		4230	ev::io w;
		4231	w.set (&f);
		4232
		4233	=item w->set<function> (void *data = 0)
		4234
		4235	Also sets a callback, but uses a static method or plain function as
		4236	callback. The optional C<data> argument will be stored in the watcher's
		4237	C<data> member and is free for you to use.
		4238
		4239	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		4240
		4241	See the method-C<set> above for more details.
		4242
		4243	Example: Use a plain function as callback.
		4244
		4245	static void io_cb (ev::io &w, int revents) { }
		4246	iow.set <io_cb> ();
		4247
1375	=item w->set (struct ev_loop *)	4248	=item w->set (loop)
1376		4249
1377	Associates a different C<struct ev_loop> with this watcher. You can only	4250	Associates a different C<struct ev_loop> with this watcher. You can only
1378	do this when the watcher is inactive (and not pending either).	4251	do this when the watcher is inactive (and not pending either).
1379		4252
1380	=item w->set ([args])	4253	=item w->set ([arguments])
1381		4254
1382	Basically the same as C<ev_TYPE_set>, with the same args. Must be	4255	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
		4256	with the same arguments. Either this method or a suitable start method
1383	called at least once. Unlike the C counterpart, an active watcher gets	4257	must be called at least once. Unlike the C counterpart, an active watcher
1384	automatically stopped and restarted.	4258	gets automatically stopped and restarted when reconfiguring it with this
		4259	method.
		4260
		4261	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4262	clashing with the C<set (loop)> method.
1385		4263
1386	=item w->start ()	4264	=item w->start ()
1387		4265
1388	Starts the watcher. Note that there is no C<loop> argument as the	4266	Starts the watcher. Note that there is no C<loop> argument, as the
1389	constructor already takes the loop.	4267	constructor already stores the event loop.
		4268
		4269	=item w->start ([arguments])
		4270
		4271	Instead of calling C<set> and C<start> methods separately, it is often
		4272	convenient to wrap them in one call. Uses the same type of arguments as
		4273	the configure C<set> method of the watcher.
1390		4274
1391	=item w->stop ()	4275	=item w->stop ()
1392		4276
1393	Stops the watcher if it is active. Again, no C<loop> argument.	4277	Stops the watcher if it is active. Again, no C<loop> argument.
1394		4278
1395	=item w->again () C<ev::timer>, C<ev::periodic> only	4279	=item w->again () (C<ev::timer>, C<ev::periodic> only)
1396		4280
1397	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding	4281	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
1398	C<ev_TYPE_again> function.	4282	C<ev_TYPE_again> function.
1399		4283
1400	=item w->sweep () C<ev::embed> only	4284	=item w->sweep () (C<ev::embed> only)
1401		4285
1402	Invokes C<ev_embed_sweep>.	4286	Invokes C<ev_embed_sweep>.
1403		4287
		4288	=item w->update () (C<ev::stat> only)
		4289
		4290	Invokes C<ev_stat_stat>.
		4291
1404	=back	4292	=back
1405		4293
1406	=back	4294	=back
1407		4295
1408	Example: Define a class with an IO and idle watcher, start one of them in	4296	Example: Define a class with two I/O and idle watchers, start the I/O
1409	the constructor.	4297	watchers in the constructor.
1410		4298
1411	class myclass	4299	class myclass
1412	{	4300	{
1413	ev_io io; void io_cb (ev::io &w, int revents);	4301	ev::io io ; void io_cb (ev::io &w, int revents);
		4302	ev::io io2 ; void io2_cb (ev::io &w, int revents);
1414	ev_idle idle void idle_cb (ev::idle &w, int revents);	4303	ev::idle idle; void idle_cb (ev::idle &w, int revents);
1415		4304
1416	myclass ();	4305	myclass (int fd)
		4306	{
		4307	io .set <myclass, &myclass::io_cb > (this);
		4308	io2 .set <myclass, &myclass::io2_cb > (this);
		4309	idle.set <myclass, &myclass::idle_cb> (this);
		4310
		4311	io.set (fd, ev::WRITE); // configure the watcher
		4312	io.start (); // start it whenever convenient
		4313
		4314	io2.start (fd, ev::READ); // set + start in one call
		4315	}
1417	}	4316	};
1418		4317
1419	myclass::myclass (int fd)	4318
1420	: io (this, &myclass::io_cb),	4319	=head1 OTHER LANGUAGE BINDINGS
1421	idle (this, &myclass::idle_cb)	4320
		4321	Libev does not offer other language bindings itself, but bindings for a
		4322	number of languages exist in the form of third-party packages. If you know
		4323	any interesting language binding in addition to the ones listed here, drop
		4324	me a note.
		4325
		4326	=over 4
		4327
		4328	=item Perl
		4329
		4330	The EV module implements the full libev API and is actually used to test
		4331	libev. EV is developed together with libev. Apart from the EV core module,
		4332	there are additional modules that implement libev-compatible interfaces
		4333	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		4334	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		4335	and C<EV::Glib>).
		4336
		4337	It can be found and installed via CPAN, its homepage is at
		4338	L<http://software.schmorp.de/pkg/EV>.
		4339
		4340	=item Python
		4341
		4342	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		4343	seems to be quite complete and well-documented.
		4344
		4345	=item Ruby
		4346
		4347	Tony Arcieri has written a ruby extension that offers access to a subset
		4348	of the libev API and adds file handle abstractions, asynchronous DNS and
		4349	more on top of it. It can be found via gem servers. Its homepage is at
		4350	L<http://rev.rubyforge.org/>.
		4351
		4352	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		4353	makes rev work even on mingw.
		4354
		4355	=item Haskell
		4356
		4357	A haskell binding to libev is available at
		4358	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		4359
		4360	=item D
		4361
		4362	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		4363	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
		4364
		4365	=item Ocaml
		4366
		4367	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4368	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4369
		4370	=item Lua
		4371
		4372	Brian Maher has written a partial interface to libev for lua (at the
		4373	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4374	L<http://github.com/brimworks/lua-ev>.
		4375
		4376	=item Javascript
		4377
		4378	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4379
		4380	=item Others
		4381
		4382	There are others, and I stopped counting.
		4383
		4384	=back
		4385
		4386
		4387	=head1 MACRO MAGIC
		4388
		4389	Libev can be compiled with a variety of options, the most fundamental
		4390	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		4391	functions and callbacks have an initial C<struct ev_loop *> argument.
		4392
		4393	To make it easier to write programs that cope with either variant, the
		4394	following macros are defined:
		4395
		4396	=over 4
		4397
		4398	=item C<EV_A>, C<EV_A_>
		4399
		4400	This provides the loop I<argument> for functions, if one is required ("ev
		4401	loop argument"). The C<EV_A> form is used when this is the sole argument,
		4402	C<EV_A_> is used when other arguments are following. Example:
		4403
		4404	ev_unref (EV_A);
		4405	ev_timer_add (EV_A_ watcher);
		4406	ev_run (EV_A_ 0);
		4407
		4408	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		4409	which is often provided by the following macro.
		4410
		4411	=item C<EV_P>, C<EV_P_>
		4412
		4413	This provides the loop I<parameter> for functions, if one is required ("ev
		4414	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		4415	C<EV_P_> is used when other parameters are following. Example:
		4416
		4417	// this is how ev_unref is being declared
		4418	static void ev_unref (EV_P);
		4419
		4420	// this is how you can declare your typical callback
		4421	static void cb (EV_P_ ev_timer *w, int revents)
		4422
		4423	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		4424	suitable for use with C<EV_A>.
		4425
		4426	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		4427
		4428	Similar to the other two macros, this gives you the value of the default
		4429	loop, if multiple loops are supported ("ev loop default"). The default loop
		4430	will be initialised if it isn't already initialised.
		4431
		4432	For non-multiplicity builds, these macros do nothing, so you always have
		4433	to initialise the loop somewhere.
