ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.90 by root, Thu Dec 20 01:18:37 2007 UTC vs.
Revision 1.336 by root, Mon Oct 25 11:10:10 2010 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 EXAMPLE PROGRAM	9	=head2 EXAMPLE PROGRAM
10		10
		11	// a single header file is required
11	#include <ev.h>	12	#include <ev.h>
12		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
13	ev_io stdin_watcher;	18	ev_io stdin_watcher;
14	ev_timer timeout_watcher;	19	ev_timer timeout_watcher;
15		20
16	/* called when data readable on stdin */	21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
17	static void	23	static void
18	stdin_cb (EV_P_ struct ev_io *w, int revents)	24	stdin_cb (EV_P_ ev_io *w, int revents)
19	{	25	{
20	/* puts ("stdin ready"); */	26	puts ("stdin ready");
21	ev_io_stop (EV_A_ w); /* just a syntax example */	27	// for one-shot events, one must manually stop the watcher
22	ev_unloop (EV_A_ EVUNLOOP_ALL); /* leave all loop calls */	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);
23	}	33	}
24		34
		35	// another callback, this time for a time-out
25	static void	36	static void
26	timeout_cb (EV_P_ struct ev_timer *w, int revents)	37	timeout_cb (EV_P_ ev_timer *w, int revents)
27	{	38	{
28	/* puts ("timeout"); */	39	puts ("timeout");
29	ev_unloop (EV_A_ EVUNLOOP_ONE); /* leave one loop call */	40	// this causes the innermost ev_run to stop iterating
		41	ev_break (EV_A_ EVBREAK_ONE);
30	}	42	}
31		43
32	int	44	int
33	main (void)	45	main (void)
34	{	46	{
35	struct ev_loop *loop = ev_default_loop (0);	47	// use the default event loop unless you have special needs
		48	struct ev_loop *loop = EV_DEFAULT;
36		49
37	/* initialise an io watcher, then start it */	50	// initialise an io watcher, then start it
		51	// this one will watch for stdin to become readable
38	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);	52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
39	ev_io_start (loop, &stdin_watcher);	53	ev_io_start (loop, &stdin_watcher);
40		54
		55	// initialise a timer watcher, then start it
41	/* simple non-repeating 5.5 second timeout */	56	// simple non-repeating 5.5 second timeout
42	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);	57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
43	ev_timer_start (loop, &timeout_watcher);	58	ev_timer_start (loop, &timeout_watcher);
44		59
45	/* loop till timeout or data ready */	60	// now wait for events to arrive
46	ev_loop (loop, 0);	61	ev_run (loop, 0);
47		62
		63	// unloop was called, so exit
48	return 0;	64	return 0;
49	}	65	}
50		66
51	=head1 DESCRIPTION	67	=head1 ABOUT THIS DOCUMENT
52		68
		69	This document documents the libev software package.
		70
53	The newest version of this document is also available as a html-formatted	71	The newest version of this document is also available as an html-formatted
54	web page you might find easier to navigate when reading it for the first	72	web page you might find easier to navigate when reading it for the first
55	time: L<http://cvs.schmorp.de/libev/ev.html>.	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
56		92
57	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
58	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
59	these event sources and provide your program with events.	95	these event sources and provide your program with events.
60		96
61	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
62	(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
63	communicate events via a callback mechanism.	99	communicate events via a callback mechanism.
…		…
65	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
66	watchers>, which are relatively small C structures you initialise with the	102	watchers>, which are relatively small C structures you initialise with the
67	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
68	watcher.	104	watcher.
69		105
70	=head1 FEATURES	106	=head2 FEATURES
71		107
72	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the	108	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
73	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms	109	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
74	for file descriptor events (C<ev_io>), the Linux C<inotify> interface	110	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
75	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers	111	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
76	with customised rescheduling (C<ev_periodic>), synchronous signals	112	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
77	(C<ev_signal>), process status change events (C<ev_child>), and event	113	timers (C<ev_timer>), absolute timers with customised rescheduling
78	watchers dealing with the event loop mechanism itself (C<ev_idle>,	114	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
79	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as	115	change events (C<ev_child>), and event watchers dealing with the event
80	file watchers (C<ev_stat>) and even limited support for fork events	116	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
81	(C<ev_fork>).	117	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		118	limited support for fork events (C<ev_fork>).
82		119
83	It also is quite fast (see this	120	It also is quite fast (see this
84	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent	121	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
85	for example).	122	for example).
86		123
87	=head1 CONVENTIONS	124	=head2 CONVENTIONS
88		125
89	Libev is very configurable. In this manual the default configuration will	126	Libev is very configurable. In this manual the default (and most common)
90	be described, which supports multiple event loops. For more info about	127	configuration will be described, which supports multiple event loops. For
91	various configuration options please have a look at B<EMBED> section in	128	more info about various configuration options please have a look at
92	this manual. If libev was configured without support for multiple event	129	B<EMBED> section in this manual. If libev was configured without support
93	loops, then all functions taking an initial argument of name C<loop>	130	for multiple event loops, then all functions taking an initial argument of
94	(which is always of type C<struct ev_loop *>) will not have this argument.	131	name C<loop> (which is always of type C<struct ev_loop *>) will not have
		132	this argument.
95		133
96	=head1 TIME REPRESENTATION	134	=head2 TIME REPRESENTATION
97		135
98	Libev represents time as a single floating point number, representing the	136	Libev represents time as a single floating point number, representing
99	(fractional) number of seconds since the (POSIX) epoch (somewhere near	137	the (fractional) number of seconds since the (POSIX) epoch (in practice
100	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
101	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
102	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
103	it, you should treat it as some floatingpoint value. Unlike the name	141	any calculations on it, you should treat it as some floating point value.
		142
104	component C<stamp> might indicate, it is also used for time differences	143	Unlike the name component C<stamp> might indicate, it is also used for
105	throughout libev.	144	time differences (e.g. delays) throughout libev.
		145
		146	=head1 ERROR HANDLING
		147
		148	Libev knows three classes of errors: operating system errors, usage errors
		149	and internal errors (bugs).
		150
		151	When libev catches an operating system error it cannot handle (for example
		152	a system call indicating a condition libev cannot fix), it calls the callback
		153	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		154	abort. The default is to print a diagnostic message and to call C<abort
		155	()>.
		156
		157	When libev detects a usage error such as a negative timer interval, then
		158	it will print a diagnostic message and abort (via the C<assert> mechanism,
		159	so C<NDEBUG> will disable this checking): these are programming errors in
		160	the libev caller and need to be fixed there.
		161
		162	Libev also has a few internal error-checking C<assert>ions, and also has
		163	extensive consistency checking code. These do not trigger under normal
		164	circumstances, as they indicate either a bug in libev or worse.
		165
106		166
107	=head1 GLOBAL FUNCTIONS	167	=head1 GLOBAL FUNCTIONS
108		168
109	These functions can be called anytime, even before initialising the	169	These functions can be called anytime, even before initialising the
110	library in any way.	170	library in any way.
…		…
113		173
114	=item ev_tstamp ev_time ()	174	=item ev_tstamp ev_time ()
115		175
116	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
117	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
118	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 ()>.
119		186
120	=item int ev_version_major ()	187	=item int ev_version_major ()
121		188
122	=item int ev_version_minor ()	189	=item int ev_version_minor ()
123		190
…		…
134	as this indicates an incompatible change. Minor versions are usually	201	as this indicates an incompatible change. Minor versions are usually
135	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
136	not a problem.	203	not a problem.
137		204
138	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
139	version.	206	version (note, however, that this will not detect other ABI mismatches,
		207	such as LFS or reentrancy).
140		208
141	assert (("libev version mismatch",	209	assert (("libev version mismatch",
142	ev_version_major () == EV_VERSION_MAJOR	210	ev_version_major () == EV_VERSION_MAJOR
143	&& ev_version_minor () >= EV_VERSION_MINOR));	211	&& ev_version_minor () >= EV_VERSION_MINOR));
144		212
145	=item unsigned int ev_supported_backends ()	213	=item unsigned int ev_supported_backends ()
146		214
147	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_*>
148	value) compiled into this binary of libev (independent of their	216	value) compiled into this binary of libev (independent of their
…		…
150	a description of the set values.	218	a description of the set values.
151		219
152	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
153	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
154		222
155	assert (("sorry, no epoll, no sex",	223	assert (("sorry, no epoll, no sex",
156	ev_supported_backends () & EVBACKEND_EPOLL));	224	ev_supported_backends () & EVBACKEND_EPOLL));
157		225
158	=item unsigned int ev_recommended_backends ()	226	=item unsigned int ev_recommended_backends ()
159		227
160	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
161	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
162	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
163	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
164	(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
165	libev will probe for if you specify no backends explicitly.	234	probe for if you specify no backends explicitly.
166		235
167	=item unsigned int ev_embeddable_backends ()	236	=item unsigned int ev_embeddable_backends ()
168		237
169	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
170	is the theoretical, all-platform, value. To find which backends	239	value is platform-specific but can include backends not available on the
171	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
172	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for	241	the current system, you would need to look at C<ev_embeddable_backends ()
173	recommended ones.	242	& ev_supported_backends ()>, likewise for recommended ones.
174		243
175	See the description of C<ev_embed> watchers for more info.	244	See the description of C<ev_embed> watchers for more info.
176		245
177	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	246	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
178		247
179	Sets the allocation function to use (the prototype is similar - the	248	Sets the allocation function to use (the prototype is similar - the
180	semantics is identical - to the realloc C function). It is used to	249	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
181	allocate and free memory (no surprises here). If it returns zero when	250	used to allocate and free memory (no surprises here). If it returns zero
182	memory needs to be allocated, the library might abort or take some	251	when memory needs to be allocated (C<size != 0>), the library might abort
183	potentially destructive action. The default is your system realloc	252	or take some potentially destructive action.
184	function.	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.
185		257
186	You could override this function in high-availability programs to, say,	258	You could override this function in high-availability programs to, say,
187	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,
188	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.
189		261
190	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
191	retries).	263	retries (example requires a standards-compliant C<realloc>).
192		264
193	static void *	265	static void *
194	persistent_realloc (void *ptr, size_t size)	266	persistent_realloc (void *ptr, size_t size)
195	{	267	{
196	for (;;)	268	for (;;)
…		…
205	}	277	}
206		278
207	...	279	...
208	ev_set_allocator (persistent_realloc);	280	ev_set_allocator (persistent_realloc);
209		281
210	=item ev_set_syserr_cb (void (*cb)(const char *msg));	282	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
211		283
212	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
213	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
214	indicating the system call or subsystem causing the problem. If this	286	indicating the system call or subsystem causing the problem. If this
215	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
216	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
217	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
218	(such as abort).	290	(such as abort).
219		291
220	Example: This is basically the same thing that libev does internally, too.	292	Example: This is basically the same thing that libev does internally, too.
…		…
229	...	301	...
230	ev_set_syserr_cb (fatal_error);	302	ev_set_syserr_cb (fatal_error);
231		303
232	=back	304	=back
233		305
234	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	306	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
235		307
236	An event loop is described by a C<struct ev_loop *>. The library knows two	308	An event loop is described by a C<struct ev_loop *> (the C<struct> is
237	types of such loops, the I<default> loop, which supports signals and child	309	I<not> optional in this case unless libev 3 compatibility is disabled, as
238	events, and dynamically created loops which do not.	310	libev 3 had an C<ev_loop> function colliding with the struct name).
239		311
240	If you use threads, a common model is to run the default event loop	312	The library knows two types of such loops, the I<default> loop, which
241	in your main thread (or in a separate thread) and for each thread you	313	supports child process events, and dynamically created event loops which
242	create, you also create another event loop. Libev itself does no locking	314	do not.
243	whatsoever, so if you mix calls to the same event loop in different
244	threads, make sure you lock (this is usually a bad idea, though, even if
245	done correctly, because it's hideous and inefficient).
246		315
247	=over 4	316	=over 4
248		317
249	=item struct ev_loop *ev_default_loop (unsigned int flags)	318	=item struct ev_loop *ev_default_loop (unsigned int flags)
250		319
251	This will initialise the default event loop if it hasn't been initialised	320	This returns the "default" event loop object, which is what you should
252	yet and return it. If the default loop could not be initialised, returns	321	normally use when you just need "the event loop". Event loop objects and
253	false. If it already was initialised it simply returns it (and ignores the	322	the C<flags> parameter are described in more detail in the entry for
254	flags. If that is troubling you, check C<ev_backend ()> afterwards).	323	C<ev_loop_new>.
		324
		325	If the default loop is already initialised then this function simply
		326	returns it (and ignores the flags. If that is troubling you, check
		327	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		328	flags, which should almost always be C<0>, unless the caller is also the
		329	one calling C<ev_run> or otherwise qualifies as "the main program".
255		330
256	If you don't know what event loop to use, use the one returned from this	331	If you don't know what event loop to use, use the one returned from this
257	function.	332	function (or via the C<EV_DEFAULT> macro).
		333
		334	Note that this function is I<not> thread-safe, so if you want to use it
		335	from multiple threads, you have to employ some kind of mutex (note also
		336	that this case is unlikely, as loops cannot be shared easily between
		337	threads anyway).
		338
		339	The default loop is the only loop that can handle C<ev_child> watchers,
		340	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		341	a problem for your application you can either create a dynamic loop with
		342	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		343	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		344
		345	Example: This is the most typical usage.
		346
		347	if (!ev_default_loop (0))
		348	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		349
		350	Example: Restrict libev to the select and poll backends, and do not allow
		351	environment settings to be taken into account:
		352
		353	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		354
		355	=item struct ev_loop *ev_loop_new (unsigned int flags)
		356
		357	This will create and initialise a new event loop object. If the loop
		358	could not be initialised, returns false.
		359
		360	Note that this function I<is> thread-safe, and one common way to use
		361	libev with threads is indeed to create one loop per thread, and using the
		362	default loop in the "main" or "initial" thread.
258		363
259	The flags argument can be used to specify special behaviour or specific	364	The flags argument can be used to specify special behaviour or specific
260	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).	365	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
261		366
262	The following flags are supported:	367	The following flags are supported:
…		…
268	The default flags value. Use this if you have no clue (it's the right	373	The default flags value. Use this if you have no clue (it's the right
269	thing, believe me).	374	thing, believe me).
270		375
271	=item C<EVFLAG_NOENV>	376	=item C<EVFLAG_NOENV>
272		377
273	If this flag bit is ored into the flag value (or the program runs setuid	378	If this flag bit is or'ed into the flag value (or the program runs setuid
274	or setgid) then libev will I<not> look at the environment variable	379	or setgid) then libev will I<not> look at the environment variable
275	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	380	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
276	override the flags completely if it is found in the environment. This is	381	override the flags completely if it is found in the environment. This is
277	useful to try out specific backends to test their performance, or to work	382	useful to try out specific backends to test their performance, or to work
278	around bugs.	383	around bugs.
279		384
280	=item C<EVFLAG_FORKCHECK>	385	=item C<EVFLAG_FORKCHECK>
281		386
282	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after	387	Instead of calling C<ev_loop_fork> manually after a fork, you can also
283	a fork, you can also make libev check for a fork in each iteration by	388	make libev check for a fork in each iteration by enabling this flag.
284	enabling this flag.
285		389
286	This works by calling C<getpid ()> on every iteration of the loop,	390	This works by calling C<getpid ()> on every iteration of the loop,
287	and thus this might slow down your event loop if you do a lot of loop	391	and thus this might slow down your event loop if you do a lot of loop
288	iterations and little real work, but is usually not noticeable (on my	392	iterations and little real work, but is usually not noticeable (on my
289	Linux system for example, C<getpid> is actually a simple 5-insn sequence	393	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
290	without a syscall and thus I<very> fast, but my Linux system also has	394	without a system call and thus I<very> fast, but my GNU/Linux system also has
291	C<pthread_atfork> which is even faster).	395	C<pthread_atfork> which is even faster).
292		396
293	The big advantage of this flag is that you can forget about fork (and	397	The big advantage of this flag is that you can forget about fork (and
294	forget about forgetting to tell libev about forking) when you use this	398	forget about forgetting to tell libev about forking) when you use this
295	flag.	399	flag.
296		400
297	This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS>	401	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
298	environment variable.	402	environment variable.
		403
		404	=item C<EVFLAG_NOINOTIFY>
		405
		406	When this flag is specified, then libev will not attempt to use the
		407	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		408	testing, this flag can be useful to conserve inotify file descriptors, as
		409	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		410
		411	=item C<EVFLAG_SIGNALFD>
		412
		413	When this flag is specified, then libev will attempt to use the
		414	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API
		415	delivers signals synchronously, which makes it both faster and might make
		416	it possible to get the queued signal data. It can also simplify signal
		417	handling with threads, as long as you properly block signals in your
		418	threads that are not interested in handling them.
		419
		420	Signalfd will not be used by default as this changes your signal mask, and
		421	there are a lot of shoddy libraries and programs (glib's threadpool for
		422	example) that can't properly initialise their signal masks.
299		423
300	=item C<EVBACKEND_SELECT> (value 1, portable select backend)	424	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
301		425
302	This is your standard select(2) backend. Not I<completely> standard, as	426	This is your standard select(2) backend. Not I<completely> standard, as
303	libev tries to roll its own fd_set with no limits on the number of fds,	427	libev tries to roll its own fd_set with no limits on the number of fds,
304	but if that fails, expect a fairly low limit on the number of fds when	428	but if that fails, expect a fairly low limit on the number of fds when
305	using this backend. It doesn't scale too well (O(highest_fd)), but its usually	429	using this backend. It doesn't scale too well (O(highest_fd)), but its
306	the fastest backend for a low number of fds.	430	usually the fastest backend for a low number of (low-numbered :) fds.
		431
		432	To get good performance out of this backend you need a high amount of
		433	parallelism (most of the file descriptors should be busy). If you are
		434	writing a server, you should C<accept ()> in a loop to accept as many
		435	connections as possible during one iteration. You might also want to have
		436	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		437	readiness notifications you get per iteration.
		438
		439	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		440	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		441	C<exceptfds> set on that platform).
307		442
308	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)	443	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
309		444
310	And this is your standard poll(2) backend. It's more complicated than	445	And this is your standard poll(2) backend. It's more complicated
311	select, but handles sparse fds better and has no artificial limit on the	446	than select, but handles sparse fds better and has no artificial
312	number of fds you can use (except it will slow down considerably with a	447	limit on the number of fds you can use (except it will slow down
313	lot of inactive fds). It scales similarly to select, i.e. O(total_fds).	448	considerably with a lot of inactive fds). It scales similarly to select,
		449	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		450	performance tips.
		451
		452	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		453	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
314		454
315	=item C<EVBACKEND_EPOLL> (value 4, Linux)	455	=item C<EVBACKEND_EPOLL> (value 4, Linux)
316		456
		457	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		458	kernels).
		459
317	For few fds, this backend is a bit little slower than poll and select,	460	For few fds, this backend is a bit little slower than poll and select,
318	but it scales phenomenally better. While poll and select usually scale like	461	but it scales phenomenally better. While poll and select usually scale
319	O(total_fds) where n is the total number of fds (or the highest fd), epoll scales	462	like O(total_fds) where n is the total number of fds (or the highest fd),
320	either O(1) or O(active_fds).	463	epoll scales either O(1) or O(active_fds).
321		464
		465	The epoll mechanism deserves honorable mention as the most misdesigned
		466	of the more advanced event mechanisms: mere annoyances include silently
		467	dropping file descriptors, requiring a system call per change per file
		468	descriptor (and unnecessary guessing of parameters), problems with dup and
		469	so on. The biggest issue is fork races, however - if a program forks then
		470	I<both> parent and child process have to recreate the epoll set, which can
		471	take considerable time (one syscall per file descriptor) and is of course
		472	hard to detect.
		473
		474	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		475	of course I<doesn't>, and epoll just loves to report events for totally
		476	I<different> file descriptors (even already closed ones, so one cannot
		477	even remove them from the set) than registered in the set (especially
		478	on SMP systems). Libev tries to counter these spurious notifications by
		479	employing an additional generation counter and comparing that against the
		480	events to filter out spurious ones, recreating the set when required. Last
		481	not least, it also refuses to work with some file descriptors which work
		482	perfectly fine with C<select> (files, many character devices...).
		483
322	While stopping and starting an I/O watcher in the same iteration will	484	While stopping, setting and starting an I/O watcher in the same iteration
323	result in some caching, there is still a syscall per such incident	485	will result in some caching, there is still a system call per such
324	(because the fd could point to a different file description now), so its	486	incident (because the same I<file descriptor> could point to a different
325	best to avoid that. Also, dup()ed file descriptors might not work very	487	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
326	well if you register events for both fds.	488	file descriptors might not work very well if you register events for both
		489	file descriptors.
327		490
328	Please note that epoll sometimes generates spurious notifications, so you	491	Best performance from this backend is achieved by not unregistering all
329	need to use non-blocking I/O or other means to avoid blocking when no data	492	watchers for a file descriptor until it has been closed, if possible,
330	(or space) is available.	493	i.e. keep at least one watcher active per fd at all times. Stopping and
		494	starting a watcher (without re-setting it) also usually doesn't cause
		495	extra overhead. A fork can both result in spurious notifications as well
		496	as in libev having to destroy and recreate the epoll object, which can
		497	take considerable time and thus should be avoided.
		498
		499	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		500	faster than epoll for maybe up to a hundred file descriptors, depending on
		501	the usage. So sad.
		502
		503	While nominally embeddable in other event loops, this feature is broken in
		504	all kernel versions tested so far.
		505
		506	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		507	C<EVBACKEND_POLL>.
331		508
332	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	509	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
333		510
334	Kqueue deserves special mention, as at the time of this writing, it	511	Kqueue deserves special mention, as at the time of this writing, it
335	was broken on all BSDs except NetBSD (usually it doesn't work with	512	was broken on all BSDs except NetBSD (usually it doesn't work reliably
336	anything but sockets and pipes, except on Darwin, where of course it's	513	with anything but sockets and pipes, except on Darwin, where of course
337	completely useless). For this reason it's not being "autodetected"	514	it's completely useless). Unlike epoll, however, whose brokenness
		515	is by design, these kqueue bugs can (and eventually will) be fixed
		516	without API changes to existing programs. For this reason it's not being
338	unless you explicitly specify it explicitly in the flags (i.e. using	517	"auto-detected" unless you explicitly specify it in the flags (i.e. using
339	C<EVBACKEND_KQUEUE>).	518	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		519	system like NetBSD.
		520
		521	You still can embed kqueue into a normal poll or select backend and use it
		522	only for sockets (after having made sure that sockets work with kqueue on
		523	the target platform). See C<ev_embed> watchers for more info.
