ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.35 by root, Fri Nov 23 19:35:09 2007 UTC vs.
Revision 1.369 by root, Mon May 30 18:34:28 2011 UTC

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

Diff Legend

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