		4434
		4435	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		4436
		4437	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		4438	default loop has been initialised (C<UC> == unchecked). Their behaviour
		4439	is undefined when the default loop has not been initialised by a previous
		4440	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		4441
		4442	It is often prudent to use C<EV_DEFAULT> when initialising the first
		4443	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		4444
		4445	=back
		4446
		4447	Example: Declare and initialise a check watcher, utilising the above
		4448	macros so it will work regardless of whether multiple loops are supported
		4449	or not.
		4450
		4451	static void
		4452	check_cb (EV_P_ ev_timer *w, int revents)
1422	{	4453	{
1423	io.start (fd, ev::READ);	4454	ev_check_stop (EV_A_ w);
1424	}	4455	}
		4456
		4457	ev_check check;
		4458	ev_check_init (&check, check_cb);
		4459	ev_check_start (EV_DEFAULT_ &check);
		4460	ev_run (EV_DEFAULT_ 0);
1425		4461
1426	=head1 EMBEDDING	4462	=head1 EMBEDDING
1427		4463
1428	Libev can (and often is) directly embedded into host	4464	Libev can (and often is) directly embedded into host
1429	applications. Examples of applications that embed it include the Deliantra	4465	applications. Examples of applications that embed it include the Deliantra
1430	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)	4466	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
1431	and rxvt-unicode.	4467	and rxvt-unicode.
1432		4468
1433	The goal is to enable you to just copy the neecssary files into your	4469	The goal is to enable you to just copy the necessary files into your
1434	source directory without having to change even a single line in them, so	4470	source directory without having to change even a single line in them, so
1435	you can easily upgrade by simply copying (or having a checked-out copy of	4471	you can easily upgrade by simply copying (or having a checked-out copy of
1436	libev somewhere in your source tree).	4472	libev somewhere in your source tree).
1437		4473
1438	=head2 FILESETS	4474	=head2 FILESETS
1439		4475
1440	Depending on what features you need you need to include one or more sets of files	4476	Depending on what features you need you need to include one or more sets of files
1441	in your app.	4477	in your application.
1442		4478
1443	=head3 CORE EVENT LOOP	4479	=head3 CORE EVENT LOOP
1444		4480
1445	To include only the libev core (all the C<ev_*> functions), with manual	4481	To include only the libev core (all the C<ev_*> functions), with manual
1446	configuration (no autoconf):	4482	configuration (no autoconf):
1447		4483
1448	#define EV_STANDALONE 1	4484	#define EV_STANDALONE 1
1449	#include "ev.c"	4485	#include "ev.c"
1450		4486
1451	This will automatically include F<ev.h>, too, and should be done in a	4487	This will automatically include F<ev.h>, too, and should be done in a
1452	single C source file only to provide the function implementations. To use	4488	single C source file only to provide the function implementations. To use
1453	it, do the same for F<ev.h> in all files wishing to use this API (best	4489	it, do the same for F<ev.h> in all files wishing to use this API (best
1454	done by writing a wrapper around F<ev.h> that you can include instead and	4490	done by writing a wrapper around F<ev.h> that you can include instead and
1455	where you can put other configuration options):	4491	where you can put other configuration options):
1456		4492
1457	#define EV_STANDALONE 1	4493	#define EV_STANDALONE 1
1458	#include "ev.h"	4494	#include "ev.h"
1459		4495
1460	Both header files and implementation files can be compiled with a C++	4496	Both header files and implementation files can be compiled with a C++
1461	compiler (at least, thats a stated goal, and breakage will be treated	4497	compiler (at least, that's a stated goal, and breakage will be treated
1462	as a bug).	4498	as a bug).
1463		4499
1464	You need the following files in your source tree, or in a directory	4500	You need the following files in your source tree, or in a directory
1465	in your include path (e.g. in libev/ when using -Ilibev):	4501	in your include path (e.g. in libev/ when using -Ilibev):
1466		4502
1467	ev.h	4503	ev.h
1468	ev.c	4504	ev.c
1469	ev_vars.h	4505	ev_vars.h
1470	ev_wrap.h	4506	ev_wrap.h
1471		4507
1472	ev_win32.c required on win32 platforms only	4508	ev_win32.c required on win32 platforms only
1473		4509
1474	ev_select.c only when select backend is enabled (which is by default)	4510	ev_select.c only when select backend is enabled
1475	ev_poll.c only when poll backend is enabled (disabled by default)	4511	ev_poll.c only when poll backend is enabled
1476	ev_epoll.c only when the epoll backend is enabled (disabled by default)	4512	ev_epoll.c only when the epoll backend is enabled
		4513	ev_linuxaio.c only when the linux aio backend is enabled
		4514	ev_iouring.c only when the linux io_uring backend is enabled
1477	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)	4515	ev_kqueue.c only when the kqueue backend is enabled
1478	ev_port.c only when the solaris port backend is enabled (disabled by default)	4516	ev_port.c only when the solaris port backend is enabled
1479		4517
1480	F<ev.c> includes the backend files directly when enabled, so you only need	4518	F<ev.c> includes the backend files directly when enabled, so you only need
1481	to compile this single file.	4519	to compile this single file.
1482		4520
1483	=head3 LIBEVENT COMPATIBILITY API	4521	=head3 LIBEVENT COMPATIBILITY API
1484		4522
1485	To include the libevent compatibility API, also include:	4523	To include the libevent compatibility API, also include:
1486		4524
1487	#include "event.c"	4525	#include "event.c"
1488		4526
1489	in the file including F<ev.c>, and:	4527	in the file including F<ev.c>, and:
1490		4528
1491	#include "event.h"	4529	#include "event.h"
1492		4530
1493	in the files that want to use the libevent API. This also includes F<ev.h>.	4531	in the files that want to use the libevent API. This also includes F<ev.h>.
1494		4532
1495	You need the following additional files for this:	4533	You need the following additional files for this:
1496		4534
1497	event.h	4535	event.h
1498	event.c	4536	event.c
1499		4537
1500	=head3 AUTOCONF SUPPORT	4538	=head3 AUTOCONF SUPPORT
1501		4539
1502	Instead of using C<EV_STANDALONE=1> and providing your config in	4540	Instead of using C<EV_STANDALONE=1> and providing your configuration in
1503	whatever way you want, you can also C<m4_include([libev.m4])> in your	4541	whatever way you want, you can also C<m4_include([libev.m4])> in your
1504	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then	4542	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
1505	include F<config.h> and configure itself accordingly.	4543	include F<config.h> and configure itself accordingly.
1506		4544
1507	For this of course you need the m4 file:	4545	For this of course you need the m4 file:
1508		4546
1509	libev.m4	4547	libev.m4
1510		4548
1511	=head2 PREPROCESSOR SYMBOLS/MACROS	4549	=head2 PREPROCESSOR SYMBOLS/MACROS
1512		4550
1513	Libev can be configured via a variety of preprocessor symbols you have to define	4551	Libev can be configured via a variety of preprocessor symbols you have to
1514	before including any of its files. The default is not to build for multiplicity	4552	define before including (or compiling) any of its files. The default in
1515	and only include the select backend.	4553	the absence of autoconf is documented for every option.
		4554
		4555	Symbols marked with "(h)" do not change the ABI, and can have different
		4556	values when compiling libev vs. including F<ev.h>, so it is permissible
		4557	to redefine them before including F<ev.h> without breaking compatibility
		4558	to a compiled library. All other symbols change the ABI, which means all
		4559	users of libev and the libev code itself must be compiled with compatible
		4560	settings.
1516		4561
1517	=over 4	4562	=over 4
1518		4563
		4564	=item EV_COMPAT3 (h)
		4565
		4566	Backwards compatibility is a major concern for libev. This is why this
		4567	release of libev comes with wrappers for the functions and symbols that
		4568	have been renamed between libev version 3 and 4.
		4569
		4570	You can disable these wrappers (to test compatibility with future
		4571	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4572	sources. This has the additional advantage that you can drop the C<struct>
		4573	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4574	typedef in that case.
		4575
		4576	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4577	and in some even more future version the compatibility code will be
		4578	removed completely.