340		524
341	It scales in the same way as the epoll backend, but the interface to the	525	It scales in the same way as the epoll backend, but the interface to the
342	kernel is more efficient (which says nothing about its actual speed, of	526	kernel is more efficient (which says nothing about its actual speed, of
343	course). While starting and stopping an I/O watcher does not cause an	527	course). While stopping, setting and starting an I/O watcher does never
344	extra syscall as with epoll, it still adds up to four event changes per	528	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
345	incident, so its best to avoid that.	529	two event changes per incident. Support for C<fork ()> is very bad (but
		530	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		531	cases
		532
		533	This backend usually performs well under most conditions.
		534
		535	While nominally embeddable in other event loops, this doesn't work
		536	everywhere, so you might need to test for this. And since it is broken
		537	almost everywhere, you should only use it when you have a lot of sockets
		538	(for which it usually works), by embedding it into another event loop
		539	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		540	also broken on OS X)) and, did I mention it, using it only for sockets.
		541
		542	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		543	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		544	C<NOTE_EOF>.
346		545
347	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)	546	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
348		547
349	This is not implemented yet (and might never be).	548	This is not implemented yet (and might never be, unless you send me an
		549	implementation). According to reports, C</dev/poll> only supports sockets
		550	and is not embeddable, which would limit the usefulness of this backend
		551	immensely.
350		552
351	=item C<EVBACKEND_PORT> (value 32, Solaris 10)	553	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
352		554
353	This uses the Solaris 10 port mechanism. As with everything on Solaris,	555	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
354	it's really slow, but it still scales very well (O(active_fds)).	556	it's really slow, but it still scales very well (O(active_fds)).
355		557
356	Please note that solaris ports can result in a lot of spurious	558	Please note that Solaris event ports can deliver a lot of spurious
357	notifications, so you need to use non-blocking I/O or other means to avoid	559	notifications, so you need to use non-blocking I/O or other means to avoid
358	blocking when no data (or space) is available.	560	blocking when no data (or space) is available.
		561
		562	While this backend scales well, it requires one system call per active
		563	file descriptor per loop iteration. For small and medium numbers of file
		564	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		565	might perform better.
		566
		567	On the positive side, with the exception of the spurious readiness
		568	notifications, this backend actually performed fully to specification
		569	in all tests and is fully embeddable, which is a rare feat among the
		570	OS-specific backends (I vastly prefer correctness over speed hacks).
		571
		572	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		573	C<EVBACKEND_POLL>.
359		574
360	=item C<EVBACKEND_ALL>	575	=item C<EVBACKEND_ALL>
361		576
362	Try all backends (even potentially broken ones that wouldn't be tried	577	Try all backends (even potentially broken ones that wouldn't be tried
363	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	578	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
364	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.	579	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
365		580
		581	It is definitely not recommended to use this flag.
		582
366	=back	583	=back
367		584
368	If one or more of these are ored into the flags value, then only these	585	If one or more of the backend flags are or'ed into the flags value,
369	backends will be tried (in the reverse order as given here). If none are	586	then only these backends will be tried (in the reverse order as listed
370	specified, most compiled-in backend will be tried, usually in reverse	587	here). If none are specified, all backends in C<ev_recommended_backends
371	order of their flag values :)	588	()> will be tried.
372
373	The most typical usage is like this:
374
375	if (!ev_default_loop (0))
376	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
377
378	Restrict libev to the select and poll backends, and do not allow
379	environment settings to be taken into account:
380
381	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
382
383	Use whatever libev has to offer, but make sure that kqueue is used if
384	available (warning, breaks stuff, best use only with your own private
385	event loop and only if you know the OS supports your types of fds):
386
387	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
388
389	=item struct ev_loop *ev_loop_new (unsigned int flags)
390
391	Similar to C<ev_default_loop>, but always creates a new event loop that is
392	always distinct from the default loop. Unlike the default loop, it cannot
393	handle signal and child watchers, and attempts to do so will be greeted by
394	undefined behaviour (or a failed assertion if assertions are enabled).
395		589
396	Example: Try to create a event loop that uses epoll and nothing else.	590	Example: Try to create a event loop that uses epoll and nothing else.
397		591
398	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);	592	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
399	if (!epoller)	593	if (!epoller)
400	fatal ("no epoll found here, maybe it hides under your chair");	594	fatal ("no epoll found here, maybe it hides under your chair");
401		595
		596	Example: Use whatever libev has to offer, but make sure that kqueue is
		597	used if available.
		598
		599	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		600
402	=item ev_default_destroy ()	601	=item ev_loop_destroy (loop)
403		602
404	Destroys the default loop again (frees all memory and kernel state	603	Destroys an event loop object (frees all memory and kernel state
405	etc.). None of the active event watchers will be stopped in the normal	604	etc.). None of the active event watchers will be stopped in the normal
406	sense, so e.g. C<ev_is_active> might still return true. It is your	605	sense, so e.g. C<ev_is_active> might still return true. It is your
407	responsibility to either stop all watchers cleanly yoursef I<before>	606	responsibility to either stop all watchers cleanly yourself I<before>
408	calling this function, or cope with the fact afterwards (which is usually	607	calling this function, or cope with the fact afterwards (which is usually
409	the easiest thing, you can just ignore the watchers and/or C<free ()> them	608	the easiest thing, you can just ignore the watchers and/or C<free ()> them
410	for example).	609	for example).
411		610
412	Note that certain global state, such as signal state, will not be freed by	611	Note that certain global state, such as signal state (and installed signal
413	this function, and related watchers (such as signal and child watchers)	612	handlers), will not be freed by this function, and related watchers (such
414	would need to be stopped manually.	613	as signal and child watchers) would need to be stopped manually.
415		614
416	In general it is not advisable to call this function except in the	615	This function is normally used on loop objects allocated by
417	rare occasion where you really need to free e.g. the signal handling	616	C<ev_loop_new>, but it can also be used on the default loop returned by
		617	C<ev_default_loop>, in which case it is not thread-safe.
		618
		619	Note that it is not advisable to call this function on the default loop
		620	except in the rare occasion where you really need to free it's resources.
418	pipe fds. If you need dynamically allocated loops it is better to use	621	If you need dynamically allocated loops it is better to use C<ev_loop_new>
419	C<ev_loop_new> and C<ev_loop_destroy>).	622	and C<ev_loop_destroy>.
420		623
421	=item ev_loop_destroy (loop)	624	=item ev_loop_fork (loop)
422		625
423	Like C<ev_default_destroy>, but destroys an event loop created by an	626	This function sets a flag that causes subsequent C<ev_run> iterations to
424	earlier call to C<ev_loop_new>.
425
426	=item ev_default_fork ()
427
428	This function reinitialises the kernel state for backends that have	627	reinitialise the kernel state for backends that have one. Despite the
429	one. Despite the name, you can call it anytime, but it makes most sense	628	name, you can call it anytime, but it makes most sense after forking, in
430	after forking, in either the parent or child process (or both, but that	629	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
431	again makes little sense).	630	child before resuming or calling C<ev_run>.
432		631
433	You I<must> call this function in the child process after forking if and	632	Again, you I<have> to call it on I<any> loop that you want to re-use after
434	only if you want to use the event library in both processes. If you just	633	a fork, I<even if you do not plan to use the loop in the parent>. This is
435	fork+exec, you don't have to call it.	634	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		635	during fork.
		636
		637	On the other hand, you only need to call this function in the child
		638	process if and only if you want to use the event loop in the child. If
		639	you just fork+exec or create a new loop in the child, you don't have to
		640	call it at all (in fact, C<epoll> is so badly broken that it makes a
		641	difference, but libev will usually detect this case on its own and do a
		642	costly reset of the backend).
436		643
437	The function itself is quite fast and it's usually not a problem to call	644	The function itself is quite fast and it's usually not a problem to call
438	it just in case after a fork. To make this easy, the function will fit in	645	it just in case after a fork.
439	quite nicely into a call to C<pthread_atfork>:
440		646
		647	Example: Automate calling C<ev_loop_fork> on the default loop when
		648	using pthreads.
		649
		650	static void
		651	post_fork_child (void)
		652	{
		653	ev_loop_fork (EV_DEFAULT);
		654	}
		655
		656	...
441	pthread_atfork (0, 0, ev_default_fork);	657	pthread_atfork (0, 0, post_fork_child);
442		658
443	At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use	659	=item int ev_is_default_loop (loop)
444	without calling this function, so if you force one of those backends you
445	do not need to care.
446		660
447	=item ev_loop_fork (loop)	661	Returns true when the given loop is, in fact, the default loop, and false
		662	otherwise.
448		663
449	Like C<ev_default_fork>, but acts on an event loop created by
450	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
451	after fork, and how you do this is entirely your own problem.
452
453	=item unsigned int ev_loop_count (loop)	664	=item unsigned int ev_iteration (loop)
454		665
455	Returns the count of loop iterations for the loop, which is identical to	666	Returns the current iteration count for the event loop, which is identical
456	the number of times libev did poll for new events. It starts at C<0> and	667	to the number of times libev did poll for new events. It starts at C<0>
457	happily wraps around with enough iterations.	668	and happily wraps around with enough iterations.
458		669
459	This value can sometimes be useful as a generation counter of sorts (it	670	This value can sometimes be useful as a generation counter of sorts (it
460	"ticks" the number of loop iterations), as it roughly corresponds with	671	"ticks" the number of loop iterations), as it roughly corresponds with
461	C<ev_prepare> and C<ev_check> calls.	672	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		673	prepare and check phases.
		674
		675	=item unsigned int ev_depth (loop)
		676
		677	Returns the number of times C<ev_run> was entered minus the number of
		678	times C<ev_run> was exited, in other words, the recursion depth.
		679
		680	Outside C<ev_run>, this number is zero. In a callback, this number is
		681	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		682	in which case it is higher.
		683
		684	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread
		685	etc.), doesn't count as "exit" - consider this as a hint to avoid such
		686	ungentleman-like behaviour unless it's really convenient.
462		687
463	=item unsigned int ev_backend (loop)	688	=item unsigned int ev_backend (loop)
464		689
465	Returns one of the C<EVBACKEND_*> flags indicating the event backend in	690	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
466	use.	691	use.
…		…
469		694
470	Returns the current "event loop time", which is the time the event loop	695	Returns the current "event loop time", which is the time the event loop
471	received events and started processing them. This timestamp does not	696	received events and started processing them. This timestamp does not
472	change as long as callbacks are being processed, and this is also the base	697	change as long as callbacks are being processed, and this is also the base
473	time used for relative timers. You can treat it as the timestamp of the	698	time used for relative timers. You can treat it as the timestamp of the
474	event occuring (or more correctly, libev finding out about it).	699	event occurring (or more correctly, libev finding out about it).
475		700
		701	=item ev_now_update (loop)
		702
		703	Establishes the current time by querying the kernel, updating the time
		704	returned by C<ev_now ()> in the progress. This is a costly operation and
		705	is usually done automatically within C<ev_run ()>.
		706
		707	This function is rarely useful, but when some event callback runs for a
		708	very long time without entering the event loop, updating libev's idea of
		709	the current time is a good idea.
		710
		711	See also L<The special problem of time updates> in the C<ev_timer> section.
		712
		713	=item ev_suspend (loop)
		714
		715	=item ev_resume (loop)
		716
		717	These two functions suspend and resume an event loop, for use when the
		718	loop is not used for a while and timeouts should not be processed.
		719
		720	A typical use case would be an interactive program such as a game: When
		721	the user presses C<^Z> to suspend the game and resumes it an hour later it
		722	would be best to handle timeouts as if no time had actually passed while
		723	the program was suspended. This can be achieved by calling C<ev_suspend>
		724	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		725	C<ev_resume> directly afterwards to resume timer processing.
		726
		727	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		728	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		729	will be rescheduled (that is, they will lose any events that would have
		730	occurred while suspended).
		731
		732	After calling C<ev_suspend> you B<must not> call I<any> function on the
		733	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		734	without a previous call to C<ev_suspend>.
		735
		736	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		737	event loop time (see C<ev_now_update>).
		738
476	=item ev_loop (loop, int flags)	739	=item ev_run (loop, int flags)
477		740
478	Finally, this is it, the event handler. This function usually is called	741	Finally, this is it, the event handler. This function usually is called
479	after you initialised all your watchers and you want to start handling	742	after you have initialised all your watchers and you want to start
480	events.	743	handling events. It will ask the operating system for any new events, call
		744	the watcher callbacks, an then repeat the whole process indefinitely: This
		745	is why event loops are called I<loops>.
481		746
482	If the flags argument is specified as C<0>, it will not return until	747	If the flags argument is specified as C<0>, it will keep handling events
483	either no event watchers are active anymore or C<ev_unloop> was called.	748	until either no event watchers are active anymore or C<ev_break> was
		749	called.
484		750
485	Please note that an explicit C<ev_unloop> is usually better than	751	Please note that an explicit C<ev_break> is usually better than
486	relying on all watchers to be stopped when deciding when a program has	752	relying on all watchers to be stopped when deciding when a program has
487	finished (especially in interactive programs), but having a program that	753	finished (especially in interactive programs), but having a program
488	automatically loops as long as it has to and no longer by virtue of	754	that automatically loops as long as it has to and no longer by virtue
489	relying on its watchers stopping correctly is a thing of beauty.	755	of relying on its watchers stopping correctly, that is truly a thing of
		756	beauty.
490		757
491	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	758	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
492	those events and any outstanding ones, but will not block your process in	759	those events and any already outstanding ones, but will not wait and
493	case there are no events and will return after one iteration of the loop.	760	block your process in case there are no events and will return after one
		761	iteration of the loop. This is sometimes useful to poll and handle new
		762	events while doing lengthy calculations, to keep the program responsive.
494		763
495	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	764	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
496	neccessary) and will handle those and any outstanding ones. It will block	765	necessary) and will handle those and any already outstanding ones. It
497	your process until at least one new event arrives, and will return after	766	will block your process until at least one new event arrives (which could
498	one iteration of the loop. This is useful if you are waiting for some	767	be an event internal to libev itself, so there is no guarantee that a
499	external event in conjunction with something not expressible using other	768	user-registered callback will be called), and will return after one
		769	iteration of the loop.
		770
		771	This is useful if you are waiting for some external event in conjunction
		772	with something not expressible using other libev watchers (i.e. "roll your
500	libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is	773	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
501	usually a better approach for this kind of thing.	774	usually a better approach for this kind of thing.
502		775
503	Here are the gory details of what C<ev_loop> does:	776	Here are the gory details of what C<ev_run> does:
504		777
		778	- Increment loop depth.
		779	- Reset the ev_break status.
505	- Before the first iteration, call any pending watchers.	780	- Before the first iteration, call any pending watchers.
506	* If there are no active watchers (reference count is zero), return.	781	LOOP:
507	- Queue all prepare watchers and then call all outstanding watchers.	782	- If EVFLAG_FORKCHECK was used, check for a fork.
		783	- If a fork was detected (by any means), queue and call all fork watchers.
		784	- Queue and call all prepare watchers.
		785	- If ev_break was called, goto FINISH.
508	- If we have been forked, recreate the kernel state.	786	- If we have been forked, detach and recreate the kernel state
		787	as to not disturb the other process.
509	- Update the kernel state with all outstanding changes.	788	- Update the kernel state with all outstanding changes.
510	- Update the "event loop time".	789	- Update the "event loop time" (ev_now ()).
511	- Calculate for how long to block.	790	- Calculate for how long to sleep or block, if at all
		791	(active idle watchers, EVRUN_NOWAIT or not having
		792	any active watchers at all will result in not sleeping).
		793	- Sleep if the I/O and timer collect interval say so.
		794	- Increment loop iteration counter.
512	- Block the process, waiting for any events.	795	- Block the process, waiting for any events.
513	- Queue all outstanding I/O (fd) events.	796	- Queue all outstanding I/O (fd) events.
514	- Update the "event loop time" and do time jump handling.	797	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
515	- Queue all outstanding timers.	798	- Queue all expired timers.
516	- Queue all outstanding periodics.	799	- Queue all expired periodics.
517	- If no events are pending now, queue all idle watchers.	800	- Queue all idle watchers with priority higher than that of pending events.
518	- Queue all check watchers.	801	- Queue all check watchers.
519	- Call all queued watchers in reverse order (i.e. check watchers first).	802	- Call all queued watchers in reverse order (i.e. check watchers first).
520	Signals and child watchers are implemented as I/O watchers, and will	803	Signals and child watchers are implemented as I/O watchers, and will
521	be handled here by queueing them when their watcher gets executed.	804	be handled here by queueing them when their watcher gets executed.
522	- If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	805	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
523	were used, return, otherwise continue with step *.	806	were used, or there are no active watchers, goto FINISH, otherwise
		807	continue with step LOOP.
		808	FINISH:
		809	- Reset the ev_break status iff it was EVBREAK_ONE.
		810	- Decrement the loop depth.
		811	- Return.
524		812
525	Example: Queue some jobs and then loop until no events are outsanding	813	Example: Queue some jobs and then loop until no events are outstanding
526	anymore.	814	anymore.
527		815
528	... queue jobs here, make sure they register event watchers as long	816	... queue jobs here, make sure they register event watchers as long
529	... as they still have work to do (even an idle watcher will do..)	817	... as they still have work to do (even an idle watcher will do..)
530	ev_loop (my_loop, 0);	818	ev_run (my_loop, 0);
531	... jobs done. yeah!	819	... jobs done or somebody called unloop. yeah!
532		820
533	=item ev_unloop (loop, how)	821	=item ev_break (loop, how)
534		822
535	Can be used to make a call to C<ev_loop> return early (but only after it	823	Can be used to make a call to C<ev_run> return early (but only after it
536	has processed all outstanding events). The C<how> argument must be either	824	has processed all outstanding events). The C<how> argument must be either
537	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	825	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
538	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	826	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		827
		828	This "break state" will be cleared when entering C<ev_run> again.
		829
		830	It is safe to call C<ev_break> from outside any C<ev_run> calls, too.
539		831
540	=item ev_ref (loop)	832	=item ev_ref (loop)
541		833
542	=item ev_unref (loop)	834	=item ev_unref (loop)
543		835
544	Ref/unref can be used to add or remove a reference count on the event	836	Ref/unref can be used to add or remove a reference count on the event
545	loop: Every watcher keeps one reference, and as long as the reference	837	loop: Every watcher keeps one reference, and as long as the reference
546	count is nonzero, C<ev_loop> will not return on its own. If you have	838	count is nonzero, C<ev_run> will not return on its own.
547	a watcher you never unregister that should not keep C<ev_loop> from	839
548	returning, ev_unref() after starting, and ev_ref() before stopping it. For	840	This is useful when you have a watcher that you never intend to
		841	unregister, but that nevertheless should not keep C<ev_run> from
		842	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		843	before stopping it.
		844
549	example, libev itself uses this for its internal signal pipe: It is not	845	As an example, libev itself uses this for its internal signal pipe: It
550	visible to the libev user and should not keep C<ev_loop> from exiting if	846	is not visible to the libev user and should not keep C<ev_run> from
551	no event watchers registered by it are active. It is also an excellent	847	exiting if no event watchers registered by it are active. It is also an
552	way to do this for generic recurring timers or from within third-party	848	excellent way to do this for generic recurring timers or from within
553	libraries. Just remember to I<unref after start> and I<ref before stop>.	849	third-party libraries. Just remember to I<unref after start> and I<ref
		850	before stop> (but only if the watcher wasn't active before, or was active
		851	before, respectively. Note also that libev might stop watchers itself
		852	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		853	in the callback).
554		854
555	Example: Create a signal watcher, but keep it from keeping C<ev_loop>	855	Example: Create a signal watcher, but keep it from keeping C<ev_run>
556	running when nothing else is active.	856	running when nothing else is active.
557		857
558	struct ev_signal exitsig;	858	ev_signal exitsig;
559	ev_signal_init (&exitsig, sig_cb, SIGINT);	859	ev_signal_init (&exitsig, sig_cb, SIGINT);
560	ev_signal_start (loop, &exitsig);	860	ev_signal_start (loop, &exitsig);
561	evf_unref (loop);	861	evf_unref (loop);
562		862
563	Example: For some weird reason, unregister the above signal handler again.	863	Example: For some weird reason, unregister the above signal handler again.
564		864
565	ev_ref (loop);	865	ev_ref (loop);
566	ev_signal_stop (loop, &exitsig);	866	ev_signal_stop (loop, &exitsig);
		867
		868	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		869
		870	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		871
		872	These advanced functions influence the time that libev will spend waiting
		873	for events. Both time intervals are by default C<0>, meaning that libev
		874	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		875	latency.
		876
		877	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		878	allows libev to delay invocation of I/O and timer/periodic callbacks
		879	to increase efficiency of loop iterations (or to increase power-saving
		880	opportunities).
		881
		882	The idea is that sometimes your program runs just fast enough to handle
		883	one (or very few) event(s) per loop iteration. While this makes the
		884	program responsive, it also wastes a lot of CPU time to poll for new
		885	events, especially with backends like C<select ()> which have a high
		886	overhead for the actual polling but can deliver many events at once.
		887
		888	By setting a higher I<io collect interval> you allow libev to spend more
		889	time collecting I/O events, so you can handle more events per iteration,
		890	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		891	C<ev_timer>) will be not affected. Setting this to a non-null value will
		892	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		893	sleep time ensures that libev will not poll for I/O events more often then
		894	once per this interval, on average.
		895
		896	Likewise, by setting a higher I<timeout collect interval> you allow libev
		897	to spend more time collecting timeouts, at the expense of increased
		898	latency/jitter/inexactness (the watcher callback will be called
		899	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		900	value will not introduce any overhead in libev.
		901
		902	Many (busy) programs can usually benefit by setting the I/O collect
		903	interval to a value near C<0.1> or so, which is often enough for
		904	interactive servers (of course not for games), likewise for timeouts. It
		905	usually doesn't make much sense to set it to a lower value than C<0.01>,
		906	as this approaches the timing granularity of most systems. Note that if
		907	you do transactions with the outside world and you can't increase the
		908	parallelity, then this setting will limit your transaction rate (if you
		909	need to poll once per transaction and the I/O collect interval is 0.01,
		910	then you can't do more than 100 transactions per second).
		911
		912	Setting the I<timeout collect interval> can improve the opportunity for
		913	saving power, as the program will "bundle" timer callback invocations that
		914	are "near" in time together, by delaying some, thus reducing the number of
		915	times the process sleeps and wakes up again. Another useful technique to
		916	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		917	they fire on, say, one-second boundaries only.
		918
		919	Example: we only need 0.1s timeout granularity, and we wish not to poll
		920	more often than 100 times per second:
		921
		922	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		923	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		924
		925	=item ev_invoke_pending (loop)
		926
		927	This call will simply invoke all pending watchers while resetting their
		928	pending state. Normally, C<ev_run> does this automatically when required,
		929	but when overriding the invoke callback this call comes handy. This
		930	function can be invoked from a watcher - this can be useful for example
		931	when you want to do some lengthy calculation and want to pass further
		932	event handling to another thread (you still have to make sure only one
		933	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		934
		935	=item int ev_pending_count (loop)
		936
		937	Returns the number of pending watchers - zero indicates that no watchers
		938	are pending.