		4579
1519	=item EV_STANDALONE	4580	=item EV_STANDALONE (h)
1520		4581
1521	Must always be C<1> if you do not use autoconf configuration, which	4582	Must always be C<1> if you do not use autoconf configuration, which
1522	keeps libev from including F<config.h>, and it also defines dummy	4583	keeps libev from including F<config.h>, and it also defines dummy
1523	implementations for some libevent functions (such as logging, which is not	4584	implementations for some libevent functions (such as logging, which is not
1524	supported). It will also not define any of the structs usually found in	4585	supported). It will also not define any of the structs usually found in
1525	F<event.h> that are not directly supported by the libev core alone.	4586	F<event.h> that are not directly supported by the libev core alone.
1526		4587
		4588	In standalone mode, libev will still try to automatically deduce the
		4589	configuration, but has to be more conservative.
		4590
		4591	=item EV_USE_FLOOR
		4592
		4593	If defined to be C<1>, libev will use the C<floor ()> function for its
		4594	periodic reschedule calculations, otherwise libev will fall back on a
		4595	portable (slower) implementation. If you enable this, you usually have to
		4596	link against libm or something equivalent. Enabling this when the C<floor>
		4597	function is not available will fail, so the safe default is to not enable
		4598	this.
		4599
1527	=item EV_USE_MONOTONIC	4600	=item EV_USE_MONOTONIC
1528		4601
1529	If defined to be C<1>, libev will try to detect the availability of the	4602	If defined to be C<1>, libev will try to detect the availability of the
1530	monotonic clock option at both compiletime and runtime. Otherwise no use	4603	monotonic clock option at both compile time and runtime. Otherwise no
1531	of the monotonic clock option will be attempted. If you enable this, you	4604	use of the monotonic clock option will be attempted. If you enable this,
1532	usually have to link against librt or something similar. Enabling it when	4605	you usually have to link against librt or something similar. Enabling it
1533	the functionality isn't available is safe, though, althoguh you have	4606	when the functionality isn't available is safe, though, although you have
1534	to make sure you link against any libraries where the C<clock_gettime>	4607	to make sure you link against any libraries where the C<clock_gettime>
1535	function is hiding in (often F<-lrt>).	4608	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
1536		4609
1537	=item EV_USE_REALTIME	4610	=item EV_USE_REALTIME
1538		4611
1539	If defined to be C<1>, libev will try to detect the availability of the	4612	If defined to be C<1>, libev will try to detect the availability of the
1540	realtime clock option at compiletime (and assume its availability at	4613	real-time clock option at compile time (and assume its availability
1541	runtime if successful). Otherwise no use of the realtime clock option will	4614	at runtime if successful). Otherwise no use of the real-time clock
1542	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	4615	option will be attempted. This effectively replaces C<gettimeofday>
1543	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See tzhe note about libraries	4616	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
1544	in the description of C<EV_USE_MONOTONIC>, though.	4617	correctness. See the note about libraries in the description of
		4618	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4619	C<EV_USE_CLOCK_SYSCALL>.
		4620
		4621	=item EV_USE_CLOCK_SYSCALL
		4622
		4623	If defined to be C<1>, libev will try to use a direct syscall instead
		4624	of calling the system-provided C<clock_gettime> function. This option
		4625	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4626	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4627	programs needlessly. Using a direct syscall is slightly slower (in
		4628	theory), because no optimised vdso implementation can be used, but avoids
		4629	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4630	higher, as it simplifies linking (no need for C<-lrt>).
		4631
		4632	=item EV_USE_NANOSLEEP
		4633
		4634	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4635	and will use it for delays. Otherwise it will use C<select ()>.
		4636
		4637	=item EV_USE_EVENTFD
		4638
		4639	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4640	available and will probe for kernel support at runtime. This will improve
		4641	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4642	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4643	2.7 or newer, otherwise disabled.
		4644
		4645	=item EV_USE_SIGNALFD
		4646
		4647	If defined to be C<1>, then libev will assume that C<signalfd ()> is
		4648	available and will probe for kernel support at runtime. This enables
		4649	the use of EVFLAG_SIGNALFD for faster and simpler signal handling. If
		4650	undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4651	2.7 or newer, otherwise disabled.
		4652
		4653	=item EV_USE_TIMERFD
		4654
		4655	If defined to be C<1>, then libev will assume that C<timerfd ()> is
		4656	available and will probe for kernel support at runtime. This allows
		4657	libev to detect time jumps accurately. If undefined, it will be enabled
		4658	if the headers indicate GNU/Linux + Glibc 2.8 or newer and define
		4659	C<TFD_TIMER_CANCEL_ON_SET>, otherwise disabled.
		4660
		4661	=item EV_USE_EVENTFD
		4662
		4663	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4664	available and will probe for kernel support at runtime. This will improve
		4665	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4666	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4667	2.7 or newer, otherwise disabled.
1545		4668
1546	=item EV_USE_SELECT	4669	=item EV_USE_SELECT
1547		4670
1548	If undefined or defined to be C<1>, libev will compile in support for the	4671	If undefined or defined to be C<1>, libev will compile in support for the
1549	C<select>(2) backend. No attempt at autodetection will be done: if no	4672	C<select>(2) backend. No attempt at auto-detection will be done: if no
1550	other method takes over, select will be it. Otherwise the select backend	4673	other method takes over, select will be it. Otherwise the select backend
1551	will not be compiled in.	4674	will not be compiled in.
1552		4675
1553	=item EV_SELECT_USE_FD_SET	4676	=item EV_SELECT_USE_FD_SET
1554		4677
1555	If defined to C<1>, then the select backend will use the system C<fd_set>	4678	If defined to C<1>, then the select backend will use the system C<fd_set>
1556	structure. This is useful if libev doesn't compile due to a missing	4679	structure. This is useful if libev doesn't compile due to a missing
1557	C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on	4680	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
1558	exotic systems. This usually limits the range of file descriptors to some	4681	on exotic systems. This usually limits the range of file descriptors to
1559	low limit such as 1024 or might have other limitations (winsocket only	4682	some low limit such as 1024 or might have other limitations (winsocket
1560	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might	4683	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
1561	influence the size of the C<fd_set> used.	4684	configures the maximum size of the C<fd_set>.
1562		4685
1563	=item EV_SELECT_IS_WINSOCKET	4686	=item EV_SELECT_IS_WINSOCKET
1564		4687
1565	When defined to C<1>, the select backend will assume that	4688	When defined to C<1>, the select backend will assume that
1566	select/socket/connect etc. don't understand file descriptors but	4689	select/socket/connect etc. don't understand file descriptors but
…		…
1568	be used is the winsock select). This means that it will call	4691	be used is the winsock select). This means that it will call
1569	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,	4692	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
1570	it is assumed that all these functions actually work on fds, even	4693	it is assumed that all these functions actually work on fds, even
1571	on win32. Should not be defined on non-win32 platforms.	4694	on win32. Should not be defined on non-win32 platforms.
1572		4695
		4696	=item EV_FD_TO_WIN32_HANDLE(fd)
		4697
		4698	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4699	file descriptors to socket handles. When not defining this symbol (the
		4700	default), then libev will call C<_get_osfhandle>, which is usually
		4701	correct. In some cases, programs use their own file descriptor management,
		4702	in which case they can provide this function to map fds to socket handles.
		4703
		4704	=item EV_WIN32_HANDLE_TO_FD(handle)
		4705
		4706	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4707	using the standard C<_open_osfhandle> function. For programs implementing
		4708	their own fd to handle mapping, overwriting this function makes it easier
		4709	to do so. This can be done by defining this macro to an appropriate value.
		4710
		4711	=item EV_WIN32_CLOSE_FD(fd)
		4712
		4713	If programs implement their own fd to handle mapping on win32, then this
		4714	macro can be used to override the C<close> function, useful to unregister
		4715	file descriptors again. Note that the replacement function has to close
		4716	the underlying OS handle.
		4717
		4718	=item EV_USE_WSASOCKET
		4719
		4720	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4721	communication socket, which works better in some environments. Otherwise,
		4722	the normal C<socket> function will be used, which works better in other
		4723	environments.