		939
		940	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		941
		942	This overrides the invoke pending functionality of the loop: Instead of
		943	invoking all pending watchers when there are any, C<ev_run> will call
		944	this callback instead. This is useful, for example, when you want to
		945	invoke the actual watchers inside another context (another thread etc.).
		946
		947	If you want to reset the callback, use C<ev_invoke_pending> as new
		948	callback.
		949
		950	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		951
		952	Sometimes you want to share the same loop between multiple threads. This
		953	can be done relatively simply by putting mutex_lock/unlock calls around
		954	each call to a libev function.
		955
		956	However, C<ev_run> can run an indefinite time, so it is not feasible
		957	to wait for it to return. One way around this is to wake up the event
		958	loop via C<ev_break> and C<av_async_send>, another way is to set these
		959	I<release> and I<acquire> callbacks on the loop.
		960
		961	When set, then C<release> will be called just before the thread is
		962	suspended waiting for new events, and C<acquire> is called just
		963	afterwards.
		964
		965	Ideally, C<release> will just call your mutex_unlock function, and
		966	C<acquire> will just call the mutex_lock function again.
		967
		968	While event loop modifications are allowed between invocations of
		969	C<release> and C<acquire> (that's their only purpose after all), no
		970	modifications done will affect the event loop, i.e. adding watchers will
		971	have no effect on the set of file descriptors being watched, or the time
		972	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		973	to take note of any changes you made.
		974
		975	In theory, threads executing C<ev_run> will be async-cancel safe between
		976	invocations of C<release> and C<acquire>.
		977
		978	See also the locking example in the C<THREADS> section later in this
		979	document.
		980
		981	=item ev_set_userdata (loop, void *data)
		982
		983	=item ev_userdata (loop)
		984
		985	Set and retrieve a single C<void *> associated with a loop. When
		986	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		987	C<0.>
		988
		989	These two functions can be used to associate arbitrary data with a loop,
		990	and are intended solely for the C<invoke_pending_cb>, C<release> and
		991	C<acquire> callbacks described above, but of course can be (ab-)used for
		992	any other purpose as well.
		993
		994	=item ev_verify (loop)
		995
		996	This function only does something when C<EV_VERIFY> support has been
		997	compiled in, which is the default for non-minimal builds. It tries to go
		998	through all internal structures and checks them for validity. If anything
		999	is found to be inconsistent, it will print an error message to standard
		1000	error and call C<abort ()>.
		1001
		1002	This can be used to catch bugs inside libev itself: under normal
		1003	circumstances, this function will never abort as of course libev keeps its
		1004	data structures consistent.
567		1005
568	=back	1006	=back
569		1007
570		1008
571	=head1 ANATOMY OF A WATCHER	1009	=head1 ANATOMY OF A WATCHER
572		1010
		1011	In the following description, uppercase C<TYPE> in names stands for the
		1012	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1013	watchers and C<ev_io_start> for I/O watchers.
		1014
573	A watcher is a structure that you create and register to record your	1015	A watcher is an opaque structure that you allocate and register to record
574	interest in some event. For instance, if you want to wait for STDIN to	1016	your interest in some event. To make a concrete example, imagine you want
575	become readable, you would create an C<ev_io> watcher for that:	1017	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1018	for that:
576		1019
577	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1020	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
578	{	1021	{
579	ev_io_stop (w);	1022	ev_io_stop (w);
580	ev_unloop (loop, EVUNLOOP_ALL);	1023	ev_break (loop, EVBREAK_ALL);
581	}	1024	}
582		1025
583	struct ev_loop *loop = ev_default_loop (0);	1026	struct ev_loop *loop = ev_default_loop (0);
		1027
584	struct ev_io stdin_watcher;	1028	ev_io stdin_watcher;
		1029
585	ev_init (&stdin_watcher, my_cb);	1030	ev_init (&stdin_watcher, my_cb);
586	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1031	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
587	ev_io_start (loop, &stdin_watcher);	1032	ev_io_start (loop, &stdin_watcher);
		1033
588	ev_loop (loop, 0);	1034	ev_run (loop, 0);
589		1035
590	As you can see, you are responsible for allocating the memory for your	1036	As you can see, you are responsible for allocating the memory for your
591	watcher structures (and it is usually a bad idea to do this on the stack,	1037	watcher structures (and it is I<usually> a bad idea to do this on the
592	although this can sometimes be quite valid).	1038	stack).
593		1039
		1040	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1041	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1042
594	Each watcher structure must be initialised by a call to C<ev_init	1043	Each watcher structure must be initialised by a call to C<ev_init (watcher
595	(watcher *, callback)>, which expects a callback to be provided. This	1044	*, callback)>, which expects a callback to be provided. This callback is
596	callback gets invoked each time the event occurs (or, in the case of io	1045	invoked each time the event occurs (or, in the case of I/O watchers, each
597	watchers, each time the event loop detects that the file descriptor given	1046	time the event loop detects that the file descriptor given is readable
598	is readable and/or writable).	1047	and/or writable).
599		1048
600	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1049	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
601	with arguments specific to this watcher type. There is also a macro	1050	macro to configure it, with arguments specific to the watcher type. There
602	to combine initialisation and setting in one call: C<< ev_<type>_init	1051	is also a macro to combine initialisation and setting in one call: C<<
603	(watcher *, callback, ...) >>.	1052	ev_TYPE_init (watcher *, callback, ...) >>.
604		1053
605	To make the watcher actually watch out for events, you have to start it	1054	To make the watcher actually watch out for events, you have to start it
606	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1055	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
607	*) >>), and you can stop watching for events at any time by calling the	1056	*) >>), and you can stop watching for events at any time by calling the
608	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1057	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
609		1058
610	As long as your watcher is active (has been started but not stopped) you	1059	As long as your watcher is active (has been started but not stopped) you
611	must not touch the values stored in it. Most specifically you must never	1060	must not touch the values stored in it. Most specifically you must never
612	reinitialise it or call its C<set> macro.	1061	reinitialise it or call its C<ev_TYPE_set> macro.
613		1062
614	Each and every callback receives the event loop pointer as first, the	1063	Each and every callback receives the event loop pointer as first, the
615	registered watcher structure as second, and a bitset of received events as	1064	registered watcher structure as second, and a bitset of received events as
616	third argument.	1065	third argument.
617		1066
…		…
626	=item C<EV_WRITE>	1075	=item C<EV_WRITE>
627		1076
628	The file descriptor in the C<ev_io> watcher has become readable and/or	1077	The file descriptor in the C<ev_io> watcher has become readable and/or
629	writable.	1078	writable.
630		1079
631	=item C<EV_TIMEOUT>	1080	=item C<EV_TIMER>
632		1081
633	The C<ev_timer> watcher has timed out.	1082	The C<ev_timer> watcher has timed out.
634		1083
635	=item C<EV_PERIODIC>	1084	=item C<EV_PERIODIC>
636		1085
…		…
654		1103
655	=item C<EV_PREPARE>	1104	=item C<EV_PREPARE>
656		1105
657	=item C<EV_CHECK>	1106	=item C<EV_CHECK>
658		1107
659	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1108	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
660	to gather new events, and all C<ev_check> watchers are invoked just after	1109	to gather new events, and all C<ev_check> watchers are invoked just after
661	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1110	C<ev_run> has gathered them, but before it invokes any callbacks for any
662	received events. Callbacks of both watcher types can start and stop as	1111	received events. Callbacks of both watcher types can start and stop as
663	many watchers as they want, and all of them will be taken into account	1112	many watchers as they want, and all of them will be taken into account
664	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1113	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
665	C<ev_loop> from blocking).	1114	C<ev_run> from blocking).
666		1115
667	=item C<EV_EMBED>	1116	=item C<EV_EMBED>
668		1117
669	The embedded event loop specified in the C<ev_embed> watcher needs attention.	1118	The embedded event loop specified in the C<ev_embed> watcher needs attention.
670		1119
671	=item C<EV_FORK>	1120	=item C<EV_FORK>
672		1121
673	The event loop has been resumed in the child process after fork (see	1122	The event loop has been resumed in the child process after fork (see
674	C<ev_fork>).	1123	C<ev_fork>).
675		1124
		1125	=item C<EV_CLEANUP>
		1126
		1127	The event loop is about to be destroyed (see C<ev_cleanup>).
		1128
		1129	=item C<EV_ASYNC>
		1130
		1131	The given async watcher has been asynchronously notified (see C<ev_async>).
		1132
		1133	=item C<EV_CUSTOM>
		1134
		1135	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1136	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1137
676	=item C<EV_ERROR>	1138	=item C<EV_ERROR>
677		1139
678	An unspecified error has occured, the watcher has been stopped. This might	1140	An unspecified error has occurred, the watcher has been stopped. This might
679	happen because the watcher could not be properly started because libev	1141	happen because the watcher could not be properly started because libev
680	ran out of memory, a file descriptor was found to be closed or any other	1142	ran out of memory, a file descriptor was found to be closed or any other
		1143	problem. Libev considers these application bugs.
		1144
681	problem. You best act on it by reporting the problem and somehow coping	1145	You best act on it by reporting the problem and somehow coping with the
682	with the watcher being stopped.	1146	watcher being stopped. Note that well-written programs should not receive
		1147	an error ever, so when your watcher receives it, this usually indicates a
		1148	bug in your program.
683		1149
684	Libev will usually signal a few "dummy" events together with an error,	1150	Libev will usually signal a few "dummy" events together with an error, for
685	for example it might indicate that a fd is readable or writable, and if	1151	example it might indicate that a fd is readable or writable, and if your
686	your callbacks is well-written it can just attempt the operation and cope	1152	callbacks is well-written it can just attempt the operation and cope with
687	with the error from read() or write(). This will not work in multithreaded	1153	the error from read() or write(). This will not work in multi-threaded
688	programs, though, so beware.	1154	programs, though, as the fd could already be closed and reused for another
		1155	thing, so beware.
689		1156
690	=back	1157	=back
691		1158
692	=head2 GENERIC WATCHER FUNCTIONS	1159	=head2 GENERIC WATCHER FUNCTIONS
693
694	In the following description, C<TYPE> stands for the watcher type,
695	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
696		1160
697	=over 4	1161	=over 4
698		1162
699	=item C<ev_init> (ev_TYPE *watcher, callback)	1163	=item C<ev_init> (ev_TYPE *watcher, callback)
700		1164
…		…
706	which rolls both calls into one.	1170	which rolls both calls into one.
707		1171
708	You can reinitialise a watcher at any time as long as it has been stopped	1172	You can reinitialise a watcher at any time as long as it has been stopped
709	(or never started) and there are no pending events outstanding.	1173	(or never started) and there are no pending events outstanding.
710		1174
711	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,	1175	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
712	int revents)>.	1176	int revents)>.
713		1177
		1178	Example: Initialise an C<ev_io> watcher in two steps.
		1179
		1180	ev_io w;
		1181	ev_init (&w, my_cb);
		1182	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1183
714	=item C<ev_TYPE_set> (ev_TYPE *, [args])	1184	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
715		1185
716	This macro initialises the type-specific parts of a watcher. You need to	1186	This macro initialises the type-specific parts of a watcher. You need to
717	call C<ev_init> at least once before you call this macro, but you can	1187	call C<ev_init> at least once before you call this macro, but you can
718	call C<ev_TYPE_set> any number of times. You must not, however, call this	1188	call C<ev_TYPE_set> any number of times. You must not, however, call this
719	macro on a watcher that is active (it can be pending, however, which is a	1189	macro on a watcher that is active (it can be pending, however, which is a
720	difference to the C<ev_init> macro).	1190	difference to the C<ev_init> macro).
721		1191
722	Although some watcher types do not have type-specific arguments	1192	Although some watcher types do not have type-specific arguments
723	(e.g. C<ev_prepare>) you still need to call its C<set> macro.	1193	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
724		1194
		1195	See C<ev_init>, above, for an example.
		1196
725	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])	1197	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
726		1198
727	This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro	1199	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
728	calls into a single call. This is the most convinient method to initialise	1200	calls into a single call. This is the most convenient method to initialise
729	a watcher. The same limitations apply, of course.	1201	a watcher. The same limitations apply, of course.
730		1202
		1203	Example: Initialise and set an C<ev_io> watcher in one step.
		1204
		1205	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1206
731	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)	1207	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
732		1208
733	Starts (activates) the given watcher. Only active watchers will receive	1209	Starts (activates) the given watcher. Only active watchers will receive
734	events. If the watcher is already active nothing will happen.	1210	events. If the watcher is already active nothing will happen.
735		1211
		1212	Example: Start the C<ev_io> watcher that is being abused as example in this
		1213	whole section.
		1214
		1215	ev_io_start (EV_DEFAULT_UC, &w);
		1216
736	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)	1217	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
737		1218
738	Stops the given watcher again (if active) and clears the pending	1219	Stops the given watcher if active, and clears the pending status (whether
		1220	the watcher was active or not).
		1221
739	status. It is possible that stopped watchers are pending (for example,	1222	It is possible that stopped watchers are pending - for example,
740	non-repeating timers are being stopped when they become pending), but	1223	non-repeating timers are being stopped when they become pending - but
741	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If	1224	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
742	you want to free or reuse the memory used by the watcher it is therefore a	1225	pending. If you want to free or reuse the memory used by the watcher it is
743	good idea to always call its C<ev_TYPE_stop> function.	1226	therefore a good idea to always call its C<ev_TYPE_stop> function.
744		1227
745	=item bool ev_is_active (ev_TYPE *watcher)	1228	=item bool ev_is_active (ev_TYPE *watcher)
746		1229
747	Returns a true value iff the watcher is active (i.e. it has been started	1230	Returns a true value iff the watcher is active (i.e. it has been started
748	and not yet been stopped). As long as a watcher is active you must not modify	1231	and not yet been stopped). As long as a watcher is active you must not modify
…		…
764	=item ev_cb_set (ev_TYPE *watcher, callback)	1247	=item ev_cb_set (ev_TYPE *watcher, callback)
765		1248
766	Change the callback. You can change the callback at virtually any time	1249	Change the callback. You can change the callback at virtually any time
767	(modulo threads).	1250	(modulo threads).
768		1251
769	=item ev_set_priority (ev_TYPE *watcher, priority)	1252	=item ev_set_priority (ev_TYPE *watcher, int priority)
770		1253
771	=item int ev_priority (ev_TYPE *watcher)	1254	=item int ev_priority (ev_TYPE *watcher)
772		1255
773	Set and query the priority of the watcher. The priority is a small	1256	Set and query the priority of the watcher. The priority is a small
774	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>	1257	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
775	(default: C<-2>). Pending watchers with higher priority will be invoked	1258	(default: C<-2>). Pending watchers with higher priority will be invoked
776	before watchers with lower priority, but priority will not keep watchers	1259	before watchers with lower priority, but priority will not keep watchers
777	from being executed (except for C<ev_idle> watchers).	1260	from being executed (except for C<ev_idle> watchers).
778		1261
779	This means that priorities are I<only> used for ordering callback
780	invocation after new events have been received. This is useful, for
781	example, to reduce latency after idling, or more often, to bind two
782	watchers on the same event and make sure one is called first.
783
784	If you need to suppress invocation when higher priority events are pending	1262	If you need to suppress invocation when higher priority events are pending
785	you need to look at C<ev_idle> watchers, which provide this functionality.	1263	you need to look at C<ev_idle> watchers, which provide this functionality.
786		1264
787	You I<must not> change the priority of a watcher as long as it is active or	1265	You I<must not> change the priority of a watcher as long as it is active or
788	pending.	1266	pending.
789		1267
		1268	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1269	fine, as long as you do not mind that the priority value you query might
		1270	or might not have been clamped to the valid range.
		1271
790	The default priority used by watchers when no priority has been set is	1272	The default priority used by watchers when no priority has been set is
791	always C<0>, which is supposed to not be too high and not be too low :).	1273	always C<0>, which is supposed to not be too high and not be too low :).
792		1274
793	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is	1275	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
794	fine, as long as you do not mind that the priority value you query might	1276	priorities.
795	or might not have been adjusted to be within valid range.
796		1277
797	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1278	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
798		1279
799	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1280	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
800	C<loop> nor C<revents> need to be valid as long as the watcher callback	1281	C<loop> nor C<revents> need to be valid as long as the watcher callback
801	can deal with that fact.	1282	can deal with that fact, as both are simply passed through to the
		1283	callback.
802		1284
803	=item int ev_clear_pending (loop, ev_TYPE *watcher)	1285	=item int ev_clear_pending (loop, ev_TYPE *watcher)
804		1286
805	If the watcher is pending, this function returns clears its pending status	1287	If the watcher is pending, this function clears its pending status and
806	and returns its C<revents> bitset (as if its callback was invoked). If the	1288	returns its C<revents> bitset (as if its callback was invoked). If the
807	watcher isn't pending it does nothing and returns C<0>.	1289	watcher isn't pending it does nothing and returns C<0>.
808		1290
		1291	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1292	callback to be invoked, which can be accomplished with this function.
		1293
		1294	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1295
		1296	Feeds the given event set into the event loop, as if the specified event
		1297	had happened for the specified watcher (which must be a pointer to an
		1298	initialised but not necessarily started event watcher). Obviously you must
		1299	not free the watcher as long as it has pending events.
		1300
		1301	Stopping the watcher, letting libev invoke it, or calling
		1302	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1303	not started in the first place.
		1304
		1305	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1306	functions that do not need a watcher.
		1307
809	=back	1308	=back
810		1309
811
812	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1310	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
813		1311
814	Each watcher has, by default, a member C<void *data> that you can change	1312	Each watcher has, by default, a member C<void *data> that you can change
815	and read at any time, libev will completely ignore it. This can be used	1313	and read at any time: libev will completely ignore it. This can be used
816	to associate arbitrary data with your watcher. If you need more data and	1314	to associate arbitrary data with your watcher. If you need more data and
817	don't want to allocate memory and store a pointer to it in that data	1315	don't want to allocate memory and store a pointer to it in that data
818	member, you can also "subclass" the watcher type and provide your own	1316	member, you can also "subclass" the watcher type and provide your own
819	data:	1317	data:
820		1318
821	struct my_io	1319	struct my_io
822	{	1320	{
823	struct ev_io io;	1321	ev_io io;
824	int otherfd;	1322	int otherfd;
825	void *somedata;	1323	void *somedata;
826	struct whatever *mostinteresting;	1324	struct whatever *mostinteresting;
827	}	1325	};
		1326
		1327	...
		1328	struct my_io w;
		1329	ev_io_init (&w.io, my_cb, fd, EV_READ);
828		1330
829	And since your callback will be called with a pointer to the watcher, you	1331	And since your callback will be called with a pointer to the watcher, you
830	can cast it back to your own type:	1332	can cast it back to your own type:
831		1333
832	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1334	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
833	{	1335	{
834	struct my_io *w = (struct my_io *)w_;	1336	struct my_io *w = (struct my_io *)w_;
835	...	1337	...
836	}	1338	}
837		1339
838	More interesting and less C-conformant ways of casting your callback type	1340	More interesting and less C-conformant ways of casting your callback type
839	instead have been omitted.	1341	instead have been omitted.
840		1342
841	Another common scenario is having some data structure with multiple	1343	Another common scenario is to use some data structure with multiple
842	watchers:	1344	embedded watchers:
843		1345
844	struct my_biggy	1346	struct my_biggy
845	{	1347	{
846	int some_data;	1348	int some_data;
847	ev_timer t1;	1349	ev_timer t1;
848	ev_timer t2;	1350	ev_timer t2;
849	}	1351	}
850		1352
851	In this case getting the pointer to C<my_biggy> is a bit more complicated,	1353	In this case getting the pointer to C<my_biggy> is a bit more
852	you need to use C<offsetof>:	1354	complicated: Either you store the address of your C<my_biggy> struct
		1355	in the C<data> member of the watcher (for woozies), or you need to use
		1356	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1357	programmers):
853		1358
854	#include <stddef.h>	1359	#include <stddef.h>
855		1360
856	static void	1361	static void
857	t1_cb (EV_P_ struct ev_timer *w, int revents)	1362	t1_cb (EV_P_ ev_timer *w, int revents)
858	{	1363	{
859	struct my_biggy big = (struct my_biggy *	1364	struct my_biggy big = (struct my_biggy *)
860	(((char *)w) - offsetof (struct my_biggy, t1));	1365	(((char *)w) - offsetof (struct my_biggy, t1));
861	}	1366	}
862		1367
863	static void	1368	static void
864	t2_cb (EV_P_ struct ev_timer *w, int revents)	1369	t2_cb (EV_P_ ev_timer *w, int revents)
865	{	1370	{
866	struct my_biggy big = (struct my_biggy *	1371	struct my_biggy big = (struct my_biggy *)
867	(((char *)w) - offsetof (struct my_biggy, t2));	1372	(((char *)w) - offsetof (struct my_biggy, t2));
868	}	1373	}
		1374
		1375	=head2 WATCHER STATES
		1376
		1377	There are various watcher states mentioned throughout this manual -
		1378	active, pending and so on. In this section these states and the rules to
		1379	transition between them will be described in more detail - and while these
		1380	rules might look complicated, they usually do "the right thing".
		1381
		1382	=over 4
		1383
		1384	=item initialiased
		1385
		1386	Before a watcher can be registered with the event looop it has to be
		1387	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1388	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1389
		1390	In this state it is simply some block of memory that is suitable for use
		1391	in an event loop. It can be moved around, freed, reused etc. at will.
		1392
		1393	=item started/running/active
		1394
		1395	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1396	property of the event loop, and is actively waiting for events. While in
		1397	this state it cannot be accessed (except in a few documented ways), moved,
		1398	freed or anything else - the only legal thing is to keep a pointer to it,
		1399	and call libev functions on it that are documented to work on active watchers.
		1400
		1401	=item pending
		1402
		1403	If a watcher is active and libev determines that an event it is interested
		1404	in has occurred (such as a timer expiring), it will become pending. It will
		1405	stay in this pending state until either it is stopped or its callback is
		1406	about to be invoked, so it is not normally pending inside the watcher
		1407	callback.
		1408
		1409	The watcher might or might not be active while it is pending (for example,
		1410	an expired non-repeating timer can be pending but no longer active). If it
		1411	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1412	but it is still property of the event loop at this time, so cannot be
		1413	moved, freed or reused. And if it is active the rules described in the
		1414	previous item still apply.
		1415
		1416	It is also possible to feed an event on a watcher that is not active (e.g.
		1417	via C<ev_feed_event>), in which case it becomes pending without being
		1418	active.
		1419
		1420	=item stopped
		1421
		1422	A watcher can be stopped implicitly by libev (in which case it might still
		1423	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1424	latter will clear any pending state the watcher might be in, regardless
		1425	of whether it was active or not, so stopping a watcher explicitly before
		1426	freeing it is often a good idea.
		1427
		1428	While stopped (and not pending) the watcher is essentially in the
		1429	initialised state, that is it can be reused, moved, modified in any way
		1430	you wish.