		4724
1573	=item EV_USE_POLL	4725	=item EV_USE_POLL
1574		4726
1575	If defined to be C<1>, libev will compile in support for the C<poll>(2)	4727	If defined to be C<1>, libev will compile in support for the C<poll>(2)
1576	backend. Otherwise it will be enabled on non-win32 platforms. It	4728	backend. Otherwise it will be enabled on non-win32 platforms. It
1577	takes precedence over select.	4729	takes precedence over select.
1578		4730
1579	=item EV_USE_EPOLL	4731	=item EV_USE_EPOLL
1580		4732
1581	If defined to be C<1>, libev will compile in support for the Linux	4733	If defined to be C<1>, libev will compile in support for the Linux
1582	C<epoll>(7) backend. Its availability will be detected at runtime,	4734	C<epoll>(7) backend. Its availability will be detected at runtime,
1583	otherwise another method will be used as fallback. This is the	4735	otherwise another method will be used as fallback. This is the preferred
1584	preferred backend for GNU/Linux systems.	4736	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4737	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4738
		4739	=item EV_USE_LINUXAIO
		4740
		4741	If defined to be C<1>, libev will compile in support for the Linux aio
		4742	backend (C<EV_USE_EPOLL> must also be enabled). If undefined, it will be
		4743	enabled on linux, otherwise disabled.
		4744
		4745	=item EV_USE_IOURING
		4746
		4747	If defined to be C<1>, libev will compile in support for the Linux
		4748	io_uring backend (C<EV_USE_EPOLL> must also be enabled). Due to it's
		4749	current limitations it has to be requested explicitly. If undefined, it
		4750	will be enabled on linux, otherwise disabled.
1585		4751
1586	=item EV_USE_KQUEUE	4752	=item EV_USE_KQUEUE
1587		4753
1588	If defined to be C<1>, libev will compile in support for the BSD style	4754	If defined to be C<1>, libev will compile in support for the BSD style
1589	C<kqueue>(2) backend. Its actual availability will be detected at runtime,	4755	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
…		…
1602	otherwise another method will be used as fallback. This is the preferred	4768	otherwise another method will be used as fallback. This is the preferred
1603	backend for Solaris 10 systems.	4769	backend for Solaris 10 systems.
1604		4770
1605	=item EV_USE_DEVPOLL	4771	=item EV_USE_DEVPOLL
1606		4772
1607	reserved for future expansion, works like the USE symbols above.	4773	Reserved for future expansion, works like the USE symbols above.
1608		4774
		4775	=item EV_USE_INOTIFY
		4776
		4777	If defined to be C<1>, libev will compile in support for the Linux inotify
		4778	interface to speed up C<ev_stat> watchers. Its actual availability will
		4779	be detected at runtime. If undefined, it will be enabled if the headers
		4780	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4781
		4782	=item EV_NO_SMP
		4783
		4784	If defined to be C<1>, libev will assume that memory is always coherent
		4785	between threads, that is, threads can be used, but threads never run on
		4786	different cpus (or different cpu cores). This reduces dependencies
		4787	and makes libev faster.
		4788
		4789	=item EV_NO_THREADS
		4790
		4791	If defined to be C<1>, libev will assume that it will never be called from
		4792	different threads (that includes signal handlers), which is a stronger
		4793	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4794	libev faster.
		4795
		4796	=item EV_ATOMIC_T
		4797
		4798	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4799	access is atomic with respect to other threads or signal contexts. No
		4800	such type is easily found in the C language, so you can provide your own
		4801	type that you know is safe for your purposes. It is used both for signal
		4802	handler "locking" as well as for signal and thread safety in C<ev_async>
		4803	watchers.
		4804
		4805	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4806	(from F<signal.h>), which is usually good enough on most platforms.
		4807
1609	=item EV_H	4808	=item EV_H (h)
1610		4809
1611	The name of the F<ev.h> header file used to include it. The default if	4810	The name of the F<ev.h> header file used to include it. The default if
1612	undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This	4811	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
1613	can be used to virtually rename the F<ev.h> header file in case of conflicts.	4812	used to virtually rename the F<ev.h> header file in case of conflicts.
1614		4813
1615	=item EV_CONFIG_H	4814	=item EV_CONFIG_H (h)
1616		4815
1617	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override	4816	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
1618	F<ev.c>'s idea of where to find the F<config.h> file, similarly to	4817	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
1619	C<EV_H>, above.	4818	C<EV_H>, above.
1620		4819
1621	=item EV_EVENT_H	4820	=item EV_EVENT_H (h)
1622		4821
1623	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea	4822	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
1624	of how the F<event.h> header can be found.	4823	of how the F<event.h> header can be found, the default is C<"event.h">.
1625		4824
1626	=item EV_PROTOTYPES	4825	=item EV_PROTOTYPES (h)
1627		4826
1628	If defined to be C<0>, then F<ev.h> will not define any function	4827	If defined to be C<0>, then F<ev.h> will not define any function
1629	prototypes, but still define all the structs and other symbols. This is	4828	prototypes, but still define all the structs and other symbols. This is
1630	occasionally useful if you want to provide your own wrapper functions	4829	occasionally useful if you want to provide your own wrapper functions
1631	around libev functions.	4830	around libev functions.
…		…
1636	will have the C<struct ev_loop *> as first argument, and you can create	4835	will have the C<struct ev_loop *> as first argument, and you can create
1637	additional independent event loops. Otherwise there will be no support	4836	additional independent event loops. Otherwise there will be no support
1638	for multiple event loops and there is no first event loop pointer	4837	for multiple event loops and there is no first event loop pointer
1639	argument. Instead, all functions act on the single default loop.	4838	argument. Instead, all functions act on the single default loop.
1640		4839
1641	=item EV_PERIODICS	4840	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4841	default loop when multiplicity is switched off - you always have to
		4842	initialise the loop manually in this case.
1642		4843
1643	If undefined or defined to be C<1>, then periodic timers are supported,	4844	=item EV_MINPRI
1644	otherwise not. This saves a few kb of code.	4845
		4846	=item EV_MAXPRI
		4847
		4848	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4849	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4850	provide for more priorities by overriding those symbols (usually defined
		4851	to be C<-2> and C<2>, respectively).
		4852
		4853	When doing priority-based operations, libev usually has to linearly search
		4854	all the priorities, so having many of them (hundreds) uses a lot of space
		4855	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4856	fine.
		4857
		4858	If your embedding application does not need any priorities, defining these
		4859	both to C<0> will save some memory and CPU.
		4860
		4861	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4862	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4863	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4864
		4865	If undefined or defined to be C<1> (and the platform supports it), then
		4866	the respective watcher type is supported. If defined to be C<0>, then it
		4867	is not. Disabling watcher types mainly saves code size.
		4868
		4869	=item EV_FEATURES
		4870
		4871	If you need to shave off some kilobytes of code at the expense of some
		4872	speed (but with the full API), you can define this symbol to request
		4873	certain subsets of functionality. The default is to enable all features
		4874	that can be enabled on the platform.
		4875
		4876	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4877	with some broad features you want) and then selectively re-enable
		4878	additional parts you want, for example if you want everything minimal,
		4879	but multiple event loop support, async and child watchers and the poll
		4880	backend, use this:
		4881
		4882	#define EV_FEATURES 0
		4883	#define EV_MULTIPLICITY 1
		4884	#define EV_USE_POLL 1
		4885	#define EV_CHILD_ENABLE 1
		4886	#define EV_ASYNC_ENABLE 1
		4887
		4888	The actual value is a bitset, it can be a combination of the following
		4889	values (by default, all of these are enabled):
		4890
		4891	=over 4
		4892
		4893	=item C<1> - faster/larger code
		4894
		4895	Use larger code to speed up some operations.
		4896
		4897	Currently this is used to override some inlining decisions (enlarging the
		4898	code size by roughly 30% on amd64).
		4899
		4900	When optimising for size, use of compiler flags such as C<-Os> with
		4901	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4902	assertions.
		4903
		4904	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4905	(e.g. gcc with C<-Os>).