		1431
		1432	=back
		1433
		1434	=head2 WATCHER PRIORITY MODELS
		1435
		1436	Many event loops support I<watcher priorities>, which are usually small
		1437	integers that influence the ordering of event callback invocation
		1438	between watchers in some way, all else being equal.
		1439
		1440	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1441	description for the more technical details such as the actual priority
		1442	range.
		1443
		1444	There are two common ways how these these priorities are being interpreted
		1445	by event loops:
		1446
		1447	In the more common lock-out model, higher priorities "lock out" invocation
		1448	of lower priority watchers, which means as long as higher priority
		1449	watchers receive events, lower priority watchers are not being invoked.
		1450
		1451	The less common only-for-ordering model uses priorities solely to order
		1452	callback invocation within a single event loop iteration: Higher priority
		1453	watchers are invoked before lower priority ones, but they all get invoked
		1454	before polling for new events.
		1455
		1456	Libev uses the second (only-for-ordering) model for all its watchers
		1457	except for idle watchers (which use the lock-out model).
		1458
		1459	The rationale behind this is that implementing the lock-out model for
		1460	watchers is not well supported by most kernel interfaces, and most event
		1461	libraries will just poll for the same events again and again as long as
		1462	their callbacks have not been executed, which is very inefficient in the
		1463	common case of one high-priority watcher locking out a mass of lower
		1464	priority ones.
		1465
		1466	Static (ordering) priorities are most useful when you have two or more
		1467	watchers handling the same resource: a typical usage example is having an
		1468	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1469	timeouts. Under load, data might be received while the program handles
		1470	other jobs, but since timers normally get invoked first, the timeout
		1471	handler will be executed before checking for data. In that case, giving
		1472	the timer a lower priority than the I/O watcher ensures that I/O will be
		1473	handled first even under adverse conditions (which is usually, but not
		1474	always, what you want).
		1475
		1476	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1477	will only be executed when no same or higher priority watchers have
		1478	received events, they can be used to implement the "lock-out" model when
		1479	required.
		1480
		1481	For example, to emulate how many other event libraries handle priorities,
		1482	you can associate an C<ev_idle> watcher to each such watcher, and in
		1483	the normal watcher callback, you just start the idle watcher. The real
		1484	processing is done in the idle watcher callback. This causes libev to
		1485	continuously poll and process kernel event data for the watcher, but when
		1486	the lock-out case is known to be rare (which in turn is rare :), this is
		1487	workable.
		1488
		1489	Usually, however, the lock-out model implemented that way will perform
		1490	miserably under the type of load it was designed to handle. In that case,
		1491	it might be preferable to stop the real watcher before starting the
		1492	idle watcher, so the kernel will not have to process the event in case
		1493	the actual processing will be delayed for considerable time.
		1494
		1495	Here is an example of an I/O watcher that should run at a strictly lower
		1496	priority than the default, and which should only process data when no
		1497	other events are pending:
		1498
		1499	ev_idle idle; // actual processing watcher
		1500	ev_io io; // actual event watcher
		1501
		1502	static void
		1503	io_cb (EV_P_ ev_io *w, int revents)
		1504	{
		1505	// stop the I/O watcher, we received the event, but
		1506	// are not yet ready to handle it.
		1507	ev_io_stop (EV_A_ w);
		1508
		1509	// start the idle watcher to handle the actual event.
		1510	// it will not be executed as long as other watchers
		1511	// with the default priority are receiving events.
		1512	ev_idle_start (EV_A_ &idle);
		1513	}
		1514
		1515	static void
		1516	idle_cb (EV_P_ ev_idle *w, int revents)
		1517	{
		1518	// actual processing
		1519	read (STDIN_FILENO, ...);
		1520
		1521	// have to start the I/O watcher again, as
		1522	// we have handled the event
		1523	ev_io_start (EV_P_ &io);
		1524	}
		1525
		1526	// initialisation
		1527	ev_idle_init (&idle, idle_cb);
		1528	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1529	ev_io_start (EV_DEFAULT_ &io);
		1530
		1531	In the "real" world, it might also be beneficial to start a timer, so that
		1532	low-priority connections can not be locked out forever under load. This
		1533	enables your program to keep a lower latency for important connections
		1534	during short periods of high load, while not completely locking out less
		1535	important ones.
869		1536
870		1537
871	=head1 WATCHER TYPES	1538	=head1 WATCHER TYPES
872		1539
873	This section describes each watcher in detail, but will not repeat	1540	This section describes each watcher in detail, but will not repeat
…		…
897	In general you can register as many read and/or write event watchers per	1564	In general you can register as many read and/or write event watchers per
898	fd as you want (as long as you don't confuse yourself). Setting all file	1565	fd as you want (as long as you don't confuse yourself). Setting all file
899	descriptors to non-blocking mode is also usually a good idea (but not	1566	descriptors to non-blocking mode is also usually a good idea (but not
900	required if you know what you are doing).	1567	required if you know what you are doing).
901		1568
902	You have to be careful with dup'ed file descriptors, though. Some backends	1569	If you cannot use non-blocking mode, then force the use of a
903	(the linux epoll backend is a notable example) cannot handle dup'ed file	1570	known-to-be-good backend (at the time of this writing, this includes only
904	descriptors correctly if you register interest in two or more fds pointing	1571	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
905	to the same underlying file/socket/etc. description (that is, they share	1572	descriptors for which non-blocking operation makes no sense (such as
906	the same underlying "file open").	1573	files) - libev doesn't guarantee any specific behaviour in that case.
907
908	If you must do this, then force the use of a known-to-be-good backend
909	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
910	C<EVBACKEND_POLL>).
911		1574
912	Another thing you have to watch out for is that it is quite easy to	1575	Another thing you have to watch out for is that it is quite easy to
913	receive "spurious" readyness notifications, that is your callback might	1576	receive "spurious" readiness notifications, that is your callback might
914	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1577	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
915	because there is no data. Not only are some backends known to create a	1578	because there is no data. Not only are some backends known to create a
916	lot of those (for example solaris ports), it is very easy to get into	1579	lot of those (for example Solaris ports), it is very easy to get into
917	this situation even with a relatively standard program structure. Thus	1580	this situation even with a relatively standard program structure. Thus
918	it is best to always use non-blocking I/O: An extra C<read>(2) returning	1581	it is best to always use non-blocking I/O: An extra C<read>(2) returning
919	C<EAGAIN> is far preferable to a program hanging until some data arrives.	1582	C<EAGAIN> is far preferable to a program hanging until some data arrives.
920		1583
921	If you cannot run the fd in non-blocking mode (for example you should not	1584	If you cannot run the fd in non-blocking mode (for example you should
922	play around with an Xlib connection), then you have to seperately re-test	1585	not play around with an Xlib connection), then you have to separately
923	whether a file descriptor is really ready with a known-to-be good interface	1586	re-test whether a file descriptor is really ready with a known-to-be good
924	such as poll (fortunately in our Xlib example, Xlib already does this on	1587	interface such as poll (fortunately in our Xlib example, Xlib already
925	its own, so its quite safe to use).	1588	does this on its own, so its quite safe to use). Some people additionally
		1589	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1590	indefinitely.
		1591
		1592	But really, best use non-blocking mode.
926		1593
927	=head3 The special problem of disappearing file descriptors	1594	=head3 The special problem of disappearing file descriptors
928		1595
929	Some backends (e.g kqueue, epoll) need to be told about closing a file	1596	Some backends (e.g. kqueue, epoll) need to be told about closing a file
930	descriptor (either by calling C<close> explicitly or by any other means,	1597	descriptor (either due to calling C<close> explicitly or any other means,
931	such as C<dup>). The reason is that you register interest in some file	1598	such as C<dup2>). The reason is that you register interest in some file
932	descriptor, but when it goes away, the operating system will silently drop	1599	descriptor, but when it goes away, the operating system will silently drop
933	this interest. If another file descriptor with the same number then is	1600	this interest. If another file descriptor with the same number then is
934	registered with libev, there is no efficient way to see that this is, in	1601	registered with libev, there is no efficient way to see that this is, in
935	fact, a different file descriptor.	1602	fact, a different file descriptor.
936		1603
…		…
943		1610
944	This is how one would do it normally anyway, the important point is that	1611	This is how one would do it normally anyway, the important point is that
945	the libev application should not optimise around libev but should leave	1612	the libev application should not optimise around libev but should leave
946	optimisations to libev.	1613	optimisations to libev.
947		1614
		1615	=head3 The special problem of dup'ed file descriptors
		1616
		1617	Some backends (e.g. epoll), cannot register events for file descriptors,
		1618	but only events for the underlying file descriptions. That means when you
		1619	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1620	events for them, only one file descriptor might actually receive events.
		1621
		1622	There is no workaround possible except not registering events
		1623	for potentially C<dup ()>'ed file descriptors, or to resort to
		1624	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1625
		1626	=head3 The special problem of fork
		1627
		1628	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1629	useless behaviour. Libev fully supports fork, but needs to be told about
		1630	it in the child.
		1631
		1632	To support fork in your programs, you either have to call
		1633	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1634	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1635	C<EVBACKEND_POLL>.
		1636
		1637	=head3 The special problem of SIGPIPE
		1638
		1639	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1640	when writing to a pipe whose other end has been closed, your program gets
		1641	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1642	this is sensible behaviour, for daemons, this is usually undesirable.
		1643
		1644	So when you encounter spurious, unexplained daemon exits, make sure you
		1645	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1646	somewhere, as that would have given you a big clue).
		1647
		1648	=head3 The special problem of accept()ing when you can't
		1649
		1650	Many implementations of the POSIX C<accept> function (for example,
		1651	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1652	connection from the pending queue in all error cases.
		1653
		1654	For example, larger servers often run out of file descriptors (because
		1655	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1656	rejecting the connection, leading to libev signalling readiness on
		1657	the next iteration again (the connection still exists after all), and
		1658	typically causing the program to loop at 100% CPU usage.
		1659
		1660	Unfortunately, the set of errors that cause this issue differs between
		1661	operating systems, there is usually little the app can do to remedy the
		1662	situation, and no known thread-safe method of removing the connection to
		1663	cope with overload is known (to me).
		1664
		1665	One of the easiest ways to handle this situation is to just ignore it
		1666	- when the program encounters an overload, it will just loop until the
		1667	situation is over. While this is a form of busy waiting, no OS offers an
		1668	event-based way to handle this situation, so it's the best one can do.
		1669
		1670	A better way to handle the situation is to log any errors other than
		1671	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1672	messages, and continue as usual, which at least gives the user an idea of
		1673	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1674	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1675	usage.
		1676
		1677	If your program is single-threaded, then you could also keep a dummy file
		1678	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1679	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1680	close that fd, and create a new dummy fd. This will gracefully refuse
		1681	clients under typical overload conditions.
		1682
		1683	The last way to handle it is to simply log the error and C<exit>, as
		1684	is often done with C<malloc> failures, but this results in an easy
		1685	opportunity for a DoS attack.
948		1686
949	=head3 Watcher-Specific Functions	1687	=head3 Watcher-Specific Functions
950		1688
951	=over 4	1689	=over 4
952		1690
953	=item ev_io_init (ev_io *, callback, int fd, int events)	1691	=item ev_io_init (ev_io *, callback, int fd, int events)
954		1692
955	=item ev_io_set (ev_io *, int fd, int events)	1693	=item ev_io_set (ev_io *, int fd, int events)
956		1694
957	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to	1695	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
958	rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or	1696	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
959	C<EV_READ | EV_WRITE> to receive the given events.	1697	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
960		1698
961	=item int fd [read-only]	1699	=item int fd [read-only]
962		1700
963	The file descriptor being watched.	1701	The file descriptor being watched.
964		1702
965	=item int events [read-only]	1703	=item int events [read-only]
966		1704
967	The events being watched.	1705	The events being watched.
968		1706
969	=back	1707	=back
		1708
		1709	=head3 Examples
970		1710
971	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well	1711	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
972	readable, but only once. Since it is likely line-buffered, you could	1712	readable, but only once. Since it is likely line-buffered, you could
973	attempt to read a whole line in the callback.	1713	attempt to read a whole line in the callback.
974		1714
975	static void	1715	static void
976	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1716	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
977	{	1717	{
978	ev_io_stop (loop, w);	1718	ev_io_stop (loop, w);
979	.. read from stdin here (or from w->fd) and haqndle any I/O errors	1719	.. read from stdin here (or from w->fd) and handle any I/O errors
980	}	1720	}
981		1721
982	...	1722	...
983	struct ev_loop *loop = ev_default_init (0);	1723	struct ev_loop *loop = ev_default_init (0);
984	struct ev_io stdin_readable;	1724	ev_io stdin_readable;
985	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);	1725	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
986	ev_io_start (loop, &stdin_readable);	1726	ev_io_start (loop, &stdin_readable);
987	ev_loop (loop, 0);	1727	ev_run (loop, 0);
988		1728
989		1729
990	=head2 C<ev_timer> - relative and optionally repeating timeouts	1730	=head2 C<ev_timer> - relative and optionally repeating timeouts
991		1731
992	Timer watchers are simple relative timers that generate an event after a	1732	Timer watchers are simple relative timers that generate an event after a
993	given time, and optionally repeating in regular intervals after that.	1733	given time, and optionally repeating in regular intervals after that.
994		1734
995	The timers are based on real time, that is, if you register an event that	1735	The timers are based on real time, that is, if you register an event that
996	times out after an hour and you reset your system clock to last years	1736	times out after an hour and you reset your system clock to January last
997	time, it will still time out after (roughly) and hour. "Roughly" because	1737	year, it will still time out after (roughly) one hour. "Roughly" because
998	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1738	detecting time jumps is hard, and some inaccuracies are unavoidable (the
999	monotonic clock option helps a lot here).	1739	monotonic clock option helps a lot here).
		1740
		1741	The callback is guaranteed to be invoked only I<after> its timeout has
		1742	passed (not I<at>, so on systems with very low-resolution clocks this
		1743	might introduce a small delay). If multiple timers become ready during the
		1744	same loop iteration then the ones with earlier time-out values are invoked
		1745	before ones of the same priority with later time-out values (but this is
		1746	no longer true when a callback calls C<ev_run> recursively).
		1747
		1748	=head3 Be smart about timeouts
		1749
		1750	Many real-world problems involve some kind of timeout, usually for error
		1751	recovery. A typical example is an HTTP request - if the other side hangs,
		1752	you want to raise some error after a while.
		1753
		1754	What follows are some ways to handle this problem, from obvious and
		1755	inefficient to smart and efficient.
		1756
		1757	In the following, a 60 second activity timeout is assumed - a timeout that
		1758	gets reset to 60 seconds each time there is activity (e.g. each time some
		1759	data or other life sign was received).
		1760
		1761	=over 4
		1762
		1763	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1764
		1765	This is the most obvious, but not the most simple way: In the beginning,
		1766	start the watcher:
		1767
		1768	ev_timer_init (timer, callback, 60., 0.);
		1769	ev_timer_start (loop, timer);
		1770
		1771	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1772	and start it again:
		1773
		1774	ev_timer_stop (loop, timer);
		1775	ev_timer_set (timer, 60., 0.);
		1776	ev_timer_start (loop, timer);
		1777
		1778	This is relatively simple to implement, but means that each time there is
		1779	some activity, libev will first have to remove the timer from its internal
		1780	data structure and then add it again. Libev tries to be fast, but it's
		1781	still not a constant-time operation.
		1782
		1783	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1784
		1785	This is the easiest way, and involves using C<ev_timer_again> instead of
		1786	C<ev_timer_start>.
		1787
		1788	To implement this, configure an C<ev_timer> with a C<repeat> value
		1789	of C<60> and then call C<ev_timer_again> at start and each time you
		1790	successfully read or write some data. If you go into an idle state where
		1791	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1792	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1793
		1794	That means you can ignore both the C<ev_timer_start> function and the
		1795	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1796	member and C<ev_timer_again>.
		1797
		1798	At start:
		1799
		1800	ev_init (timer, callback);
		1801	timer->repeat = 60.;
		1802	ev_timer_again (loop, timer);
		1803
		1804	Each time there is some activity:
		1805
		1806	ev_timer_again (loop, timer);
		1807
		1808	It is even possible to change the time-out on the fly, regardless of
		1809	whether the watcher is active or not:
		1810
		1811	timer->repeat = 30.;
		1812	ev_timer_again (loop, timer);
		1813
		1814	This is slightly more efficient then stopping/starting the timer each time
		1815	you want to modify its timeout value, as libev does not have to completely
		1816	remove and re-insert the timer from/into its internal data structure.
		1817
		1818	It is, however, even simpler than the "obvious" way to do it.
		1819
		1820	=item 3. Let the timer time out, but then re-arm it as required.
		1821
		1822	This method is more tricky, but usually most efficient: Most timeouts are
		1823	relatively long compared to the intervals between other activity - in
		1824	our example, within 60 seconds, there are usually many I/O events with
		1825	associated activity resets.
		1826
		1827	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1828	but remember the time of last activity, and check for a real timeout only
		1829	within the callback:
		1830
		1831	ev_tstamp last_activity; // time of last activity
		1832
		1833	static void
		1834	callback (EV_P_ ev_timer *w, int revents)
		1835	{
		1836	ev_tstamp now = ev_now (EV_A);
		1837	ev_tstamp timeout = last_activity + 60.;
		1838
		1839	// if last_activity + 60. is older than now, we did time out
		1840	if (timeout < now)
		1841	{
		1842	// timeout occurred, take action
		1843	}
		1844	else
		1845	{
		1846	// callback was invoked, but there was some activity, re-arm
		1847	// the watcher to fire in last_activity + 60, which is
		1848	// guaranteed to be in the future, so "again" is positive:
		1849	w->repeat = timeout - now;
		1850	ev_timer_again (EV_A_ w);
		1851	}
		1852	}
		1853
		1854	To summarise the callback: first calculate the real timeout (defined
		1855	as "60 seconds after the last activity"), then check if that time has
		1856	been reached, which means something I<did>, in fact, time out. Otherwise
		1857	the callback was invoked too early (C<timeout> is in the future), so
		1858	re-schedule the timer to fire at that future time, to see if maybe we have
		1859	a timeout then.
		1860
		1861	Note how C<ev_timer_again> is used, taking advantage of the
		1862	C<ev_timer_again> optimisation when the timer is already running.
		1863
		1864	This scheme causes more callback invocations (about one every 60 seconds
		1865	minus half the average time between activity), but virtually no calls to
		1866	libev to change the timeout.
		1867
		1868	To start the timer, simply initialise the watcher and set C<last_activity>
		1869	to the current time (meaning we just have some activity :), then call the
		1870	callback, which will "do the right thing" and start the timer:
		1871
		1872	ev_init (timer, callback);
		1873	last_activity = ev_now (loop);
		1874	callback (loop, timer, EV_TIMER);
		1875
		1876	And when there is some activity, simply store the current time in
		1877	C<last_activity>, no libev calls at all:
		1878
		1879	last_activity = ev_now (loop);
		1880
		1881	This technique is slightly more complex, but in most cases where the
		1882	time-out is unlikely to be triggered, much more efficient.
		1883
		1884	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1885	callback :) - just change the timeout and invoke the callback, which will
		1886	fix things for you.
		1887
		1888	=item 4. Wee, just use a double-linked list for your timeouts.
		1889
		1890	If there is not one request, but many thousands (millions...), all
		1891	employing some kind of timeout with the same timeout value, then one can
		1892	do even better:
		1893
		1894	When starting the timeout, calculate the timeout value and put the timeout
		1895	at the I<end> of the list.
		1896
		1897	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1898	the list is expected to fire (for example, using the technique #3).
		1899
		1900	When there is some activity, remove the timer from the list, recalculate
		1901	the timeout, append it to the end of the list again, and make sure to
		1902	update the C<ev_timer> if it was taken from the beginning of the list.
		1903
		1904	This way, one can manage an unlimited number of timeouts in O(1) time for
		1905	starting, stopping and updating the timers, at the expense of a major
		1906	complication, and having to use a constant timeout. The constant timeout
		1907	ensures that the list stays sorted.
		1908
		1909	=back
		1910
		1911	So which method the best?
		1912
		1913	Method #2 is a simple no-brain-required solution that is adequate in most
		1914	situations. Method #3 requires a bit more thinking, but handles many cases
		1915	better, and isn't very complicated either. In most case, choosing either
		1916	one is fine, with #3 being better in typical situations.
		1917
		1918	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1919	rather complicated, but extremely efficient, something that really pays
		1920	off after the first million or so of active timers, i.e. it's usually
		1921	overkill :)
		1922
		1923	=head3 The special problem of time updates
		1924
		1925	Establishing the current time is a costly operation (it usually takes at
		1926	least two system calls): EV therefore updates its idea of the current
		1927	time only before and after C<ev_run> collects new events, which causes a
		1928	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1929	lots of events in one iteration.
1000		1930
1001	The relative timeouts are calculated relative to the C<ev_now ()>	1931	The relative timeouts are calculated relative to the C<ev_now ()>
1002	time. This is usually the right thing as this timestamp refers to the time	1932	time. This is usually the right thing as this timestamp refers to the time
1003	of the event triggering whatever timeout you are modifying/starting. If	1933	of the event triggering whatever timeout you are modifying/starting. If
1004	you suspect event processing to be delayed and you I<need> to base the timeout	1934	you suspect event processing to be delayed and you I<need> to base the
1005	on the current time, use something like this to adjust for this:	1935	timeout on the current time, use something like this to adjust for this:
1006		1936
1007	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1937	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
1008		1938
1009	The callback is guarenteed to be invoked only when its timeout has passed,	1939	If the event loop is suspended for a long time, you can also force an
1010	but if multiple timers become ready during the same loop iteration then	1940	update of the time returned by C<ev_now ()> by calling C<ev_now_update
1011	order of execution is undefined.	1941	()>.
		1942
		1943	=head3 The special problems of suspended animation
		1944
		1945	When you leave the server world it is quite customary to hit machines that
		1946	can suspend/hibernate - what happens to the clocks during such a suspend?
		1947
		1948	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1949	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1950	to run until the system is suspended, but they will not advance while the
		1951	system is suspended. That means, on resume, it will be as if the program
		1952	was frozen for a few seconds, but the suspend time will not be counted
		1953	towards C<ev_timer> when a monotonic clock source is used. The real time
		1954	clock advanced as expected, but if it is used as sole clocksource, then a
		1955	long suspend would be detected as a time jump by libev, and timers would
		1956	be adjusted accordingly.
		1957
		1958	I would not be surprised to see different behaviour in different between
		1959	operating systems, OS versions or even different hardware.
		1960
		1961	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1962	time jump in the monotonic clocks and the realtime clock. If the program
		1963	is suspended for a very long time, and monotonic clock sources are in use,
		1964	then you can expect C<ev_timer>s to expire as the full suspension time
		1965	will be counted towards the timers. When no monotonic clock source is in
		1966	use, then libev will again assume a timejump and adjust accordingly.