		4906
		4907	=item C<2> - faster/larger data structures
		4908
		4909	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4910	hash table sizes and so on. This will usually further increase code size
		4911	and can additionally have an effect on the size of data structures at
		4912	runtime.
		4913
		4914	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4915	(e.g. gcc with C<-Os>).
		4916
		4917	=item C<4> - full API configuration
		4918
		4919	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4920	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4921
		4922	=item C<8> - full API
		4923
		4924	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4925	details on which parts of the API are still available without this
		4926	feature, and do not complain if this subset changes over time.
		4927
		4928	=item C<16> - enable all optional watcher types
		4929
		4930	Enables all optional watcher types. If you want to selectively enable
		4931	only some watcher types other than I/O and timers (e.g. prepare,
		4932	embed, async, child...) you can enable them manually by defining
		4933	C<EV_watchertype_ENABLE> to C<1> instead.
		4934
		4935	=item C<32> - enable all backends
		4936
		4937	This enables all backends - without this feature, you need to enable at
		4938	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4939
		4940	=item C<64> - enable OS-specific "helper" APIs
		4941
		4942	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4943	default.
		4944
		4945	=back
		4946
		4947	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4948	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4949	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4950	watchers, timers and monotonic clock support.
		4951
		4952	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4953	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4954	your program might be left out as well - a binary starting a timer and an
		4955	I/O watcher then might come out at only 5Kb.
		4956
		4957	=item EV_API_STATIC
		4958
		4959	If this symbol is defined (by default it is not), then all identifiers
		4960	will have static linkage. This means that libev will not export any
		4961	identifiers, and you cannot link against libev anymore. This can be useful
		4962	when you embed libev, only want to use libev functions in a single file,
		4963	and do not want its identifiers to be visible.
		4964
		4965	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4966	wants to use libev.
		4967
		4968	This option only works when libev is compiled with a C compiler, as C++
		4969	doesn't support the required declaration syntax.
		4970
		4971	=item EV_AVOID_STDIO
		4972
		4973	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4974	functions (printf, scanf, perror etc.). This will increase the code size
		4975	somewhat, but if your program doesn't otherwise depend on stdio and your
		4976	libc allows it, this avoids linking in the stdio library which is quite
		4977	big.
		4978
		4979	Note that error messages might become less precise when this option is
		4980	enabled.
		4981
		4982	=item EV_NSIG
		4983
		4984	The highest supported signal number, +1 (or, the number of
		4985	signals): Normally, libev tries to deduce the maximum number of signals
		4986	automatically, but sometimes this fails, in which case it can be
		4987	specified. Also, using a lower number than detected (C<32> should be
		4988	good for about any system in existence) can save some memory, as libev
		4989	statically allocates some 12-24 bytes per signal number.
		4990
		4991	=item EV_PID_HASHSIZE
		4992
		4993	C<ev_child> watchers use a small hash table to distribute workload by
		4994	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4995	usually more than enough. If you need to manage thousands of children you
		4996	might want to increase this value (I<must> be a power of two).
		4997
		4998	=item EV_INOTIFY_HASHSIZE
		4999
		5000	C<ev_stat> watchers use a small hash table to distribute workload by
		5001	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		5002	disabled), usually more than enough. If you need to manage thousands of
		5003	C<ev_stat> watchers you might want to increase this value (I<must> be a
		5004	power of two).
		5005
		5006	=item EV_USE_4HEAP
		5007
		5008	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		5009	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		5010	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		5011	faster performance with many (thousands) of watchers.
		5012
		5013	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		5014	will be C<0>.
		5015
		5016	=item EV_HEAP_CACHE_AT
		5017
		5018	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		5019	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		5020	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		5021	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		5022	but avoids random read accesses on heap changes. This improves performance
		5023	noticeably with many (hundreds) of watchers.
		5024
		5025	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		5026	will be C<0>.
		5027
		5028	=item EV_VERIFY
		5029
		5030	Controls how much internal verification (see C<ev_verify ()>) will
		5031	be done: If set to C<0>, no internal verification code will be compiled
		5032	in. If set to C<1>, then verification code will be compiled in, but not
		5033	called. If set to C<2>, then the internal verification code will be
		5034	called once per loop, which can slow down libev. If set to C<3>, then the
		5035	verification code will be called very frequently, which will slow down
		5036	libev considerably.
		5037
		5038	Verification errors are reported via C's C<assert> mechanism, so if you
		5039	disable that (e.g. by defining C<NDEBUG>) then no errors will be reported.
		5040
		5041	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		5042	will be C<0>.
1645		5043
1646	=item EV_COMMON	5044	=item EV_COMMON
1647		5045
1648	By default, all watchers have a C<void *data> member. By redefining	5046	By default, all watchers have a C<void *data> member. By redefining
1649	this macro to a something else you can include more and other types of	5047	this macro to something else you can include more and other types of
1650	members. You have to define it each time you include one of the files,	5048	members. You have to define it each time you include one of the files,
1651	though, and it must be identical each time.	5049	though, and it must be identical each time.
1652		5050
1653	For example, the perl EV module uses something like this:	5051	For example, the perl EV module uses something like this:
1654		5052
1655	#define EV_COMMON \	5053	#define EV_COMMON \
1656	SV *self; /* contains this struct */ \	5054	SV *self; /* contains this struct */ \
1657	SV *cb_sv, *fh /* note no trailing ";" */	5055	SV *cb_sv, *fh /* note no trailing ";" */
1658		5056
1659	=item EV_CB_DECLARE(type)	5057	=item EV_CB_DECLARE (type)
1660		5058
1661	=item EV_CB_INVOKE(watcher,revents)	5059	=item EV_CB_INVOKE (watcher, revents)
1662		5060
1663	=item ev_set_cb(ev,cb)	5061	=item ev_set_cb (ev, cb)
1664		5062
1665	Can be used to change the callback member declaration in each watcher,	5063	Can be used to change the callback member declaration in each watcher,
1666	and the way callbacks are invoked and set. Must expand to a struct member	5064	and the way callbacks are invoked and set. Must expand to a struct member
1667	definition and a statement, respectively. See the F<ev.v> header file for	5065	definition and a statement, respectively. See the F<ev.h> header file for
1668	their default definitions. One possible use for overriding these is to	5066	their default definitions. One possible use for overriding these is to
1669	avoid the ev_loop pointer as first argument in all cases, or to use method	5067	avoid the C<struct ev_loop *> as first argument in all cases, or to use
1670	calls instead of plain function calls in C++.	5068	method calls instead of plain function calls in C++.
		5069
		5070	=back
		5071
		5072	=head2 EXPORTED API SYMBOLS
		5073
		5074	If you need to re-export the API (e.g. via a DLL) and you need a list of
		5075	exported symbols, you can use the provided F<Symbol.*> files which list
		5076	all public symbols, one per line:
		5077
		5078	Symbols.ev for libev proper
		5079	Symbols.event for the libevent emulation
		5080
		5081	This can also be used to rename all public symbols to avoid clashes with
		5082	multiple versions of libev linked together (which is obviously bad in
		5083	itself, but sometimes it is inconvenient to avoid this).
		5084
		5085	A sed command like this will create wrapper C<#define>'s that you need to
		5086	include before including F<ev.h>:
		5087
		5088	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		5089
		5090	This would create a file F<wrap.h> which essentially looks like this:
		5091
		5092	#define ev_backend myprefix_ev_backend
		5093	#define ev_check_start myprefix_ev_check_start
		5094	#define ev_check_stop myprefix_ev_check_stop
		5095	...
1671		5096
1672	=head2 EXAMPLES	5097	=head2 EXAMPLES
1673		5098
1674	For a real-world example of a program the includes libev	5099	For a real-world example of a program the includes libev
1675	verbatim, you can have a look at the EV perl module	5100	verbatim, you can have a look at the EV perl module
…		…
1678	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file	5103	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
1679	will be compiled. It is pretty complex because it provides its own header	5104	will be compiled. It is pretty complex because it provides its own header
1680	file.	5105	file.
1681		5106
1682	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file	5107	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
1683	that everybody includes and which overrides some autoconf choices:	5108	that everybody includes and which overrides some configure choices:
1684		5109
		5110	#define EV_FEATURES 8
		5111	#define EV_USE_SELECT 1
		5112	#define EV_PREPARE_ENABLE 1
		5113	#define EV_IDLE_ENABLE 1
		5114	#define EV_SIGNAL_ENABLE 1
		5115	#define EV_CHILD_ENABLE 1
1685	#define EV_USE_POLL 0	5116	#define EV_USE_STDEXCEPT 0
1686	#define EV_MULTIPLICITY 0
1687	#define EV_PERIODICS 0
1688	#define EV_CONFIG_H <config.h>	5117	#define EV_CONFIG_H <config.h>
1689		5118
1690	#include "ev++.h"	5119	#include "ev++.h"
1691		5120
1692	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	5121	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
1693		5122
1694	#include "ev_cpp.h"	5123	#include "ev_cpp.h"
1695	#include "ev.c"	5124	#include "ev.c"
		5125
		5126	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
		5127
		5128	=head2 THREADS AND COROUTINES
		5129
		5130	=head3 THREADS
		5131
		5132	All libev functions are reentrant and thread-safe unless explicitly
		5133	documented otherwise, but libev implements no locking itself. This means
		5134	that you can use as many loops as you want in parallel, as long as there
		5135	are no concurrent calls into any libev function with the same loop
		5136	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		5137	of course): libev guarantees that different event loops share no data
		5138	structures that need any locking.
		5139
		5140	Or to put it differently: calls with different loop parameters can be done
		5141	concurrently from multiple threads, calls with the same loop parameter
		5142	must be done serially (but can be done from different threads, as long as
		5143	only one thread ever is inside a call at any point in time, e.g. by using
		5144	a mutex per loop).
		5145
		5146	Specifically to support threads (and signal handlers), libev implements
		5147	so-called C<ev_async> watchers, which allow some limited form of
		5148	concurrency on the same event loop, namely waking it up "from the
		5149	outside".
		5150
		5151	If you want to know which design (one loop, locking, or multiple loops
		5152	without or something else still) is best for your problem, then I cannot
		5153	help you, but here is some generic advice:
		5154
		5155	=over 4
		5156
		5157	=item * most applications have a main thread: use the default libev loop
		5158	in that thread, or create a separate thread running only the default loop.
		5159
		5160	This helps integrating other libraries or software modules that use libev
		5161	themselves and don't care/know about threading.
		5162
		5163	=item * one loop per thread is usually a good model.
		5164
		5165	Doing this is almost never wrong, sometimes a better-performance model
		5166	exists, but it is always a good start.
		5167
		5168	=item * other models exist, such as the leader/follower pattern, where one
		5169	loop is handed through multiple threads in a kind of round-robin fashion.
		5170
		5171	Choosing a model is hard - look around, learn, know that usually you can do
		5172	better than you currently do :-)
		5173
		5174	=item * often you need to talk to some other thread which blocks in the
		5175	event loop.
		5176
		5177	C<ev_async> watchers can be used to wake them up from other threads safely
		5178	(or from signal contexts...).
		5179
		5180	An example use would be to communicate signals or other events that only
		5181	work in the default loop by registering the signal watcher with the
		5182	default loop and triggering an C<ev_async> watcher from the default loop
		5183	watcher callback into the event loop interested in the signal.
		5184
		5185	=back
		5186
		5187	See also L</THREAD LOCKING EXAMPLE>.
		5188
		5189	=head3 COROUTINES
		5190
		5191	Libev is very accommodating to coroutines ("cooperative threads"):
		5192	libev fully supports nesting calls to its functions from different
		5193	coroutines (e.g. you can call C<ev_run> on the same loop from two
		5194	different coroutines, and switch freely between both coroutines running
		5195	the loop, as long as you don't confuse yourself). The only exception is
		5196	that you must not do this from C<ev_periodic> reschedule callbacks.
		5197
		5198	Care has been taken to ensure that libev does not keep local state inside
		5199	C<ev_run>, and other calls do not usually allow for coroutine switches as
		5200	they do not call any callbacks.
		5201
		5202	=head2 COMPILER WARNINGS
		5203
		5204	Depending on your compiler and compiler settings, you might get no or a
		5205	lot of warnings when compiling libev code. Some people are apparently
		5206	scared by this.
		5207
		5208	However, these are unavoidable for many reasons. For one, each compiler
		5209	has different warnings, and each user has different tastes regarding
		5210	warning options. "Warn-free" code therefore cannot be a goal except when
		5211	targeting a specific compiler and compiler-version.
		5212
		5213	Another reason is that some compiler warnings require elaborate
		5214	workarounds, or other changes to the code that make it less clear and less
		5215	maintainable.
		5216
		5217	And of course, some compiler warnings are just plain stupid, or simply
		5218	wrong (because they don't actually warn about the condition their message
		5219	seems to warn about). For example, certain older gcc versions had some
		5220	warnings that resulted in an extreme number of false positives. These have
		5221	been fixed, but some people still insist on making code warn-free with
		5222	such buggy versions.
		5223
		5224	While libev is written to generate as few warnings as possible,
		5225	"warn-free" code is not a goal, and it is recommended not to build libev
		5226	with any compiler warnings enabled unless you are prepared to cope with
		5227	them (e.g. by ignoring them). Remember that warnings are just that:
		5228	warnings, not errors, or proof of bugs.
		5229
		5230
		5231	=head2 VALGRIND
		5232
		5233	Valgrind has a special section here because it is a popular tool that is
		5234	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		5235
		5236	If you think you found a bug (memory leak, uninitialised data access etc.)
		5237	in libev, then check twice: If valgrind reports something like:
		5238
		5239	==2274== definitely lost: 0 bytes in 0 blocks.
		5240	==2274== possibly lost: 0 bytes in 0 blocks.
		5241	==2274== still reachable: 256 bytes in 1 blocks.
		5242
		5243	Then there is no memory leak, just as memory accounted to global variables
		5244	is not a memleak - the memory is still being referenced, and didn't leak.
		5245
		5246	Similarly, under some circumstances, valgrind might report kernel bugs
		5247	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		5248	although an acceptable workaround has been found here), or it might be
		5249	confused.
		5250
		5251	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		5252	make it into some kind of religion.
		5253
		5254	If you are unsure about something, feel free to contact the mailing list
		5255	with the full valgrind report and an explanation on why you think this
		5256	is a bug in libev (best check the archives, too :). However, don't be
		5257	annoyed when you get a brisk "this is no bug" answer and take the chance
		5258	of learning how to interpret valgrind properly.
		5259
		5260	If you need, for some reason, empty reports from valgrind for your project
		5261	I suggest using suppression lists.
		5262
		5263
		5264	=head1 PORTABILITY NOTES
		5265
		5266	=head2 GNU/LINUX 32 BIT LIMITATIONS
		5267
		5268	GNU/Linux is the only common platform that supports 64 bit file/large file
		5269	interfaces but I<disables> them by default.
		5270
		5271	That means that libev compiled in the default environment doesn't support
		5272	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		5273
		5274	Unfortunately, many programs try to work around this GNU/Linux issue
		5275	by enabling the large file API, which makes them incompatible with the
		5276	standard libev compiled for their system.
		5277
		5278	Likewise, libev cannot enable the large file API itself as this would
		5279	suddenly make it incompatible to the default compile time environment,
		5280	i.e. all programs not using special compile switches.
		5281
		5282	=head2 OS/X AND DARWIN BUGS
		5283
		5284	The whole thing is a bug if you ask me - basically any system interface
		5285	you touch is broken, whether it is locales, poll, kqueue or even the
		5286	OpenGL drivers.
		5287
		5288	=head3 C<kqueue> is buggy
		5289
		5290	The kqueue syscall is broken in all known versions - most versions support
		5291	only sockets, many support pipes.