		1967
		1968	It might be beneficial for this latter case to call C<ev_suspend>
		1969	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1970	deterministic behaviour in this case (you can do nothing against
		1971	C<SIGSTOP>).
1012		1972
1013	=head3 Watcher-Specific Functions and Data Members	1973	=head3 Watcher-Specific Functions and Data Members
1014		1974
1015	=over 4	1975	=over 4
1016		1976
1017	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1977	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
1018		1978
1019	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1979	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
1020		1980
1021	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1981	Configure the timer to trigger after C<after> seconds. If C<repeat>
1022	C<0.>, then it will automatically be stopped. If it is positive, then the	1982	is C<0.>, then it will automatically be stopped once the timeout is
1023	timer will automatically be configured to trigger again C<repeat> seconds	1983	reached. If it is positive, then the timer will automatically be
1024	later, again, and again, until stopped manually.	1984	configured to trigger again C<repeat> seconds later, again, and again,
		1985	until stopped manually.
1025		1986
1026	The timer itself will do a best-effort at avoiding drift, that is, if you	1987	The timer itself will do a best-effort at avoiding drift, that is, if
1027	configure a timer to trigger every 10 seconds, then it will trigger at	1988	you configure a timer to trigger every 10 seconds, then it will normally
1028	exactly 10 second intervals. If, however, your program cannot keep up with	1989	trigger at exactly 10 second intervals. If, however, your program cannot
1029	the timer (because it takes longer than those 10 seconds to do stuff) the	1990	keep up with the timer (because it takes longer than those 10 seconds to
1030	timer will not fire more than once per event loop iteration.	1991	do stuff) the timer will not fire more than once per event loop iteration.
1031		1992
1032	=item ev_timer_again (loop)	1993	=item ev_timer_again (loop, ev_timer *)
1033		1994
1034	This will act as if the timer timed out and restart it again if it is	1995	This will act as if the timer timed out and restart it again if it is
1035	repeating. The exact semantics are:	1996	repeating. The exact semantics are:
1036		1997
1037	If the timer is pending, its pending status is cleared.	1998	If the timer is pending, its pending status is cleared.
1038		1999
1039	If the timer is started but nonrepeating, stop it (as if it timed out).	2000	If the timer is started but non-repeating, stop it (as if it timed out).
1040		2001
1041	If the timer is repeating, either start it if necessary (with the	2002	If the timer is repeating, either start it if necessary (with the
1042	C<repeat> value), or reset the running timer to the C<repeat> value.	2003	C<repeat> value), or reset the running timer to the C<repeat> value.
1043		2004
1044	This sounds a bit complicated, but here is a useful and typical	2005	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
1045	example: Imagine you have a tcp connection and you want a so-called idle	2006	usage example.
1046	timeout, that is, you want to be called when there have been, say, 60
1047	seconds of inactivity on the socket. The easiest way to do this is to
1048	configure an C<ev_timer> with a C<repeat> value of C<60> and then call
1049	C<ev_timer_again> each time you successfully read or write some data. If
1050	you go into an idle state where you do not expect data to travel on the
1051	socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
1052	automatically restart it if need be.
1053		2007
1054	That means you can ignore the C<after> value and C<ev_timer_start>	2008	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
1055	altogether and only ever use the C<repeat> value and C<ev_timer_again>:
1056		2009
1057	ev_timer_init (timer, callback, 0., 5.);	2010	Returns the remaining time until a timer fires. If the timer is active,
1058	ev_timer_again (loop, timer);	2011	then this time is relative to the current event loop time, otherwise it's
1059	...	2012	the timeout value currently configured.
1060	timer->again = 17.;
1061	ev_timer_again (loop, timer);
1062	...
1063	timer->again = 10.;
1064	ev_timer_again (loop, timer);
1065		2013
1066	This is more slightly efficient then stopping/starting the timer each time	2014	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
1067	you want to modify its timeout value.	2015	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2016	will return C<4>. When the timer expires and is restarted, it will return
		2017	roughly C<7> (likely slightly less as callback invocation takes some time,
		2018	too), and so on.
1068		2019
1069	=item ev_tstamp repeat [read-write]	2020	=item ev_tstamp repeat [read-write]
1070		2021
1071	The current C<repeat> value. Will be used each time the watcher times out	2022	The current C<repeat> value. Will be used each time the watcher times out
1072	or C<ev_timer_again> is called and determines the next timeout (if any),	2023	or C<ev_timer_again> is called, and determines the next timeout (if any),
1073	which is also when any modifications are taken into account.	2024	which is also when any modifications are taken into account.
1074		2025
1075	=back	2026	=back
1076		2027
		2028	=head3 Examples
		2029
1077	Example: Create a timer that fires after 60 seconds.	2030	Example: Create a timer that fires after 60 seconds.
1078		2031
1079	static void	2032	static void
1080	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2033	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
1081	{	2034	{
1082	.. one minute over, w is actually stopped right here	2035	.. one minute over, w is actually stopped right here
1083	}	2036	}
1084		2037
1085	struct ev_timer mytimer;	2038	ev_timer mytimer;
1086	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);	2039	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
1087	ev_timer_start (loop, &mytimer);	2040	ev_timer_start (loop, &mytimer);
1088		2041
1089	Example: Create a timeout timer that times out after 10 seconds of	2042	Example: Create a timeout timer that times out after 10 seconds of
1090	inactivity.	2043	inactivity.
1091		2044
1092	static void	2045	static void
1093	timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)	2046	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
1094	{	2047	{
1095	.. ten seconds without any activity	2048	.. ten seconds without any activity
1096	}	2049	}
1097		2050
1098	struct ev_timer mytimer;	2051	ev_timer mytimer;
1099	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */	2052	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
1100	ev_timer_again (&mytimer); /* start timer */	2053	ev_timer_again (&mytimer); /* start timer */
1101	ev_loop (loop, 0);	2054	ev_run (loop, 0);
1102		2055
1103	// and in some piece of code that gets executed on any "activity":	2056	// and in some piece of code that gets executed on any "activity":
1104	// reset the timeout to start ticking again at 10 seconds	2057	// reset the timeout to start ticking again at 10 seconds
1105	ev_timer_again (&mytimer);	2058	ev_timer_again (&mytimer);
1106		2059
1107		2060
1108	=head2 C<ev_periodic> - to cron or not to cron?	2061	=head2 C<ev_periodic> - to cron or not to cron?
1109		2062
1110	Periodic watchers are also timers of a kind, but they are very versatile	2063	Periodic watchers are also timers of a kind, but they are very versatile
1111	(and unfortunately a bit complex).	2064	(and unfortunately a bit complex).
1112		2065
1113	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2066	Unlike C<ev_timer>, periodic watchers are not based on real time (or
1114	but on wallclock time (absolute time). You can tell a periodic watcher	2067	relative time, the physical time that passes) but on wall clock time
1115	to trigger "at" some specific point in time. For example, if you tell a	2068	(absolute time, the thing you can read on your calender or clock). The
1116	periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()	2069	difference is that wall clock time can run faster or slower than real
1117	+ 10.>) and then reset your system clock to the last year, then it will	2070	time, and time jumps are not uncommon (e.g. when you adjust your
1118	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2071	wrist-watch).
1119	roughly 10 seconds later).
1120		2072
1121	They can also be used to implement vastly more complex timers, such as	2073	You can tell a periodic watcher to trigger after some specific point
1122	triggering an event on each midnight, local time or other, complicated,	2074	in time: for example, if you tell a periodic watcher to trigger "in 10
1123	rules.	2075	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2076	not a delay) and then reset your system clock to January of the previous
		2077	year, then it will take a year or more to trigger the event (unlike an
		2078	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2079	it, as it uses a relative timeout).
1124		2080
		2081	C<ev_periodic> watchers can also be used to implement vastly more complex
		2082	timers, such as triggering an event on each "midnight, local time", or
		2083	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		2084	those cannot react to time jumps.
		2085
1125	As with timers, the callback is guarenteed to be invoked only when the	2086	As with timers, the callback is guaranteed to be invoked only when the
1126	time (C<at>) has been passed, but if multiple periodic timers become ready	2087	point in time where it is supposed to trigger has passed. If multiple
1127	during the same loop iteration then order of execution is undefined.	2088	timers become ready during the same loop iteration then the ones with
		2089	earlier time-out values are invoked before ones with later time-out values
		2090	(but this is no longer true when a callback calls C<ev_run> recursively).
1128		2091
1129	=head3 Watcher-Specific Functions and Data Members	2092	=head3 Watcher-Specific Functions and Data Members
1130		2093
1131	=over 4	2094	=over 4
1132		2095
1133	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2096	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1134		2097
1135	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2098	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1136		2099
1137	Lots of arguments, lets sort it out... There are basically three modes of	2100	Lots of arguments, let's sort it out... There are basically three modes of
1138	operation, and we will explain them from simplest to complex:	2101	operation, and we will explain them from simplest to most complex:
1139		2102
1140	=over 4	2103	=over 4
1141		2104
1142	=item * absolute timer (at = time, interval = reschedule_cb = 0)	2105	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
1143		2106
1144	In this configuration the watcher triggers an event at the wallclock time	2107	In this configuration the watcher triggers an event after the wall clock
1145	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2108	time C<offset> has passed. It will not repeat and will not adjust when a
1146	that is, if it is to be run at January 1st 2011 then it will run when the	2109	time jump occurs, that is, if it is to be run at January 1st 2011 then it
1147	system time reaches or surpasses this time.	2110	will be stopped and invoked when the system clock reaches or surpasses
		2111	this point in time.
1148		2112
1149	=item * non-repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	2113	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
1150		2114
1151	In this mode the watcher will always be scheduled to time out at the next	2115	In this mode the watcher will always be scheduled to time out at the next
1152	C<at + N * interval> time (for some integer N, which can also be negative)	2116	C<offset + N * interval> time (for some integer N, which can also be
1153	and then repeat, regardless of any time jumps.	2117	negative) and then repeat, regardless of any time jumps. The C<offset>
		2118	argument is merely an offset into the C<interval> periods.
1154		2119
1155	This can be used to create timers that do not drift with respect to system	2120	This can be used to create timers that do not drift with respect to the
1156	time:	2121	system clock, for example, here is an C<ev_periodic> that triggers each
		2122	hour, on the hour (with respect to UTC):
1157		2123
1158	ev_periodic_set (&periodic, 0., 3600., 0);	2124	ev_periodic_set (&periodic, 0., 3600., 0);
1159		2125
1160	This doesn't mean there will always be 3600 seconds in between triggers,	2126	This doesn't mean there will always be 3600 seconds in between triggers,
1161	but only that the the callback will be called when the system time shows a	2127	but only that the callback will be called when the system time shows a
1162	full hour (UTC), or more correctly, when the system time is evenly divisible	2128	full hour (UTC), or more correctly, when the system time is evenly divisible
1163	by 3600.	2129	by 3600.
1164		2130
1165	Another way to think about it (for the mathematically inclined) is that	2131	Another way to think about it (for the mathematically inclined) is that
1166	C<ev_periodic> will try to run the callback in this mode at the next possible	2132	C<ev_periodic> will try to run the callback in this mode at the next possible
1167	time where C<time = at (mod interval)>, regardless of any time jumps.	2133	time where C<time = offset (mod interval)>, regardless of any time jumps.
1168		2134
1169	For numerical stability it is preferable that the C<at> value is near	2135	For numerical stability it is preferable that the C<offset> value is near
1170	C<ev_now ()> (the current time), but there is no range requirement for	2136	C<ev_now ()> (the current time), but there is no range requirement for
1171	this value.	2137	this value, and in fact is often specified as zero.
1172		2138
		2139	Note also that there is an upper limit to how often a timer can fire (CPU
		2140	speed for example), so if C<interval> is very small then timing stability
		2141	will of course deteriorate. Libev itself tries to be exact to be about one
		2142	millisecond (if the OS supports it and the machine is fast enough).
		2143
1173	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)	2144	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
1174		2145
1175	In this mode the values for C<interval> and C<at> are both being	2146	In this mode the values for C<interval> and C<offset> are both being
1176	ignored. Instead, each time the periodic watcher gets scheduled, the	2147	ignored. Instead, each time the periodic watcher gets scheduled, the
1177	reschedule callback will be called with the watcher as first, and the	2148	reschedule callback will be called with the watcher as first, and the
1178	current time as second argument.	2149	current time as second argument.
1179		2150
1180	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2151	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
1181	ever, or make any event loop modifications>. If you need to stop it,	2152	or make ANY other event loop modifications whatsoever, unless explicitly
1182	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2153	allowed by documentation here>.
1183	starting an C<ev_prepare> watcher, which is legal).
1184		2154
		2155	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2156	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2157	only event loop modification you are allowed to do).
		2158
1185	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2159	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
1186	ev_tstamp now)>, e.g.:	2160	*w, ev_tstamp now)>, e.g.:
1187		2161
		2162	static ev_tstamp
1188	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2163	my_rescheduler (ev_periodic *w, ev_tstamp now)
1189	{	2164	{
1190	return now + 60.;	2165	return now + 60.;
1191	}	2166	}
1192		2167
1193	It must return the next time to trigger, based on the passed time value	2168	It must return the next time to trigger, based on the passed time value
1194	(that is, the lowest time value larger than to the second argument). It	2169	(that is, the lowest time value larger than to the second argument). It
1195	will usually be called just before the callback will be triggered, but	2170	will usually be called just before the callback will be triggered, but
1196	might be called at other times, too.	2171	might be called at other times, too.
1197		2172
1198	NOTE: I<< This callback must always return a time that is later than the	2173	NOTE: I<< This callback must always return a time that is higher than or
1199	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2174	equal to the passed C<now> value >>.
1200		2175
1201	This can be used to create very complex timers, such as a timer that	2176	This can be used to create very complex timers, such as a timer that
1202	triggers on each midnight, local time. To do this, you would calculate the	2177	triggers on "next midnight, local time". To do this, you would calculate the
1203	next midnight after C<now> and return the timestamp value for this. How	2178	next midnight after C<now> and return the timestamp value for this. How
1204	you do this is, again, up to you (but it is not trivial, which is the main	2179	you do this is, again, up to you (but it is not trivial, which is the main
1205	reason I omitted it as an example).	2180	reason I omitted it as an example).
1206		2181
1207	=back	2182	=back
…		…
1211	Simply stops and restarts the periodic watcher again. This is only useful	2186	Simply stops and restarts the periodic watcher again. This is only useful
1212	when you changed some parameters or the reschedule callback would return	2187	when you changed some parameters or the reschedule callback would return
1213	a different time than the last time it was called (e.g. in a crond like	2188	a different time than the last time it was called (e.g. in a crond like
1214	program when the crontabs have changed).	2189	program when the crontabs have changed).
1215		2190
		2191	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2192
		2193	When active, returns the absolute time that the watcher is supposed
		2194	to trigger next. This is not the same as the C<offset> argument to
		2195	C<ev_periodic_set>, but indeed works even in interval and manual
		2196	rescheduling modes.
		2197
1216	=item ev_tstamp offset [read-write]	2198	=item ev_tstamp offset [read-write]
1217		2199
1218	When repeating, this contains the offset value, otherwise this is the	2200	When repeating, this contains the offset value, otherwise this is the
1219	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	2201	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2202	although libev might modify this value for better numerical stability).
1220		2203
1221	Can be modified any time, but changes only take effect when the periodic	2204	Can be modified any time, but changes only take effect when the periodic
1222	timer fires or C<ev_periodic_again> is being called.	2205	timer fires or C<ev_periodic_again> is being called.
1223		2206
1224	=item ev_tstamp interval [read-write]	2207	=item ev_tstamp interval [read-write]
1225		2208
1226	The current interval value. Can be modified any time, but changes only	2209	The current interval value. Can be modified any time, but changes only
1227	take effect when the periodic timer fires or C<ev_periodic_again> is being	2210	take effect when the periodic timer fires or C<ev_periodic_again> is being
1228	called.	2211	called.
1229		2212
1230	=item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]	2213	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
1231		2214
1232	The current reschedule callback, or C<0>, if this functionality is	2215	The current reschedule callback, or C<0>, if this functionality is
1233	switched off. Can be changed any time, but changes only take effect when	2216	switched off. Can be changed any time, but changes only take effect when
1234	the periodic timer fires or C<ev_periodic_again> is being called.	2217	the periodic timer fires or C<ev_periodic_again> is being called.
1235		2218
1236	=item ev_tstamp at [read-only]
1237
1238	When active, contains the absolute time that the watcher is supposed to
1239	trigger next.
1240
1241	=back	2219	=back
1242		2220
		2221	=head3 Examples
		2222
1243	Example: Call a callback every hour, or, more precisely, whenever the	2223	Example: Call a callback every hour, or, more precisely, whenever the
1244	system clock is divisible by 3600. The callback invocation times have	2224	system time is divisible by 3600. The callback invocation times have
1245	potentially a lot of jittering, but good long-term stability.	2225	potentially a lot of jitter, but good long-term stability.
1246		2226
1247	static void	2227	static void
1248	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)	2228	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
1249	{	2229	{
1250	... its now a full hour (UTC, or TAI or whatever your clock follows)	2230	... its now a full hour (UTC, or TAI or whatever your clock follows)
1251	}	2231	}
1252		2232
1253	struct ev_periodic hourly_tick;	2233	ev_periodic hourly_tick;
1254	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);	2234	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
1255	ev_periodic_start (loop, &hourly_tick);	2235	ev_periodic_start (loop, &hourly_tick);
1256		2236
1257	Example: The same as above, but use a reschedule callback to do it:	2237	Example: The same as above, but use a reschedule callback to do it:
1258		2238
1259	#include <math.h>	2239	#include <math.h>
1260		2240
1261	static ev_tstamp	2241	static ev_tstamp
1262	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)	2242	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
1263	{	2243	{
1264	return fmod (now, 3600.) + 3600.;	2244	return now + (3600. - fmod (now, 3600.));
1265	}	2245	}
1266		2246
1267	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);	2247	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
1268		2248
1269	Example: Call a callback every hour, starting now:	2249	Example: Call a callback every hour, starting now:
1270		2250
1271	struct ev_periodic hourly_tick;	2251	ev_periodic hourly_tick;
1272	ev_periodic_init (&hourly_tick, clock_cb,	2252	ev_periodic_init (&hourly_tick, clock_cb,
1273	fmod (ev_now (loop), 3600.), 3600., 0);	2253	fmod (ev_now (loop), 3600.), 3600., 0);
1274	ev_periodic_start (loop, &hourly_tick);	2254	ev_periodic_start (loop, &hourly_tick);
1275		2255
1276		2256
1277	=head2 C<ev_signal> - signal me when a signal gets signalled!	2257	=head2 C<ev_signal> - signal me when a signal gets signalled!
1278		2258
1279	Signal watchers will trigger an event when the process receives a specific	2259	Signal watchers will trigger an event when the process receives a specific
1280	signal one or more times. Even though signals are very asynchronous, libev	2260	signal one or more times. Even though signals are very asynchronous, libev
1281	will try it's best to deliver signals synchronously, i.e. as part of the	2261	will try it's best to deliver signals synchronously, i.e. as part of the
1282	normal event processing, like any other event.	2262	normal event processing, like any other event.
1283		2263
		2264	If you want signals to be delivered truly asynchronously, just use
		2265	C<sigaction> as you would do without libev and forget about sharing
		2266	the signal. You can even use C<ev_async> from a signal handler to
		2267	synchronously wake up an event loop.
		2268
1284	You can configure as many watchers as you like per signal. Only when the	2269	You can configure as many watchers as you like for the same signal, but
		2270	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2271	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2272	C<SIGINT> in both the default loop and another loop at the same time. At
		2273	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2274
1285	first watcher gets started will libev actually register a signal watcher	2275	When the first watcher gets started will libev actually register something
1286	with the kernel (thus it coexists with your own signal handlers as long	2276	with the kernel (thus it coexists with your own signal handlers as long as
1287	as you don't register any with libev). Similarly, when the last signal	2277	you don't register any with libev for the same signal).
1288	watcher for a signal is stopped libev will reset the signal handler to	2278
1289	SIG_DFL (regardless of what it was set to before).	2279	If possible and supported, libev will install its handlers with
		2280	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2281	not be unduly interrupted. If you have a problem with system calls getting
		2282	interrupted by signals you can block all signals in an C<ev_check> watcher
		2283	and unblock them in an C<ev_prepare> watcher.
		2284
		2285	=head3 The special problem of inheritance over fork/execve/pthread_create
		2286
		2287	Both the signal mask (C<sigprocmask>) and the signal disposition
		2288	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2289	stopping it again), that is, libev might or might not block the signal,
		2290	and might or might not set or restore the installed signal handler.
		2291
		2292	While this does not matter for the signal disposition (libev never
		2293	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2294	C<execve>), this matters for the signal mask: many programs do not expect
		2295	certain signals to be blocked.
		2296
		2297	This means that before calling C<exec> (from the child) you should reset
		2298	the signal mask to whatever "default" you expect (all clear is a good
		2299	choice usually).
		2300
		2301	The simplest way to ensure that the signal mask is reset in the child is
		2302	to install a fork handler with C<pthread_atfork> that resets it. That will
		2303	catch fork calls done by libraries (such as the libc) as well.
		2304
		2305	In current versions of libev, the signal will not be blocked indefinitely
		2306	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2307	the window of opportunity for problems, it will not go away, as libev
		2308	I<has> to modify the signal mask, at least temporarily.
		2309
		2310	So I can't stress this enough: I<If you do not reset your signal mask when
		2311	you expect it to be empty, you have a race condition in your code>. This
		2312	is not a libev-specific thing, this is true for most event libraries.
1290		2313
1291	=head3 Watcher-Specific Functions and Data Members	2314	=head3 Watcher-Specific Functions and Data Members
1292		2315
1293	=over 4	2316	=over 4
1294		2317
…		…
1303		2326
1304	The signal the watcher watches out for.	2327	The signal the watcher watches out for.
1305		2328
1306	=back	2329	=back
1307		2330
		2331	=head3 Examples
		2332
		2333	Example: Try to exit cleanly on SIGINT.
		2334
		2335	static void
		2336	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2337	{
		2338	ev_break (loop, EVBREAK_ALL);
		2339	}
		2340
		2341	ev_signal signal_watcher;
		2342	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2343	ev_signal_start (loop, &signal_watcher);
		2344
1308		2345
1309	=head2 C<ev_child> - watch out for process status changes	2346	=head2 C<ev_child> - watch out for process status changes
1310		2347
1311	Child watchers trigger when your process receives a SIGCHLD in response to	2348	Child watchers trigger when your process receives a SIGCHLD in response to
1312	some child status changes (most typically when a child of yours dies).	2349	some child status changes (most typically when a child of yours dies or
		2350	exits). It is permissible to install a child watcher I<after> the child
		2351	has been forked (which implies it might have already exited), as long
		2352	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2353	forking and then immediately registering a watcher for the child is fine,
		2354	but forking and registering a watcher a few event loop iterations later or
		2355	in the next callback invocation is not.
		2356
		2357	Only the default event loop is capable of handling signals, and therefore
		2358	you can only register child watchers in the default event loop.