		5292
		5293	Libev tries to work around this by not using C<kqueue> by default on this
		5294	rotten platform, but of course you can still ask for it when creating a
		5295	loop - embedding a socket-only kqueue loop into a select-based one is
		5296	probably going to work well.
		5297
		5298	=head3 C<poll> is buggy
		5299
		5300	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		5301	implementation by something calling C<kqueue> internally around the 10.5.6
		5302	release, so now C<kqueue> I<and> C<poll> are broken.
		5303
		5304	Libev tries to work around this by not using C<poll> by default on
		5305	this rotten platform, but of course you can still ask for it when creating
		5306	a loop.
		5307
		5308	=head3 C<select> is buggy
		5309
		5310	All that's left is C<select>, and of course Apple found a way to fuck this
		5311	one up as well: On OS/X, C<select> actively limits the number of file
		5312	descriptors you can pass in to 1024 - your program suddenly crashes when
		5313	you use more.
		5314
		5315	There is an undocumented "workaround" for this - defining
		5316	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		5317	work on OS/X.
		5318
		5319	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		5320
		5321	=head3 C<errno> reentrancy
		5322
		5323	The default compile environment on Solaris is unfortunately so
		5324	thread-unsafe that you can't even use components/libraries compiled
		5325	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		5326	defined by default. A valid, if stupid, implementation choice.
		5327
		5328	If you want to use libev in threaded environments you have to make sure
		5329	it's compiled with C<_REENTRANT> defined.
		5330
		5331	=head3 Event port backend
		5332
		5333	The scalable event interface for Solaris is called "event
		5334	ports". Unfortunately, this mechanism is very buggy in all major
		5335	releases. If you run into high CPU usage, your program freezes or you get
		5336	a large number of spurious wakeups, make sure you have all the relevant
		5337	and latest kernel patches applied. No, I don't know which ones, but there
		5338	are multiple ones to apply, and afterwards, event ports actually work
		5339	great.
		5340
		5341	If you can't get it to work, you can try running the program by setting
		5342	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		5343	C<select> backends.
		5344
		5345	=head2 AIX POLL BUG
		5346
		5347	AIX unfortunately has a broken C<poll.h> header. Libev works around
		5348	this by trying to avoid the poll backend altogether (i.e. it's not even
		5349	compiled in), which normally isn't a big problem as C<select> works fine
		5350	with large bitsets on AIX, and AIX is dead anyway.
		5351
		5352	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		5353
		5354	=head3 General issues
		5355
		5356	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		5357	requires, and its I/O model is fundamentally incompatible with the POSIX
		5358	model. Libev still offers limited functionality on this platform in
		5359	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		5360	descriptors. This only applies when using Win32 natively, not when using
		5361	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		5362	as every compiler comes with a slightly differently broken/incompatible
		5363	environment.
		5364
		5365	Lifting these limitations would basically require the full
		5366	re-implementation of the I/O system. If you are into this kind of thing,
		5367	then note that glib does exactly that for you in a very portable way (note
		5368	also that glib is the slowest event library known to man).
		5369
		5370	There is no supported compilation method available on windows except
		5371	embedding it into other applications.
		5372
		5373	Sensible signal handling is officially unsupported by Microsoft - libev
		5374	tries its best, but under most conditions, signals will simply not work.
		5375
		5376	Not a libev limitation but worth mentioning: windows apparently doesn't
		5377	accept large writes: instead of resulting in a partial write, windows will
		5378	either accept everything or return C<ENOBUFS> if the buffer is too large,
		5379	so make sure you only write small amounts into your sockets (less than a
		5380	megabyte seems safe, but this apparently depends on the amount of memory
		5381	available).
		5382
		5383	Due to the many, low, and arbitrary limits on the win32 platform and
		5384	the abysmal performance of winsockets, using a large number of sockets
		5385	is not recommended (and not reasonable). If your program needs to use
		5386	more than a hundred or so sockets, then likely it needs to use a totally
		5387	different implementation for windows, as libev offers the POSIX readiness
		5388	notification model, which cannot be implemented efficiently on windows
		5389	(due to Microsoft monopoly games).
		5390
		5391	A typical way to use libev under windows is to embed it (see the embedding
		5392	section for details) and use the following F<evwrap.h> header file instead
		5393	of F<ev.h>:
		5394
		5395	#define EV_STANDALONE /* keeps ev from requiring config.h */
		5396	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		5397
		5398	#include "ev.h"
		5399
		5400	And compile the following F<evwrap.c> file into your project (make sure
		5401	you do I<not> compile the F<ev.c> or any other embedded source files!):
		5402
		5403	#include "evwrap.h"
		5404	#include "ev.c"
		5405
		5406	=head3 The winsocket C<select> function
		5407
		5408	The winsocket C<select> function doesn't follow POSIX in that it
		5409	requires socket I<handles> and not socket I<file descriptors> (it is
		5410	also extremely buggy). This makes select very inefficient, and also
		5411	requires a mapping from file descriptors to socket handles (the Microsoft
		5412	C runtime provides the function C<_open_osfhandle> for this). See the
		5413	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		5414	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		5415
		5416	The configuration for a "naked" win32 using the Microsoft runtime
		5417	libraries and raw winsocket select is:
		5418
		5419	#define EV_USE_SELECT 1
		5420	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		5421
		5422	Note that winsockets handling of fd sets is O(n), so you can easily get a
		5423	complexity in the O(n²) range when using win32.
		5424
		5425	=head3 Limited number of file descriptors
		5426
		5427	Windows has numerous arbitrary (and low) limits on things.
		5428
		5429	Early versions of winsocket's select only supported waiting for a maximum
		5430	of C<64> handles (probably owning to the fact that all windows kernels
		5431	can only wait for C<64> things at the same time internally; Microsoft
		5432	recommends spawning a chain of threads and wait for 63 handles and the
		5433	previous thread in each. Sounds great!).
		5434
		5435	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		5436	to some high number (e.g. C<2048>) before compiling the winsocket select
		5437	call (which might be in libev or elsewhere, for example, perl and many
		5438	other interpreters do their own select emulation on windows).
		5439
		5440	Another limit is the number of file descriptors in the Microsoft runtime
		5441	libraries, which by default is C<64> (there must be a hidden I<64>
		5442	fetish or something like this inside Microsoft). You can increase this
		5443	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		5444	(another arbitrary limit), but is broken in many versions of the Microsoft
		5445	runtime libraries. This might get you to about C<512> or C<2048> sockets
		5446	(depending on windows version and/or the phase of the moon). To get more,
		5447	you need to wrap all I/O functions and provide your own fd management, but
		5448	the cost of calling select (O(n²)) will likely make this unworkable.
		5449
		5450	=head2 PORTABILITY REQUIREMENTS
		5451
		5452	In addition to a working ISO-C implementation and of course the
		5453	backend-specific APIs, libev relies on a few additional extensions:
		5454
		5455	=over 4
		5456
		5457	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		5458	calling conventions regardless of C<ev_watcher_type *>.
		5459
		5460	Libev assumes not only that all watcher pointers have the same internal
		5461	structure (guaranteed by POSIX but not by ISO C for example), but it also
		5462	assumes that the same (machine) code can be used to call any watcher
		5463	callback: The watcher callbacks have different type signatures, but libev
		5464	calls them using an C<ev_watcher *> internally.
		5465
		5466	=item null pointers and integer zero are represented by 0 bytes
		5467
		5468	Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
		5469	relies on this setting pointers and integers to null.
		5470
		5471	=item pointer accesses must be thread-atomic
		5472
		5473	Accessing a pointer value must be atomic, it must both be readable and
		5474	writable in one piece - this is the case on all current architectures.
		5475
		5476	=item C<sig_atomic_t volatile> must be thread-atomic as well
		5477
		5478	The type C<sig_atomic_t volatile> (or whatever is defined as
		5479	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		5480	threads. This is not part of the specification for C<sig_atomic_t>, but is
		5481	believed to be sufficiently portable.