		2359
		2360	Due to some design glitches inside libev, child watchers will always be
		2361	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2362	libev)
		2363
		2364	=head3 Process Interaction
		2365
		2366	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2367	initialised. This is necessary to guarantee proper behaviour even if the
		2368	first child watcher is started after the child exits. The occurrence
		2369	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2370	synchronously as part of the event loop processing. Libev always reaps all
		2371	children, even ones not watched.
		2372
		2373	=head3 Overriding the Built-In Processing
		2374
		2375	Libev offers no special support for overriding the built-in child
		2376	processing, but if your application collides with libev's default child
		2377	handler, you can override it easily by installing your own handler for
		2378	C<SIGCHLD> after initialising the default loop, and making sure the
		2379	default loop never gets destroyed. You are encouraged, however, to use an
		2380	event-based approach to child reaping and thus use libev's support for
		2381	that, so other libev users can use C<ev_child> watchers freely.
		2382
		2383	=head3 Stopping the Child Watcher
		2384
		2385	Currently, the child watcher never gets stopped, even when the
		2386	child terminates, so normally one needs to stop the watcher in the
		2387	callback. Future versions of libev might stop the watcher automatically
		2388	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2389	problem).
1313		2390
1314	=head3 Watcher-Specific Functions and Data Members	2391	=head3 Watcher-Specific Functions and Data Members
1315		2392
1316	=over 4	2393	=over 4
1317		2394
1318	=item ev_child_init (ev_child *, callback, int pid)	2395	=item ev_child_init (ev_child *, callback, int pid, int trace)
1319		2396
1320	=item ev_child_set (ev_child *, int pid)	2397	=item ev_child_set (ev_child *, int pid, int trace)
1321		2398
1322	Configures the watcher to wait for status changes of process C<pid> (or	2399	Configures the watcher to wait for status changes of process C<pid> (or
1323	I<any> process if C<pid> is specified as C<0>). The callback can look	2400	I<any> process if C<pid> is specified as C<0>). The callback can look
1324	at the C<rstatus> member of the C<ev_child> watcher structure to see	2401	at the C<rstatus> member of the C<ev_child> watcher structure to see
1325	the status word (use the macros from C<sys/wait.h> and see your systems	2402	the status word (use the macros from C<sys/wait.h> and see your systems
1326	C<waitpid> documentation). The C<rpid> member contains the pid of the	2403	C<waitpid> documentation). The C<rpid> member contains the pid of the
1327	process causing the status change.	2404	process causing the status change. C<trace> must be either C<0> (only
		2405	activate the watcher when the process terminates) or C<1> (additionally
		2406	activate the watcher when the process is stopped or continued).
1328		2407
1329	=item int pid [read-only]	2408	=item int pid [read-only]
1330		2409
1331	The process id this watcher watches out for, or C<0>, meaning any process id.	2410	The process id this watcher watches out for, or C<0>, meaning any process id.
1332		2411
…		…
1339	The process exit/trace status caused by C<rpid> (see your systems	2418	The process exit/trace status caused by C<rpid> (see your systems
1340	C<waitpid> and C<sys/wait.h> documentation for details).	2419	C<waitpid> and C<sys/wait.h> documentation for details).
1341		2420
1342	=back	2421	=back
1343		2422
1344	Example: Try to exit cleanly on SIGINT and SIGTERM.	2423	=head3 Examples
1345		2424
		2425	Example: C<fork()> a new process and install a child handler to wait for
		2426	its completion.
		2427
		2428	ev_child cw;
		2429
1346	static void	2430	static void
1347	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)	2431	child_cb (EV_P_ ev_child *w, int revents)
1348	{	2432	{
1349	ev_unloop (loop, EVUNLOOP_ALL);	2433	ev_child_stop (EV_A_ w);
		2434	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
1350	}	2435	}
1351		2436
1352	struct ev_signal signal_watcher;	2437	pid_t pid = fork ();
1353	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	2438
1354	ev_signal_start (loop, &sigint_cb);	2439	if (pid < 0)
		2440	// error
		2441	else if (pid == 0)
		2442	{
		2443	// the forked child executes here
		2444	exit (1);
		2445	}
		2446	else
		2447	{
		2448	ev_child_init (&cw, child_cb, pid, 0);
		2449	ev_child_start (EV_DEFAULT_ &cw);
		2450	}
1355		2451
1356		2452
1357	=head2 C<ev_stat> - did the file attributes just change?	2453	=head2 C<ev_stat> - did the file attributes just change?
1358		2454
1359	This watches a filesystem path for attribute changes. That is, it calls	2455	This watches a file system path for attribute changes. That is, it calls
1360	C<stat> regularly (or when the OS says it changed) and sees if it changed	2456	C<stat> on that path in regular intervals (or when the OS says it changed)
1361	compared to the last time, invoking the callback if it did.	2457	and sees if it changed compared to the last time, invoking the callback if
		2458	it did.
1362		2459
1363	The path does not need to exist: changing from "path exists" to "path does	2460	The path does not need to exist: changing from "path exists" to "path does
1364	not exist" is a status change like any other. The condition "path does	2461	not exist" is a status change like any other. The condition "path does not
1365	not exist" is signified by the C<st_nlink> field being zero (which is	2462	exist" (or more correctly "path cannot be stat'ed") is signified by the
1366	otherwise always forced to be at least one) and all the other fields of	2463	C<st_nlink> field being zero (which is otherwise always forced to be at
1367	the stat buffer having unspecified contents.	2464	least one) and all the other fields of the stat buffer having unspecified
		2465	contents.
1368		2466
1369	The path I<should> be absolute and I<must not> end in a slash. If it is	2467	The path I<must not> end in a slash or contain special components such as
		2468	C<.> or C<..>. The path I<should> be absolute: If it is relative and
1370	relative and your working directory changes, the behaviour is undefined.	2469	your working directory changes, then the behaviour is undefined.
1371		2470
1372	Since there is no standard to do this, the portable implementation simply	2471	Since there is no portable change notification interface available, the
1373	calls C<stat (2)> regularly on the path to see if it changed somehow. You	2472	portable implementation simply calls C<stat(2)> regularly on the path
1374	can specify a recommended polling interval for this case. If you specify	2473	to see if it changed somehow. You can specify a recommended polling
1375	a polling interval of C<0> (highly recommended!) then a I<suitable,	2474	interval for this case. If you specify a polling interval of C<0> (highly
1376	unspecified default> value will be used (which you can expect to be around	2475	recommended!) then a I<suitable, unspecified default> value will be used
1377	five seconds, although this might change dynamically). Libev will also	2476	(which you can expect to be around five seconds, although this might
1378	impose a minimum interval which is currently around C<0.1>, but thats	2477	change dynamically). Libev will also impose a minimum interval which is
1379	usually overkill.	2478	currently around C<0.1>, but that's usually overkill.
1380		2479
1381	This watcher type is not meant for massive numbers of stat watchers,	2480	This watcher type is not meant for massive numbers of stat watchers,
1382	as even with OS-supported change notifications, this can be	2481	as even with OS-supported change notifications, this can be
1383	resource-intensive.	2482	resource-intensive.
1384		2483
1385	At the time of this writing, only the Linux inotify interface is	2484	At the time of this writing, the only OS-specific interface implemented
1386	implemented (implementing kqueue support is left as an exercise for the	2485	is the Linux inotify interface (implementing kqueue support is left as an
1387	reader). Inotify will be used to give hints only and should not change the	2486	exercise for the reader. Note, however, that the author sees no way of
1388	semantics of C<ev_stat> watchers, which means that libev sometimes needs	2487	implementing C<ev_stat> semantics with kqueue, except as a hint).
1389	to fall back to regular polling again even with inotify, but changes are	2488
1390	usually detected immediately, and if the file exists there will be no	2489	=head3 ABI Issues (Largefile Support)
1391	polling.	2490
		2491	Libev by default (unless the user overrides this) uses the default
		2492	compilation environment, which means that on systems with large file
		2493	support disabled by default, you get the 32 bit version of the stat
		2494	structure. When using the library from programs that change the ABI to
		2495	use 64 bit file offsets the programs will fail. In that case you have to
		2496	compile libev with the same flags to get binary compatibility. This is
		2497	obviously the case with any flags that change the ABI, but the problem is
		2498	most noticeably displayed with ev_stat and large file support.
		2499
		2500	The solution for this is to lobby your distribution maker to make large
		2501	file interfaces available by default (as e.g. FreeBSD does) and not
		2502	optional. Libev cannot simply switch on large file support because it has
		2503	to exchange stat structures with application programs compiled using the
		2504	default compilation environment.
		2505
		2506	=head3 Inotify and Kqueue
		2507
		2508	When C<inotify (7)> support has been compiled into libev and present at
		2509	runtime, it will be used to speed up change detection where possible. The
		2510	inotify descriptor will be created lazily when the first C<ev_stat>
		2511	watcher is being started.
		2512
		2513	Inotify presence does not change the semantics of C<ev_stat> watchers
		2514	except that changes might be detected earlier, and in some cases, to avoid
		2515	making regular C<stat> calls. Even in the presence of inotify support
		2516	there are many cases where libev has to resort to regular C<stat> polling,
		2517	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2518	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2519	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2520	xfs are fully working) libev usually gets away without polling.
		2521
		2522	There is no support for kqueue, as apparently it cannot be used to
		2523	implement this functionality, due to the requirement of having a file
		2524	descriptor open on the object at all times, and detecting renames, unlinks
		2525	etc. is difficult.
		2526
		2527	=head3 C<stat ()> is a synchronous operation
		2528
		2529	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2530	the process. The exception are C<ev_stat> watchers - those call C<stat
		2531	()>, which is a synchronous operation.
		2532
		2533	For local paths, this usually doesn't matter: unless the system is very
		2534	busy or the intervals between stat's are large, a stat call will be fast,
		2535	as the path data is usually in memory already (except when starting the
		2536	watcher).
		2537
		2538	For networked file systems, calling C<stat ()> can block an indefinite
		2539	time due to network issues, and even under good conditions, a stat call
		2540	often takes multiple milliseconds.
		2541
		2542	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2543	paths, although this is fully supported by libev.
		2544
		2545	=head3 The special problem of stat time resolution
		2546
		2547	The C<stat ()> system call only supports full-second resolution portably,
		2548	and even on systems where the resolution is higher, most file systems
		2549	still only support whole seconds.
		2550
		2551	That means that, if the time is the only thing that changes, you can
		2552	easily miss updates: on the first update, C<ev_stat> detects a change and
		2553	calls your callback, which does something. When there is another update
		2554	within the same second, C<ev_stat> will be unable to detect unless the
		2555	stat data does change in other ways (e.g. file size).
		2556
		2557	The solution to this is to delay acting on a change for slightly more
		2558	than a second (or till slightly after the next full second boundary), using
		2559	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2560	ev_timer_again (loop, w)>).
		2561
		2562	The C<.02> offset is added to work around small timing inconsistencies
		2563	of some operating systems (where the second counter of the current time
		2564	might be be delayed. One such system is the Linux kernel, where a call to
		2565	C<gettimeofday> might return a timestamp with a full second later than
		2566	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2567	update file times then there will be a small window where the kernel uses
		2568	the previous second to update file times but libev might already execute
		2569	the timer callback).
1392		2570
1393	=head3 Watcher-Specific Functions and Data Members	2571	=head3 Watcher-Specific Functions and Data Members
1394		2572
1395	=over 4	2573	=over 4
1396		2574
…		…
1402	C<path>. The C<interval> is a hint on how quickly a change is expected to	2580	C<path>. The C<interval> is a hint on how quickly a change is expected to
1403	be detected and should normally be specified as C<0> to let libev choose	2581	be detected and should normally be specified as C<0> to let libev choose
1404	a suitable value. The memory pointed to by C<path> must point to the same	2582	a suitable value. The memory pointed to by C<path> must point to the same
1405	path for as long as the watcher is active.	2583	path for as long as the watcher is active.
1406		2584
1407	The callback will be receive C<EV_STAT> when a change was detected,	2585	The callback will receive an C<EV_STAT> event when a change was detected,
1408	relative to the attributes at the time the watcher was started (or the	2586	relative to the attributes at the time the watcher was started (or the
1409	last change was detected).	2587	last change was detected).
1410		2588
1411	=item ev_stat_stat (ev_stat *)	2589	=item ev_stat_stat (loop, ev_stat *)
1412		2590
1413	Updates the stat buffer immediately with new values. If you change the	2591	Updates the stat buffer immediately with new values. If you change the
1414	watched path in your callback, you could call this fucntion to avoid	2592	watched path in your callback, you could call this function to avoid
1415	detecting this change (while introducing a race condition). Can also be	2593	detecting this change (while introducing a race condition if you are not
1416	useful simply to find out the new values.	2594	the only one changing the path). Can also be useful simply to find out the
		2595	new values.
1417		2596
1418	=item ev_statdata attr [read-only]	2597	=item ev_statdata attr [read-only]
1419		2598
1420	The most-recently detected attributes of the file. Although the type is of	2599	The most-recently detected attributes of the file. Although the type is
1421	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types	2600	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
1422	suitable for your system. If the C<st_nlink> member is C<0>, then there	2601	suitable for your system, but you can only rely on the POSIX-standardised
		2602	members to be present. If the C<st_nlink> member is C<0>, then there was
1423	was some error while C<stat>ing the file.	2603	some error while C<stat>ing the file.
1424		2604
1425	=item ev_statdata prev [read-only]	2605	=item ev_statdata prev [read-only]
1426		2606
1427	The previous attributes of the file. The callback gets invoked whenever	2607	The previous attributes of the file. The callback gets invoked whenever
1428	C<prev> != C<attr>.	2608	C<prev> != C<attr>, or, more precisely, one or more of these members
		2609	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2610	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
1429		2611
1430	=item ev_tstamp interval [read-only]	2612	=item ev_tstamp interval [read-only]
1431		2613
1432	The specified interval.	2614	The specified interval.
1433		2615
1434	=item const char *path [read-only]	2616	=item const char *path [read-only]
1435		2617
1436	The filesystem path that is being watched.	2618	The file system path that is being watched.
1437		2619
1438	=back	2620	=back
1439		2621
		2622	=head3 Examples
		2623
1440	Example: Watch C</etc/passwd> for attribute changes.	2624	Example: Watch C</etc/passwd> for attribute changes.
1441		2625
1442	static void	2626	static void
1443	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)	2627	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
1444	{	2628	{
1445	/* /etc/passwd changed in some way */	2629	/* /etc/passwd changed in some way */
1446	if (w->attr.st_nlink)	2630	if (w->attr.st_nlink)
1447	{	2631	{
1448	printf ("passwd current size %ld\n", (long)w->attr.st_size);	2632	printf ("passwd current size %ld\n", (long)w->attr.st_size);
1449	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);	2633	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
1450	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);	2634	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
1451	}	2635	}
1452	else	2636	else
1453	/* you shalt not abuse printf for puts */	2637	/* you shalt not abuse printf for puts */
1454	puts ("wow, /etc/passwd is not there, expect problems. "	2638	puts ("wow, /etc/passwd is not there, expect problems. "
1455	"if this is windows, they already arrived\n");	2639	"if this is windows, they already arrived\n");
1456	}	2640	}
1457		2641
1458	...	2642	...
1459	ev_stat passwd;	2643	ev_stat passwd;
1460		2644
1461	ev_stat_init (&passwd, passwd_cb, "/etc/passwd");	2645	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
1462	ev_stat_start (loop, &passwd);	2646	ev_stat_start (loop, &passwd);
		2647
		2648	Example: Like above, but additionally use a one-second delay so we do not
		2649	miss updates (however, frequent updates will delay processing, too, so
		2650	one might do the work both on C<ev_stat> callback invocation I<and> on
		2651	C<ev_timer> callback invocation).
		2652
		2653	static ev_stat passwd;
		2654	static ev_timer timer;
		2655
		2656	static void
		2657	timer_cb (EV_P_ ev_timer *w, int revents)
		2658	{
		2659	ev_timer_stop (EV_A_ w);
		2660
		2661	/* now it's one second after the most recent passwd change */
		2662	}
		2663
		2664	static void
		2665	stat_cb (EV_P_ ev_stat *w, int revents)
		2666	{
		2667	/* reset the one-second timer */
		2668	ev_timer_again (EV_A_ &timer);
		2669	}
		2670
		2671	...
		2672	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2673	ev_stat_start (loop, &passwd);
		2674	ev_timer_init (&timer, timer_cb, 0., 1.02);
1463		2675
1464		2676
1465	=head2 C<ev_idle> - when you've got nothing better to do...	2677	=head2 C<ev_idle> - when you've got nothing better to do...
1466		2678
1467	Idle watchers trigger events when no other events of the same or higher	2679	Idle watchers trigger events when no other events of the same or higher
1468	priority are pending (prepare, check and other idle watchers do not	2680	priority are pending (prepare, check and other idle watchers do not count
1469	count).	2681	as receiving "events").
1470		2682
1471	That is, as long as your process is busy handling sockets or timeouts	2683	That is, as long as your process is busy handling sockets or timeouts
1472	(or even signals, imagine) of the same or higher priority it will not be	2684	(or even signals, imagine) of the same or higher priority it will not be
1473	triggered. But when your process is idle (or only lower-priority watchers	2685	triggered. But when your process is idle (or only lower-priority watchers
1474	are pending), the idle watchers are being called once per event loop	2686	are pending), the idle watchers are being called once per event loop
…		…
1485		2697
1486	=head3 Watcher-Specific Functions and Data Members	2698	=head3 Watcher-Specific Functions and Data Members
1487		2699
1488	=over 4	2700	=over 4
1489		2701
1490	=item ev_idle_init (ev_signal *, callback)	2702	=item ev_idle_init (ev_idle *, callback)
1491		2703
1492	Initialises and configures the idle watcher - it has no parameters of any	2704	Initialises and configures the idle watcher - it has no parameters of any
1493	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2705	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
1494	believe me.	2706	believe me.
1495		2707
1496	=back	2708	=back
1497		2709
		2710	=head3 Examples
		2711
1498	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the	2712	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
1499	callback, free it. Also, use no error checking, as usual.	2713	callback, free it. Also, use no error checking, as usual.
1500		2714
1501	static void	2715	static void
1502	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)	2716	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
1503	{	2717	{
1504	free (w);	2718	free (w);
1505	// now do something you wanted to do when the program has	2719	// now do something you wanted to do when the program has
1506	// no longer asnything immediate to do.	2720	// no longer anything immediate to do.
1507	}	2721	}
1508		2722
1509	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));	2723	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
1510	ev_idle_init (idle_watcher, idle_cb);	2724	ev_idle_init (idle_watcher, idle_cb);
1511	ev_idle_start (loop, idle_cb);	2725	ev_idle_start (loop, idle_watcher);
1512		2726
1513		2727
1514	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	2728	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
1515		2729
1516	Prepare and check watchers are usually (but not always) used in tandem:	2730	Prepare and check watchers are usually (but not always) used in pairs:
1517	prepare watchers get invoked before the process blocks and check watchers	2731	prepare watchers get invoked before the process blocks and check watchers
1518	afterwards.	2732	afterwards.
1519		2733
1520	You I<must not> call C<ev_loop> or similar functions that enter	2734	You I<must not> call C<ev_run> or similar functions that enter
1521	the current event loop from either C<ev_prepare> or C<ev_check>	2735	the current event loop from either C<ev_prepare> or C<ev_check>
1522	watchers. Other loops than the current one are fine, however. The	2736	watchers. Other loops than the current one are fine, however. The
1523	rationale behind this is that you do not need to check for recursion in	2737	rationale behind this is that you do not need to check for recursion in
1524	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,	2738	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
1525	C<ev_check> so if you have one watcher of each kind they will always be	2739	C<ev_check> so if you have one watcher of each kind they will always be
1526	called in pairs bracketing the blocking call.	2740	called in pairs bracketing the blocking call.
1527		2741
1528	Their main purpose is to integrate other event mechanisms into libev and	2742	Their main purpose is to integrate other event mechanisms into libev and
1529	their use is somewhat advanced. This could be used, for example, to track	2743	their use is somewhat advanced. They could be used, for example, to track
1530	variable changes, implement your own watchers, integrate net-snmp or a	2744	variable changes, implement your own watchers, integrate net-snmp or a
1531	coroutine library and lots more. They are also occasionally useful if	2745	coroutine library and lots more. They are also occasionally useful if
1532	you cache some data and want to flush it before blocking (for example,	2746	you cache some data and want to flush it before blocking (for example,
1533	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>	2747	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
1534	watcher).	2748	watcher).
1535		2749
1536	This is done by examining in each prepare call which file descriptors need	2750	This is done by examining in each prepare call which file descriptors
1537	to be watched by the other library, registering C<ev_io> watchers for	2751	need to be watched by the other library, registering C<ev_io> watchers
1538	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2752	for them and starting an C<ev_timer> watcher for any timeouts (many
1539	provide just this functionality). Then, in the check watcher you check for	2753	libraries provide exactly this functionality). Then, in the check watcher,
1540	any events that occured (by checking the pending status of all watchers	2754	you check for any events that occurred (by checking the pending status
1541	and stopping them) and call back into the library. The I/O and timer	2755	of all watchers and stopping them) and call back into the library. The
1542	callbacks will never actually be called (but must be valid nevertheless,	2756	I/O and timer callbacks will never actually be called (but must be valid
1543	because you never know, you know?).	2757	nevertheless, because you never know, you know?).
1544		2758
1545	As another example, the Perl Coro module uses these hooks to integrate	2759	As another example, the Perl Coro module uses these hooks to integrate
1546	coroutines into libev programs, by yielding to other active coroutines	2760	coroutines into libev programs, by yielding to other active coroutines
1547	during each prepare and only letting the process block if no coroutines	2761	during each prepare and only letting the process block if no coroutines
1548	are ready to run (it's actually more complicated: it only runs coroutines	2762	are ready to run (it's actually more complicated: it only runs coroutines
…		…
1551	loop from blocking if lower-priority coroutines are active, thus mapping	2765	loop from blocking if lower-priority coroutines are active, thus mapping
1552	low-priority coroutines to idle/background tasks).	2766	low-priority coroutines to idle/background tasks).
1553		2767
1554	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)	2768	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
1555	priority, to ensure that they are being run before any other watchers	2769	priority, to ensure that they are being run before any other watchers
		2770	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2771
1556	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,	2772	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
1557	too) should not activate ("feed") events into libev. While libev fully	2773	activate ("feed") events into libev. While libev fully supports this, they
1558	supports this, they will be called before other C<ev_check> watchers did	2774	might get executed before other C<ev_check> watchers did their job. As
1559	their job. As C<ev_check> watchers are often used to embed other event	2775	C<ev_check> watchers are often used to embed other (non-libev) event
1560	loops those other event loops might be in an unusable state until their	2776	loops those other event loops might be in an unusable state until their
1561	C<ev_check> watcher ran (always remind yourself to coexist peacefully with	2777	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
1562	others).	2778	others).
1563		2779
1564	=head3 Watcher-Specific Functions and Data Members	2780	=head3 Watcher-Specific Functions and Data Members
…		…
1569		2785
1570	=item ev_check_init (ev_check *, callback)	2786	=item ev_check_init (ev_check *, callback)
1571		2787
1572	Initialises and configures the prepare or check watcher - they have no	2788	Initialises and configures the prepare or check watcher - they have no
1573	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2789	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
1574	macros, but using them is utterly, utterly and completely pointless.	2790	macros, but using them is utterly, utterly, utterly and completely
		2791	pointless.
1575		2792
1576	=back	2793	=back
		2794
		2795	=head3 Examples
1577		2796
1578	There are a number of principal ways to embed other event loops or modules	2797	There are a number of principal ways to embed other event loops or modules
1579	into libev. Here are some ideas on how to include libadns into libev	2798	into libev. Here are some ideas on how to include libadns into libev
1580	(there is a Perl module named C<EV::ADNS> that does this, which you could	2799	(there is a Perl module named C<EV::ADNS> that does this, which you could
1581	use for an actually working example. Another Perl module named C<EV::Glib>	2800	use as a working example. Another Perl module named C<EV::Glib> embeds a
1582	embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV	2801	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
1583	into the Glib event loop).	2802	Glib event loop).
1584		2803
1585	Method 1: Add IO watchers and a timeout watcher in a prepare handler,	2804	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
1586	and in a check watcher, destroy them and call into libadns. What follows	2805	and in a check watcher, destroy them and call into libadns. What follows
1587	is pseudo-code only of course. This requires you to either use a low	2806	is pseudo-code only of course. This requires you to either use a low
1588	priority for the check watcher or use C<ev_clear_pending> explicitly, as	2807	priority for the check watcher or use C<ev_clear_pending> explicitly, as
1589	the callbacks for the IO/timeout watchers might not have been called yet.	2808	the callbacks for the IO/timeout watchers might not have been called yet.
1590		2809
1591	static ev_io iow [nfd];	2810	static ev_io iow [nfd];
1592	static ev_timer tw;	2811	static ev_timer tw;
1593		2812
1594	static void	2813	static void
1595	io_cb (ev_loop *loop, ev_io *w, int revents)	2814	io_cb (struct ev_loop *loop, ev_io *w, int revents)
1596	{	2815	{
1597	}	2816	}
1598		2817
1599	// create io watchers for each fd and a timer before blocking	2818	// create io watchers for each fd and a timer before blocking
1600	static void	2819	static void
1601	adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)	2820	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
1602	{	2821	{
1603	int timeout = 3600000;	2822	int timeout = 3600000;
1604	struct pollfd fds [nfd];	2823	struct pollfd fds [nfd];
1605	// actual code will need to loop here and realloc etc.	2824	// actual code will need to loop here and realloc etc.
1606	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));	2825	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1607		2826
1608	/* the callback is illegal, but won't be called as we stop during check */	2827	/* the callback is illegal, but won't be called as we stop during check */
1609	ev_timer_init (&tw, 0, timeout * 1e-3);	2828	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
1610	ev_timer_start (loop, &tw);	2829	ev_timer_start (loop, &tw);
1611		2830
1612	// create one ev_io per pollfd	2831	// create one ev_io per pollfd
1613	for (int i = 0; i < nfd; ++i)	2832	for (int i = 0; i < nfd; ++i)
1614	{	2833	{
1615	ev_io_init (iow + i, io_cb, fds [i].fd,	2834	ev_io_init (iow + i, io_cb, fds [i].fd,
1616	((fds [i].events & POLLIN ? EV_READ : 0)	2835	((fds [i].events & POLLIN ? EV_READ : 0)
1617	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));	2836	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
1618		2837
1619	fds [i].revents = 0;	2838	fds [i].revents = 0;
1620	ev_io_start (loop, iow + i);	2839	ev_io_start (loop, iow + i);
1621	}	2840	}
1622	}	2841	}
1623		2842
1624	// stop all watchers after blocking	2843	// stop all watchers after blocking
1625	static void	2844	static void
1626	adns_check_cb (ev_loop *loop, ev_check *w, int revents)	2845	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
1627	{	2846	{
1628	ev_timer_stop (loop, &tw);	2847	ev_timer_stop (loop, &tw);
1629		2848
1630	for (int i = 0; i < nfd; ++i)	2849	for (int i = 0; i < nfd; ++i)
1631	{	2850	{
1632	// set the relevant poll flags	2851	// set the relevant poll flags
1633	// could also call adns_processreadable etc. here	2852	// could also call adns_processreadable etc. here
1634	struct pollfd *fd = fds + i;	2853	struct pollfd *fd = fds + i;
1635	int revents = ev_clear_pending (iow + i);	2854	int revents = ev_clear_pending (iow + i);
1636	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;	2855	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1637	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;	2856	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1638		2857
1639	// now stop the watcher	2858	// now stop the watcher
1640	ev_io_stop (loop, iow + i);	2859	ev_io_stop (loop, iow + i);
1641	}	2860	}
1642		2861
1643	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));	2862	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
1644	}	2863	}
1645		2864
1646	Method 2: This would be just like method 1, but you run C<adns_afterpoll>	2865	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
1647	in the prepare watcher and would dispose of the check watcher.	2866	in the prepare watcher and would dispose of the check watcher.
1648		2867
1649	Method 3: If the module to be embedded supports explicit event	2868	Method 3: If the module to be embedded supports explicit event
1650	notification (adns does), you can also make use of the actual watcher	2869	notification (libadns does), you can also make use of the actual watcher
1651	callbacks, and only destroy/create the watchers in the prepare watcher.	2870	callbacks, and only destroy/create the watchers in the prepare watcher.
1652		2871
1653	static void	2872	static void
1654	timer_cb (EV_P_ ev_timer *w, int revents)	2873	timer_cb (EV_P_ ev_timer *w, int revents)
1655	{	2874	{
1656	adns_state ads = (adns_state)w->data;	2875	adns_state ads = (adns_state)w->data;
1657	update_now (EV_A);	2876	update_now (EV_A);
1658		2877
1659	adns_processtimeouts (ads, &tv_now);	2878	adns_processtimeouts (ads, &tv_now);
1660	}	2879	}
1661		2880
1662	static void	2881	static void
1663	io_cb (EV_P_ ev_io *w, int revents)	2882	io_cb (EV_P_ ev_io *w, int revents)
1664	{	2883	{
1665	adns_state ads = (adns_state)w->data;	2884	adns_state ads = (adns_state)w->data;
1666	update_now (EV_A);	2885	update_now (EV_A);
1667		2886
1668	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);	2887	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
1669	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);	2888	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
1670	}	2889	}
1671		2890
1672	// do not ever call adns_afterpoll	2891	// do not ever call adns_afterpoll
1673		2892
1674	Method 4: Do not use a prepare or check watcher because the module you	2893	Method 4: Do not use a prepare or check watcher because the module you
1675	want to embed is too inflexible to support it. Instead, youc na override	2894	want to embed is not flexible enough to support it. Instead, you can
1676	their poll function. The drawback with this solution is that the main	2895	override their poll function. The drawback with this solution is that the
1677	loop is now no longer controllable by EV. The C<Glib::EV> module does	2896	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
1678	this.	2897	this approach, effectively embedding EV as a client into the horrible
		2898	libglib event loop.
1679		2899
1680	static gint	2900	static gint
1681	event_poll_func (GPollFD *fds, guint nfds, gint timeout)	2901	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
1682	{	2902	{
1683	int got_events = 0;	2903	int got_events = 0;
1684		2904
1685	for (n = 0; n < nfds; ++n)	2905	for (n = 0; n < nfds; ++n)
1686	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events	2906	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
1687		2907
1688	if (timeout >= 0)	2908	if (timeout >= 0)
1689	// create/start timer	2909	// create/start timer
1690		2910
1691	// poll	2911	// poll
1692	ev_loop (EV_A_ 0);	2912	ev_run (EV_A_ 0);
1693		2913
1694	// stop timer again	2914	// stop timer again
1695	if (timeout >= 0)	2915	if (timeout >= 0)
1696	ev_timer_stop (EV_A_ &to);	2916	ev_timer_stop (EV_A_ &to);
1697		2917
1698	// stop io watchers again - their callbacks should have set	2918	// stop io watchers again - their callbacks should have set
1699	for (n = 0; n < nfds; ++n)	2919	for (n = 0; n < nfds; ++n)
1700	ev_io_stop (EV_A_ iow [n]);	2920	ev_io_stop (EV_A_ iow [n]);
1701		2921
1702	return got_events;	2922	return got_events;
1703	}	2923	}
1704		2924
1705		2925
1706	=head2 C<ev_embed> - when one backend isn't enough...	2926	=head2 C<ev_embed> - when one backend isn't enough...
1707		2927
1708	This is a rather advanced watcher type that lets you embed one event loop	2928	This is a rather advanced watcher type that lets you embed one event loop
…		…
1714	prioritise I/O.	2934	prioritise I/O.
1715		2935
1716	As an example for a bug workaround, the kqueue backend might only support	2936	As an example for a bug workaround, the kqueue backend might only support
1717	sockets on some platform, so it is unusable as generic backend, but you	2937	sockets on some platform, so it is unusable as generic backend, but you
1718	still want to make use of it because you have many sockets and it scales	2938	still want to make use of it because you have many sockets and it scales
1719	so nicely. In this case, you would create a kqueue-based loop and embed it	2939	so nicely. In this case, you would create a kqueue-based loop and embed
1720	into your default loop (which might use e.g. poll). Overall operation will	2940	it into your default loop (which might use e.g. poll). Overall operation
1721	be a bit slower because first libev has to poll and then call kevent, but	2941	will be a bit slower because first libev has to call C<poll> and then
1722	at least you can use both at what they are best.	2942	C<kevent>, but at least you can use both mechanisms for what they are
		2943	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
1723		2944
1724	As for prioritising I/O: rarely you have the case where some fds have	2945	As for prioritising I/O: under rare circumstances you have the case where
1725	to be watched and handled very quickly (with low latency), and even	2946	some fds have to be watched and handled very quickly (with low latency),
1726	priorities and idle watchers might have too much overhead. In this case	2947	and even priorities and idle watchers might have too much overhead. In
1727	you would put all the high priority stuff in one loop and all the rest in	2948	this case you would put all the high priority stuff in one loop and all
1728	a second one, and embed the second one in the first.	2949	the rest in a second one, and embed the second one in the first.
1729		2950
1730	As long as the watcher is active, the callback will be invoked every time	2951	As long as the watcher is active, the callback will be invoked every
1731	there might be events pending in the embedded loop. The callback must then	2952	time there might be events pending in the embedded loop. The callback
1732	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	2953	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
1733	their callbacks (you could also start an idle watcher to give the embedded	2954	sweep and invoke their callbacks (the callback doesn't need to invoke the
1734	loop strictly lower priority for example). You can also set the callback	2955	C<ev_embed_sweep> function directly, it could also start an idle watcher
1735	to C<0>, in which case the embed watcher will automatically execute the	2956	to give the embedded loop strictly lower priority for example).
1736	embedded loop sweep.
1737		2957
1738	As long as the watcher is started it will automatically handle events. The	2958	You can also set the callback to C<0>, in which case the embed watcher
1739	callback will be invoked whenever some events have been handled. You can	2959	will automatically execute the embedded loop sweep whenever necessary.
1740	set the callback to C<0> to avoid having to specify one if you are not
1741	interested in that.
1742		2960
1743	Also, there have not currently been made special provisions for forking:	2961	Fork detection will be handled transparently while the C<ev_embed> watcher
1744	when you fork, you not only have to call C<ev_loop_fork> on both loops,	2962	is active, i.e., the embedded loop will automatically be forked when the
1745	but you will also have to stop and restart any C<ev_embed> watchers	2963	embedding loop forks. In other cases, the user is responsible for calling
1746	yourself.	2964	C<ev_loop_fork> on the embedded loop.
1747		2965
1748	Unfortunately, not all backends are embeddable, only the ones returned by	2966	Unfortunately, not all backends are embeddable: only the ones returned by
1749	C<ev_embeddable_backends> are, which, unfortunately, does not include any	2967	C<ev_embeddable_backends> are, which, unfortunately, does not include any
1750	portable one.	2968	portable one.
1751		2969
1752	So when you want to use this feature you will always have to be prepared	2970	So when you want to use this feature you will always have to be prepared
1753	that you cannot get an embeddable loop. The recommended way to get around	2971	that you cannot get an embeddable loop. The recommended way to get around
1754	this is to have a separate variables for your embeddable loop, try to	2972	this is to have a separate variables for your embeddable loop, try to
1755	create it, and if that fails, use the normal loop for everything:	2973	create it, and if that fails, use the normal loop for everything.
1756		2974
1757	struct ev_loop *loop_hi = ev_default_init (0);	2975	=head3 C<ev_embed> and fork
1758	struct ev_loop *loop_lo = 0;
1759	struct ev_embed embed;
1760
1761	// see if there is a chance of getting one that works
1762	// (remember that a flags value of 0 means autodetection)
1763	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
1764	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
1765	: 0;
1766		2976
1767	// if we got one, then embed it, otherwise default to loop_hi	2977	While the C<ev_embed> watcher is running, forks in the embedding loop will
1768	if (loop_lo)	2978	automatically be applied to the embedded loop as well, so no special
1769	{	2979	fork handling is required in that case. When the watcher is not running,
1770	ev_embed_init (&embed, 0, loop_lo);	2980	however, it is still the task of the libev user to call C<ev_loop_fork ()>
1771	ev_embed_start (loop_hi, &embed);	2981	as applicable.
1772	}
1773	else
1774	loop_lo = loop_hi;
1775		2982
1776	=head3 Watcher-Specific Functions and Data Members	2983	=head3 Watcher-Specific Functions and Data Members
1777		2984
1778	=over 4	2985	=over 4
1779		2986
…		…
1783		2990
1784	Configures the watcher to embed the given loop, which must be	2991	Configures the watcher to embed the given loop, which must be
1785	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be	2992	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
1786	invoked automatically, otherwise it is the responsibility of the callback	2993	invoked automatically, otherwise it is the responsibility of the callback
1787	to invoke it (it will continue to be called until the sweep has been done,	2994	to invoke it (it will continue to be called until the sweep has been done,
1788	if you do not want thta, you need to temporarily stop the embed watcher).	2995	if you do not want that, you need to temporarily stop the embed watcher).
1789		2996
1790	=item ev_embed_sweep (loop, ev_embed *)	2997	=item ev_embed_sweep (loop, ev_embed *)
1791		2998
1792	Make a single, non-blocking sweep over the embedded loop. This works	2999	Make a single, non-blocking sweep over the embedded loop. This works
1793	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most	3000	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
1794	apropriate way for embedded loops.	3001	appropriate way for embedded loops.
1795		3002
1796	=item struct ev_loop *loop [read-only]	3003	=item struct ev_loop *other [read-only]
1797		3004
1798	The embedded event loop.	3005	The embedded event loop.
1799		3006
1800	=back	3007	=back
		3008
		3009	=head3 Examples
		3010
		3011	Example: Try to get an embeddable event loop and embed it into the default
		3012	event loop. If that is not possible, use the default loop. The default
		3013	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3014	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3015	used).
		3016
		3017	struct ev_loop *loop_hi = ev_default_init (0);
		3018	struct ev_loop *loop_lo = 0;
		3019	ev_embed embed;
		3020
		3021	// see if there is a chance of getting one that works
		3022	// (remember that a flags value of 0 means autodetection)
		3023	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3024	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3025	: 0;
		3026
		3027	// if we got one, then embed it, otherwise default to loop_hi
		3028	if (loop_lo)
		3029	{
		3030	ev_embed_init (&embed, 0, loop_lo);
		3031	ev_embed_start (loop_hi, &embed);
		3032	}
		3033	else
		3034	loop_lo = loop_hi;
		3035
		3036	Example: Check if kqueue is available but not recommended and create
		3037	a kqueue backend for use with sockets (which usually work with any
		3038	kqueue implementation). Store the kqueue/socket-only event loop in
		3039	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3040
		3041	struct ev_loop *loop = ev_default_init (0);
		3042	struct ev_loop *loop_socket = 0;
		3043	ev_embed embed;
		3044
		3045	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3046	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3047	{
		3048	ev_embed_init (&embed, 0, loop_socket);
		3049	ev_embed_start (loop, &embed);
		3050	}
		3051
		3052	if (!loop_socket)
		3053	loop_socket = loop;
		3054
		3055	// now use loop_socket for all sockets, and loop for everything else
1801		3056
1802		3057
1803	=head2 C<ev_fork> - the audacity to resume the event loop after a fork	3058	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
1804		3059
1805	Fork watchers are called when a C<fork ()> was detected (usually because	3060	Fork watchers are called when a C<fork ()> was detected (usually because
…		…
1808	event loop blocks next and before C<ev_check> watchers are being called,	3063	event loop blocks next and before C<ev_check> watchers are being called,
1809	and only in the child after the fork. If whoever good citizen calling	3064	and only in the child after the fork. If whoever good citizen calling
1810	C<ev_default_fork> cheats and calls it in the wrong process, the fork	3065	C<ev_default_fork> cheats and calls it in the wrong process, the fork
1811	handlers will be invoked, too, of course.	3066	handlers will be invoked, too, of course.
1812		3067
		3068	=head3 The special problem of life after fork - how is it possible?
		3069
		3070	Most uses of C<fork()> consist of forking, then some simple calls to set
		3071	up/change the process environment, followed by a call to C<exec()>. This
		3072	sequence should be handled by libev without any problems.
		3073
		3074	This changes when the application actually wants to do event handling
		3075	in the child, or both parent in child, in effect "continuing" after the
		3076	fork.
		3077
		3078	The default mode of operation (for libev, with application help to detect
		3079	forks) is to duplicate all the state in the child, as would be expected
		3080	when I<either> the parent I<or> the child process continues.
		3081
		3082	When both processes want to continue using libev, then this is usually the
		3083	wrong result. In that case, usually one process (typically the parent) is
		3084	supposed to continue with all watchers in place as before, while the other
		3085	process typically wants to start fresh, i.e. without any active watchers.
		3086
		3087	The cleanest and most efficient way to achieve that with libev is to
		3088	simply create a new event loop, which of course will be "empty", and
		3089	use that for new watchers. This has the advantage of not touching more
		3090	memory than necessary, and thus avoiding the copy-on-write, and the
		3091	disadvantage of having to use multiple event loops (which do not support
		3092	signal watchers).
		3093
		3094	When this is not possible, or you want to use the default loop for
		3095	other reasons, then in the process that wants to start "fresh", call
		3096	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3097	Destroying the default loop will "orphan" (not stop) all registered
		3098	watchers, so you have to be careful not to execute code that modifies
		3099	those watchers. Note also that in that case, you have to re-register any
		3100	signal watchers.
		3101
1813	=head3 Watcher-Specific Functions and Data Members	3102	=head3 Watcher-Specific Functions and Data Members
1814		3103
1815	=over 4	3104	=over 4
1816		3105
1817	=item ev_fork_init (ev_signal *, callback)	3106	=item ev_fork_init (ev_fork *, callback)
1818		3107
1819	Initialises and configures the fork watcher - it has no parameters of any	3108	Initialises and configures the fork watcher - it has no parameters of any
1820	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,	3109	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
1821	believe me.	3110	really.
1822		3111
1823	=back	3112	=back
1824		3113
1825		3114
		3115	=head2 C<ev_cleanup> - even the best things end
		3116
		3117	Cleanup watchers are called just before the event loop is being destroyed
		3118	by a call to C<ev_loop_destroy>.
		3119
		3120	While there is no guarantee that the event loop gets destroyed, cleanup
		3121	watchers provide a convenient method to install cleanup hooks for your
		3122	program, worker threads and so on - you just to make sure to destroy the
		3123	loop when you want them to be invoked.
		3124
		3125	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3126	all other watchers, they do not keep a reference to the event loop (which
		3127	makes a lot of sense if you think about it). Like all other watchers, you
		3128	can call libev functions in the callback, except C<ev_cleanup_start>.
		3129
		3130	=head3 Watcher-Specific Functions and Data Members
		3131
		3132	=over 4
		3133
		3134	=item ev_cleanup_init (ev_cleanup *, callback)
		3135
		3136	Initialises and configures the cleanup watcher - it has no parameters of
		3137	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3138	pointless, I assure you.
		3139
		3140	=back
		3141
		3142	Example: Register an atexit handler to destroy the default loop, so any
		3143	cleanup functions are called.
		3144
		3145	static void
		3146	program_exits (void)
		3147	{
		3148	ev_loop_destroy (EV_DEFAULT_UC);
		3149	}
		3150
		3151	...
		3152	atexit (program_exits);
		3153
		3154
		3155	=head2 C<ev_async> - how to wake up an event loop
		3156
		3157	In general, you cannot use an C<ev_run> from multiple threads or other
		3158	asynchronous sources such as signal handlers (as opposed to multiple event
		3159	loops - those are of course safe to use in different threads).
		3160
		3161	Sometimes, however, you need to wake up an event loop you do not control,
		3162	for example because it belongs to another thread. This is what C<ev_async>
		3163	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3164	it by calling C<ev_async_send>, which is thread- and signal safe.
		3165
		3166	This functionality is very similar to C<ev_signal> watchers, as signals,
		3167	too, are asynchronous in nature, and signals, too, will be compressed
		3168	(i.e. the number of callback invocations may be less than the number of
		3169	C<ev_async_sent> calls).
		3170
		3171	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3172	just the default loop.
		3173
		3174	=head3 Queueing
		3175
		3176	C<ev_async> does not support queueing of data in any way. The reason
		3177	is that the author does not know of a simple (or any) algorithm for a
		3178	multiple-writer-single-reader queue that works in all cases and doesn't
		3179	need elaborate support such as pthreads or unportable memory access
		3180	semantics.
		3181
		3182	That means that if you want to queue data, you have to provide your own
		3183	queue. But at least I can tell you how to implement locking around your
		3184	queue:
		3185
		3186	=over 4
		3187
		3188	=item queueing from a signal handler context
		3189
		3190	To implement race-free queueing, you simply add to the queue in the signal
		3191	handler but you block the signal handler in the watcher callback. Here is
		3192	an example that does that for some fictitious SIGUSR1 handler:
		3193
		3194	static ev_async mysig;
		3195
		3196	static void
		3197	sigusr1_handler (void)
		3198	{
		3199	sometype data;
		3200
		3201	// no locking etc.
		3202	queue_put (data);
		3203	ev_async_send (EV_DEFAULT_ &mysig);
		3204	}
		3205
		3206	static void
		3207	mysig_cb (EV_P_ ev_async *w, int revents)
		3208	{
		3209	sometype data;
		3210	sigset_t block, prev;
		3211
		3212	sigemptyset (&block);
		3213	sigaddset (&block, SIGUSR1);
		3214	sigprocmask (SIG_BLOCK, &block, &prev);
		3215
		3216	while (queue_get (&data))
		3217	process (data);
		3218
		3219	if (sigismember (&prev, SIGUSR1)
		3220	sigprocmask (SIG_UNBLOCK, &block, 0);
		3221	}
		3222
		3223	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3224	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3225	either...).