		5482
		5483	=item C<sigprocmask> must work in a threaded environment
		5484
		5485	Libev uses C<sigprocmask> to temporarily block signals. This is not
		5486	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		5487	pthread implementations will either allow C<sigprocmask> in the "main
		5488	thread" or will block signals process-wide, both behaviours would
		5489	be compatible with libev. Interaction between C<sigprocmask> and
		5490	C<pthread_sigmask> could complicate things, however.
		5491
		5492	The most portable way to handle signals is to block signals in all threads
		5493	except the initial one, and run the signal handling loop in the initial
		5494	thread as well.
		5495
		5496	=item C<long> must be large enough for common memory allocation sizes
		5497
		5498	To improve portability and simplify its API, libev uses C<long> internally
		5499	instead of C<size_t> when allocating its data structures. On non-POSIX
		5500	systems (Microsoft...) this might be unexpectedly low, but is still at
		5501	least 31 bits everywhere, which is enough for hundreds of millions of
		5502	watchers.
		5503
		5504	=item C<double> must hold a time value in seconds with enough accuracy
		5505
		5506	The type C<double> is used to represent timestamps. It is required to
		5507	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		5508	good enough for at least into the year 4000 with millisecond accuracy
		5509	(the design goal for libev). This requirement is overfulfilled by
		5510	implementations using IEEE 754, which is basically all existing ones.
		5511
		5512	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5513	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5514	is either obsolete or somebody patched it to use C<long double> or
		5515	something like that, just kidding).
		5516
		5517	=back
		5518
		5519	If you know of other additional requirements drop me a note.
		5520
		5521
		5522	=head1 ALGORITHMIC COMPLEXITIES
		5523
		5524	In this section the complexities of (many of) the algorithms used inside
		5525	libev will be documented. For complexity discussions about backends see
		5526	the documentation for C<ev_default_init>.
		5527
		5528	All of the following are about amortised time: If an array needs to be
		5529	extended, libev needs to realloc and move the whole array, but this
		5530	happens asymptotically rarer with higher number of elements, so O(1) might
		5531	mean that libev does a lengthy realloc operation in rare cases, but on
		5532	average it is much faster and asymptotically approaches constant time.
		5533
		5534	=over 4
		5535
		5536	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		5537
		5538	This means that, when you have a watcher that triggers in one hour and
		5539	there are 100 watchers that would trigger before that, then inserting will
		5540	have to skip roughly seven (C<ld 100>) of these watchers.
		5541
		5542	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		5543
		5544	That means that changing a timer costs less than removing/adding them,
		5545	as only the relative motion in the event queue has to be paid for.
		5546
		5547	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		5548
		5549	These just add the watcher into an array or at the head of a list.
		5550
		5551	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		5552
		5553	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		5554
		5555	These watchers are stored in lists, so they need to be walked to find the
		5556	correct watcher to remove. The lists are usually short (you don't usually
		5557	have many watchers waiting for the same fd or signal: one is typical, two
		5558	is rare).
		5559
		5560	=item Finding the next timer in each loop iteration: O(1)
		5561
		5562	By virtue of using a binary or 4-heap, the next timer is always found at a
		5563	fixed position in the storage array.
		5564
		5565	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5566
		5567	A change means an I/O watcher gets started or stopped, which requires
		5568	libev to recalculate its status (and possibly tell the kernel, depending
		5569	on backend and whether C<ev_io_set> was used).
		5570
		5571	=item Activating one watcher (putting it into the pending state): O(1)
		5572
		5573	=item Priority handling: O(number_of_priorities)
		5574
		5575	Priorities are implemented by allocating some space for each
		5576	priority. When doing priority-based operations, libev usually has to
		5577	linearly search all the priorities, but starting/stopping and activating
		5578	watchers becomes O(1) with respect to priority handling.
		5579
		5580	=item Sending an ev_async: O(1)
		5581
		5582	=item Processing ev_async_send: O(number_of_async_watchers)
		5583
		5584	=item Processing signals: O(max_signal_number)
		5585
		5586	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5587	calls in the current loop iteration and the loop is currently
		5588	blocked. Checking for async and signal events involves iterating over all
		5589	running async watchers or all signal numbers.
		5590
		5591	=back
		5592
		5593
		5594	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5595
		5596	The major version 4 introduced some incompatible changes to the API.
		5597
		5598	At the moment, the C<ev.h> header file provides compatibility definitions
		5599	for all changes, so most programs should still compile. The compatibility
		5600	layer might be removed in later versions of libev, so better update to the
		5601	new API early than late.
		5602
		5603	=over 4
		5604
		5605	=item C<EV_COMPAT3> backwards compatibility mechanism
		5606
		5607	The backward compatibility mechanism can be controlled by
		5608	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
		5609	section.
		5610
		5611	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5612
		5613	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5614
		5615	ev_loop_destroy (EV_DEFAULT_UC);
		5616	ev_loop_fork (EV_DEFAULT);
		5617
		5618	=item function/symbol renames
		5619
		5620	A number of functions and symbols have been renamed:
		5621
		5622	ev_loop => ev_run
		5623	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5624	EVLOOP_ONESHOT => EVRUN_ONCE
		5625
		5626	ev_unloop => ev_break
		5627	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5628	EVUNLOOP_ONE => EVBREAK_ONE
		5629	EVUNLOOP_ALL => EVBREAK_ALL
		5630
		5631	EV_TIMEOUT => EV_TIMER
		5632
		5633	ev_loop_count => ev_iteration
		5634	ev_loop_depth => ev_depth
		5635	ev_loop_verify => ev_verify
		5636
		5637	Most functions working on C<struct ev_loop> objects don't have an
		5638	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5639	associated constants have been renamed to not collide with the C<struct
		5640	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5641	as all other watcher types. Note that C<ev_loop_fork> is still called
		5642	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5643	typedef.
		5644
		5645	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5646
		5647	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5648	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5649	and work, but the library code will of course be larger.
		5650
		5651	=back
		5652
		5653
		5654	=head1 GLOSSARY
		5655
		5656	=over 4
		5657
		5658	=item active
		5659
		5660	A watcher is active as long as it has been started and not yet stopped.
		5661	See L</WATCHER STATES> for details.
		5662
		5663	=item application
		5664
		5665	In this document, an application is whatever is using libev.
		5666
		5667	=item backend
		5668
		5669	The part of the code dealing with the operating system interfaces.
		5670
		5671	=item callback
		5672
		5673	The address of a function that is called when some event has been
		5674	detected. Callbacks are being passed the event loop, the watcher that
		5675	received the event, and the actual event bitset.
		5676
		5677	=item callback/watcher invocation
		5678
		5679	The act of calling the callback associated with a watcher.
		5680
		5681	=item event
		5682
		5683	A change of state of some external event, such as data now being available
		5684	for reading on a file descriptor, time having passed or simply not having
		5685	any other events happening anymore.
		5686
		5687	In libev, events are represented as single bits (such as C<EV_READ> or
		5688	C<EV_TIMER>).
		5689
		5690	=item event library
		5691
		5692	A software package implementing an event model and loop.
		5693
		5694	=item event loop
		5695
		5696	An entity that handles and processes external events and converts them
		5697	into callback invocations.
		5698
		5699	=item event model
		5700
		5701	The model used to describe how an event loop handles and processes
		5702	watchers and events.
		5703
		5704	=item pending
		5705
		5706	A watcher is pending as soon as the corresponding event has been
		5707	detected. See L</WATCHER STATES> for details.
		5708
		5709	=item real time
		5710
		5711	The physical time that is observed. It is apparently strictly monotonic :)
		5712
		5713	=item wall-clock time
		5714
		5715	The time and date as shown on clocks. Unlike real time, it can actually
		5716	be wrong and jump forwards and backwards, e.g. when you adjust your
		5717	clock.
		5718
		5719	=item watcher
		5720
		5721	A data structure that describes interest in certain events. Watchers need
		5722	to be started (attached to an event loop) before they can receive events.
		5723
		5724	=back
1696		5725
1697	=head1 AUTHOR	5726	=head1 AUTHOR
1698		5727
1699	Marc Lehmann <libev@schmorp.de>.	5728	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5729	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
1700		5730

Diff Legend

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