		3226
		3227	=item queueing from a thread context
		3228
		3229	The strategy for threads is different, as you cannot (easily) block
		3230	threads but you can easily preempt them, so to queue safely you need to
		3231	employ a traditional mutex lock, such as in this pthread example:
		3232
		3233	static ev_async mysig;
		3234	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3235
		3236	static void
		3237	otherthread (void)
		3238	{
		3239	// only need to lock the actual queueing operation
		3240	pthread_mutex_lock (&mymutex);
		3241	queue_put (data);
		3242	pthread_mutex_unlock (&mymutex);
		3243
		3244	ev_async_send (EV_DEFAULT_ &mysig);
		3245	}
		3246
		3247	static void
		3248	mysig_cb (EV_P_ ev_async *w, int revents)
		3249	{
		3250	pthread_mutex_lock (&mymutex);
		3251
		3252	while (queue_get (&data))
		3253	process (data);
		3254
		3255	pthread_mutex_unlock (&mymutex);
		3256	}
		3257
		3258	=back
		3259
		3260
		3261	=head3 Watcher-Specific Functions and Data Members
		3262
		3263	=over 4
		3264
		3265	=item ev_async_init (ev_async *, callback)
		3266
		3267	Initialises and configures the async watcher - it has no parameters of any
		3268	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3269	trust me.
		3270
		3271	=item ev_async_send (loop, ev_async *)
		3272
		3273	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3274	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3275	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3276	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3277	section below on what exactly this means).
		3278
		3279	Note that, as with other watchers in libev, multiple events might get
		3280	compressed into a single callback invocation (another way to look at this
		3281	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3282	reset when the event loop detects that).
		3283
		3284	This call incurs the overhead of a system call only once per event loop
		3285	iteration, so while the overhead might be noticeable, it doesn't apply to
		3286	repeated calls to C<ev_async_send> for the same event loop.
		3287
		3288	=item bool = ev_async_pending (ev_async *)
		3289
		3290	Returns a non-zero value when C<ev_async_send> has been called on the
		3291	watcher but the event has not yet been processed (or even noted) by the
		3292	event loop.
		3293
		3294	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3295	the loop iterates next and checks for the watcher to have become active,
		3296	it will reset the flag again. C<ev_async_pending> can be used to very
		3297	quickly check whether invoking the loop might be a good idea.
		3298
		3299	Not that this does I<not> check whether the watcher itself is pending,
		3300	only whether it has been requested to make this watcher pending: there
		3301	is a time window between the event loop checking and resetting the async
		3302	notification, and the callback being invoked.
		3303
		3304	=back
		3305
		3306
1826	=head1 OTHER FUNCTIONS	3307	=head1 OTHER FUNCTIONS
1827		3308
1828	There are some other functions of possible interest. Described. Here. Now.	3309	There are some other functions of possible interest. Described. Here. Now.
1829		3310
1830	=over 4	3311	=over 4
1831		3312
1832	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3313	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
1833		3314
1834	This function combines a simple timer and an I/O watcher, calls your	3315	This function combines a simple timer and an I/O watcher, calls your
1835	callback on whichever event happens first and automatically stop both	3316	callback on whichever event happens first and automatically stops both
1836	watchers. This is useful if you want to wait for a single event on an fd	3317	watchers. This is useful if you want to wait for a single event on an fd
1837	or timeout without having to allocate/configure/start/stop/free one or	3318	or timeout without having to allocate/configure/start/stop/free one or
1838	more watchers yourself.	3319	more watchers yourself.
1839		3320
1840	If C<fd> is less than 0, then no I/O watcher will be started and events	3321	If C<fd> is less than 0, then no I/O watcher will be started and the
1841	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3322	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
1842	C<events> set will be craeted and started.	3323	the given C<fd> and C<events> set will be created and started.
1843		3324
1844	If C<timeout> is less than 0, then no timeout watcher will be	3325	If C<timeout> is less than 0, then no timeout watcher will be
1845	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3326	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
1846	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3327	repeat = 0) will be started. C<0> is a valid timeout.
1847	dubious value.
1848		3328
1849	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3329	The callback has the type C<void (*cb)(int revents, void *arg)> and is
1850	passed an C<revents> set like normal event callbacks (a combination of	3330	passed an C<revents> set like normal event callbacks (a combination of
1851	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3331	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
1852	value passed to C<ev_once>:	3332	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3333	a timeout and an io event at the same time - you probably should give io
		3334	events precedence.
1853		3335
		3336	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3337
1854	static void stdin_ready (int revents, void *arg)	3338	static void stdin_ready (int revents, void *arg)
1855	{	3339	{
1856	if (revents & EV_TIMEOUT)
1857	/* doh, nothing entered */;
1858	else if (revents & EV_READ)	3340	if (revents & EV_READ)
1859	/* stdin might have data for us, joy! */;	3341	/* stdin might have data for us, joy! */;
		3342	else if (revents & EV_TIMER)
		3343	/* doh, nothing entered */;
1860	}	3344	}
1861		3345
1862	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3346	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
1863		3347
1864	=item ev_feed_event (ev_loop *, watcher *, int revents)
1865
1866	Feeds the given event set into the event loop, as if the specified event
1867	had happened for the specified watcher (which must be a pointer to an
1868	initialised but not necessarily started event watcher).
1869
1870	=item ev_feed_fd_event (ev_loop *, int fd, int revents)	3348	=item ev_feed_fd_event (loop, int fd, int revents)
1871		3349
1872	Feed an event on the given fd, as if a file descriptor backend detected	3350	Feed an event on the given fd, as if a file descriptor backend detected
1873	the given events it.	3351	the given events it.
1874		3352
1875	=item ev_feed_signal_event (ev_loop *loop, int signum)	3353	=item ev_feed_signal_event (loop, int signum)
1876		3354
1877	Feed an event as if the given signal occured (C<loop> must be the default	3355	Feed an event as if the given signal occurred (C<loop> must be the default
1878	loop!).	3356	loop!).
1879		3357
1880	=back	3358	=back
1881		3359
1882		3360
…		…
1898		3376
1899	=item * Priorities are not currently supported. Initialising priorities	3377	=item * Priorities are not currently supported. Initialising priorities
1900	will fail and all watchers will have the same priority, even though there	3378	will fail and all watchers will have the same priority, even though there
1901	is an ev_pri field.	3379	is an ev_pri field.
1902		3380
		3381	=item * In libevent, the last base created gets the signals, in libev, the
		3382	first base created (== the default loop) gets the signals.
		3383
1903	=item * Other members are not supported.	3384	=item * Other members are not supported.
1904		3385
1905	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3386	=item * The libev emulation is I<not> ABI compatible to libevent, you need
1906	to use the libev header file and library.	3387	to use the libev header file and library.
1907		3388
1908	=back	3389	=back
1909		3390
1910	=head1 C++ SUPPORT	3391	=head1 C++ SUPPORT
1911		3392
1912	Libev comes with some simplistic wrapper classes for C++ that mainly allow	3393	Libev comes with some simplistic wrapper classes for C++ that mainly allow
1913	you to use some convinience methods to start/stop watchers and also change	3394	you to use some convenience methods to start/stop watchers and also change
1914	the callback model to a model using method callbacks on objects.	3395	the callback model to a model using method callbacks on objects.
1915		3396
1916	To use it,	3397	To use it,
1917		3398
1918	#include <ev++.h>	3399	#include <ev++.h>
1919		3400
1920	This automatically includes F<ev.h> and puts all of its definitions (many	3401	This automatically includes F<ev.h> and puts all of its definitions (many
1921	of them macros) into the global namespace. All C++ specific things are	3402	of them macros) into the global namespace. All C++ specific things are
1922	put into the C<ev> namespace. It should support all the same embedding	3403	put into the C<ev> namespace. It should support all the same embedding
1923	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.	3404	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
…		…
1957		3438
1958	=over 4	3439	=over 4
1959		3440
1960	=item ev::TYPE::TYPE ()	3441	=item ev::TYPE::TYPE ()
1961		3442
1962	=item ev::TYPE::TYPE (struct ev_loop *)	3443	=item ev::TYPE::TYPE (loop)
1963		3444
1964	=item ev::TYPE::~TYPE	3445	=item ev::TYPE::~TYPE
1965		3446
1966	The constructor (optionally) takes an event loop to associate the watcher	3447	The constructor (optionally) takes an event loop to associate the watcher
1967	with. If it is omitted, it will use C<EV_DEFAULT>.	3448	with. If it is omitted, it will use C<EV_DEFAULT>.
…		…
1990	your compiler is good :), then the method will be fully inlined into the	3471	your compiler is good :), then the method will be fully inlined into the
1991	thunking function, making it as fast as a direct C callback.	3472	thunking function, making it as fast as a direct C callback.
1992		3473
1993	Example: simple class declaration and watcher initialisation	3474	Example: simple class declaration and watcher initialisation
1994		3475
1995	struct myclass	3476	struct myclass
1996	{	3477	{
1997	void io_cb (ev::io &w, int revents) { }	3478	void io_cb (ev::io &w, int revents) { }
1998	}	3479	}
1999		3480
2000	myclass obj;	3481	myclass obj;
2001	ev::io iow;	3482	ev::io iow;
2002	iow.set <myclass, &myclass::io_cb> (&obj);	3483	iow.set <myclass, &myclass::io_cb> (&obj);
		3484
		3485	=item w->set (object *)
		3486
		3487	This is a variation of a method callback - leaving out the method to call
		3488	will default the method to C<operator ()>, which makes it possible to use
		3489	functor objects without having to manually specify the C<operator ()> all
		3490	the time. Incidentally, you can then also leave out the template argument
		3491	list.
		3492
		3493	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3494	int revents)>.
		3495
		3496	See the method-C<set> above for more details.
		3497
		3498	Example: use a functor object as callback.
		3499
		3500	struct myfunctor
		3501	{
		3502	void operator() (ev::io &w, int revents)
		3503	{
		3504	...
		3505	}
		3506	}
		3507
		3508	myfunctor f;
		3509
		3510	ev::io w;
		3511	w.set (&f);
2003		3512
2004	=item w->set<function> (void *data = 0)	3513	=item w->set<function> (void *data = 0)
2005		3514
2006	Also sets a callback, but uses a static method or plain function as	3515	Also sets a callback, but uses a static method or plain function as
2007	callback. The optional C<data> argument will be stored in the watcher's	3516	callback. The optional C<data> argument will be stored in the watcher's
…		…
2009		3518
2010	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.	3519	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
2011		3520
2012	See the method-C<set> above for more details.	3521	See the method-C<set> above for more details.
2013		3522
2014	Example:	3523	Example: Use a plain function as callback.
2015		3524
2016	static void io_cb (ev::io &w, int revents) { }	3525	static void io_cb (ev::io &w, int revents) { }
2017	iow.set <io_cb> ();	3526	iow.set <io_cb> ();
2018		3527
2019	=item w->set (struct ev_loop *)	3528	=item w->set (loop)
2020		3529
2021	Associates a different C<struct ev_loop> with this watcher. You can only	3530	Associates a different C<struct ev_loop> with this watcher. You can only
2022	do this when the watcher is inactive (and not pending either).	3531	do this when the watcher is inactive (and not pending either).
2023		3532
2024	=item w->set ([args])	3533	=item w->set ([arguments])
2025		3534
2026	Basically the same as C<ev_TYPE_set>, with the same args. Must be	3535	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
2027	called at least once. Unlike the C counterpart, an active watcher gets	3536	method or a suitable start method must be called at least once. Unlike the
2028	automatically stopped and restarted when reconfiguring it with this	3537	C counterpart, an active watcher gets automatically stopped and restarted
2029	method.	3538	when reconfiguring it with this method.
2030		3539
2031	=item w->start ()	3540	=item w->start ()
2032		3541
2033	Starts the watcher. Note that there is no C<loop> argument, as the	3542	Starts the watcher. Note that there is no C<loop> argument, as the
2034	constructor already stores the event loop.	3543	constructor already stores the event loop.
2035		3544
		3545	=item w->start ([arguments])
		3546
		3547	Instead of calling C<set> and C<start> methods separately, it is often
		3548	convenient to wrap them in one call. Uses the same type of arguments as
		3549	the configure C<set> method of the watcher.
		3550
2036	=item w->stop ()	3551	=item w->stop ()
2037		3552
2038	Stops the watcher if it is active. Again, no C<loop> argument.	3553	Stops the watcher if it is active. Again, no C<loop> argument.
2039		3554
2040	=item w->again () (C<ev::timer>, C<ev::periodic> only)	3555	=item w->again () (C<ev::timer>, C<ev::periodic> only)
…		…
2052		3567
2053	=back	3568	=back
2054		3569
2055	=back	3570	=back
2056		3571
2057	Example: Define a class with an IO and idle watcher, start one of them in	3572	Example: Define a class with two I/O and idle watchers, start the I/O
2058	the constructor.	3573	watchers in the constructor.
2059		3574
2060	class myclass	3575	class myclass
2061	{	3576	{
2062	ev_io io; void io_cb (ev::io &w, int revents);	3577	ev::io io ; void io_cb (ev::io &w, int revents);
		3578	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
2063	ev_idle idle void idle_cb (ev::idle &w, int revents);	3579	ev::idle idle; void idle_cb (ev::idle &w, int revents);
2064		3580
2065	myclass ();	3581	myclass (int fd)
2066	}	3582	{
2067
2068	myclass::myclass (int fd)
2069	{
2070	io .set <myclass, &myclass::io_cb > (this);	3583	io .set <myclass, &myclass::io_cb > (this);
		3584	io2 .set <myclass, &myclass::io2_cb > (this);
2071	idle.set <myclass, &myclass::idle_cb> (this);	3585	idle.set <myclass, &myclass::idle_cb> (this);
2072		3586
2073	io.start (fd, ev::READ);	3587	io.set (fd, ev::WRITE); // configure the watcher
		3588	io.start (); // start it whenever convenient
		3589
		3590	io2.start (fd, ev::READ); // set + start in one call
		3591	}
2074	}	3592	};
		3593
		3594
		3595	=head1 OTHER LANGUAGE BINDINGS
		3596
		3597	Libev does not offer other language bindings itself, but bindings for a
		3598	number of languages exist in the form of third-party packages. If you know
		3599	any interesting language binding in addition to the ones listed here, drop
		3600	me a note.
		3601
		3602	=over 4
		3603
		3604	=item Perl
		3605
		3606	The EV module implements the full libev API and is actually used to test
		3607	libev. EV is developed together with libev. Apart from the EV core module,
		3608	there are additional modules that implement libev-compatible interfaces
		3609	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3610	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3611	and C<EV::Glib>).
		3612
		3613	It can be found and installed via CPAN, its homepage is at
		3614	L<http://software.schmorp.de/pkg/EV>.
		3615
		3616	=item Python
		3617
		3618	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3619	seems to be quite complete and well-documented.
		3620
		3621	=item Ruby
		3622
		3623	Tony Arcieri has written a ruby extension that offers access to a subset
		3624	of the libev API and adds file handle abstractions, asynchronous DNS and
		3625	more on top of it. It can be found via gem servers. Its homepage is at
		3626	L<http://rev.rubyforge.org/>.
		3627
		3628	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3629	makes rev work even on mingw.
		3630
		3631	=item Haskell
		3632
		3633	A haskell binding to libev is available at
		3634	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3635
		3636	=item D
		3637
		3638	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3639	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3640
		3641	=item Ocaml
		3642
		3643	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3644	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3645
		3646	=item Lua
		3647
		3648	Brian Maher has written a partial interface to libev for lua (at the
		3649	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3650	L<http://github.com/brimworks/lua-ev>.
		3651
		3652	=back
2075		3653
2076		3654
2077	=head1 MACRO MAGIC	3655	=head1 MACRO MAGIC
2078		3656
2079	Libev can be compiled with a variety of options, the most fundamantal	3657	Libev can be compiled with a variety of options, the most fundamental
2080	of which is C<EV_MULTIPLICITY>. This option determines whether (most)	3658	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
2081	functions and callbacks have an initial C<struct ev_loop *> argument.	3659	functions and callbacks have an initial C<struct ev_loop *> argument.
2082		3660
2083	To make it easier to write programs that cope with either variant, the	3661	To make it easier to write programs that cope with either variant, the
2084	following macros are defined:	3662	following macros are defined:
…		…
2089		3667
2090	This provides the loop I<argument> for functions, if one is required ("ev	3668	This provides the loop I<argument> for functions, if one is required ("ev
2091	loop argument"). The C<EV_A> form is used when this is the sole argument,	3669	loop argument"). The C<EV_A> form is used when this is the sole argument,
2092	C<EV_A_> is used when other arguments are following. Example:	3670	C<EV_A_> is used when other arguments are following. Example:
2093		3671
2094	ev_unref (EV_A);	3672	ev_unref (EV_A);
2095	ev_timer_add (EV_A_ watcher);	3673	ev_timer_add (EV_A_ watcher);
2096	ev_loop (EV_A_ 0);	3674	ev_run (EV_A_ 0);
2097		3675
2098	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,	3676	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
2099	which is often provided by the following macro.	3677	which is often provided by the following macro.
2100		3678
2101	=item C<EV_P>, C<EV_P_>	3679	=item C<EV_P>, C<EV_P_>
2102		3680
2103	This provides the loop I<parameter> for functions, if one is required ("ev	3681	This provides the loop I<parameter> for functions, if one is required ("ev
2104	loop parameter"). The C<EV_P> form is used when this is the sole parameter,	3682	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
2105	C<EV_P_> is used when other parameters are following. Example:	3683	C<EV_P_> is used when other parameters are following. Example:
2106		3684
2107	// this is how ev_unref is being declared	3685	// this is how ev_unref is being declared
2108	static void ev_unref (EV_P);	3686	static void ev_unref (EV_P);
2109		3687
2110	// this is how you can declare your typical callback	3688	// this is how you can declare your typical callback
2111	static void cb (EV_P_ ev_timer *w, int revents)	3689	static void cb (EV_P_ ev_timer *w, int revents)
2112		3690
2113	It declares a parameter C<loop> of type C<struct ev_loop *>, quite	3691	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
2114	suitable for use with C<EV_A>.	3692	suitable for use with C<EV_A>.
2115		3693
2116	=item C<EV_DEFAULT>, C<EV_DEFAULT_>	3694	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
2117		3695
2118	Similar to the other two macros, this gives you the value of the default	3696	Similar to the other two macros, this gives you the value of the default
2119	loop, if multiple loops are supported ("ev loop default").	3697	loop, if multiple loops are supported ("ev loop default").
		3698
		3699	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3700
		3701	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3702	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3703	is undefined when the default loop has not been initialised by a previous
		3704	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3705
		3706	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3707	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
2120		3708
2121	=back	3709	=back
2122		3710
2123	Example: Declare and initialise a check watcher, utilising the above	3711	Example: Declare and initialise a check watcher, utilising the above
2124	macros so it will work regardless of whether multiple loops are supported	3712	macros so it will work regardless of whether multiple loops are supported
2125	or not.	3713	or not.
2126		3714
2127	static void	3715	static void
2128	check_cb (EV_P_ ev_timer *w, int revents)	3716	check_cb (EV_P_ ev_timer *w, int revents)
2129	{	3717	{
2130	ev_check_stop (EV_A_ w);	3718	ev_check_stop (EV_A_ w);
2131	}	3719	}
2132		3720
2133	ev_check check;	3721	ev_check check;
2134	ev_check_init (&check, check_cb);	3722	ev_check_init (&check, check_cb);
2135	ev_check_start (EV_DEFAULT_ &check);	3723	ev_check_start (EV_DEFAULT_ &check);
2136	ev_loop (EV_DEFAULT_ 0);	3724	ev_run (EV_DEFAULT_ 0);
2137		3725
2138	=head1 EMBEDDING	3726	=head1 EMBEDDING
2139		3727
2140	Libev can (and often is) directly embedded into host	3728	Libev can (and often is) directly embedded into host
2141	applications. Examples of applications that embed it include the Deliantra	3729	applications. Examples of applications that embed it include the Deliantra
2142	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)	3730	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
2143	and rxvt-unicode.	3731	and rxvt-unicode.
2144		3732
2145	The goal is to enable you to just copy the neecssary files into your	3733	The goal is to enable you to just copy the necessary files into your
2146	source directory without having to change even a single line in them, so	3734	source directory without having to change even a single line in them, so
2147	you can easily upgrade by simply copying (or having a checked-out copy of	3735	you can easily upgrade by simply copying (or having a checked-out copy of
2148	libev somewhere in your source tree).	3736	libev somewhere in your source tree).
2149		3737
2150	=head2 FILESETS	3738	=head2 FILESETS
2151		3739
2152	Depending on what features you need you need to include one or more sets of files	3740	Depending on what features you need you need to include one or more sets of files
2153	in your app.	3741	in your application.
2154		3742
2155	=head3 CORE EVENT LOOP	3743	=head3 CORE EVENT LOOP
2156		3744
2157	To include only the libev core (all the C<ev_*> functions), with manual	3745	To include only the libev core (all the C<ev_*> functions), with manual
2158	configuration (no autoconf):	3746	configuration (no autoconf):
2159		3747
2160	#define EV_STANDALONE 1	3748	#define EV_STANDALONE 1
2161	#include "ev.c"	3749	#include "ev.c"
2162		3750
2163	This will automatically include F<ev.h>, too, and should be done in a	3751	This will automatically include F<ev.h>, too, and should be done in a
2164	single C source file only to provide the function implementations. To use	3752	single C source file only to provide the function implementations. To use
2165	it, do the same for F<ev.h> in all files wishing to use this API (best	3753	it, do the same for F<ev.h> in all files wishing to use this API (best
2166	done by writing a wrapper around F<ev.h> that you can include instead and	3754	done by writing a wrapper around F<ev.h> that you can include instead and
2167	where you can put other configuration options):	3755	where you can put other configuration options):
2168		3756
2169	#define EV_STANDALONE 1	3757	#define EV_STANDALONE 1
2170	#include "ev.h"	3758	#include "ev.h"
2171		3759
2172	Both header files and implementation files can be compiled with a C++	3760	Both header files and implementation files can be compiled with a C++
2173	compiler (at least, thats a stated goal, and breakage will be treated	3761	compiler (at least, that's a stated goal, and breakage will be treated
2174	as a bug).	3762	as a bug).
2175		3763
2176	You need the following files in your source tree, or in a directory	3764	You need the following files in your source tree, or in a directory
2177	in your include path (e.g. in libev/ when using -Ilibev):	3765	in your include path (e.g. in libev/ when using -Ilibev):
2178		3766
2179	ev.h	3767	ev.h
2180	ev.c	3768	ev.c
2181	ev_vars.h	3769	ev_vars.h
2182	ev_wrap.h	3770	ev_wrap.h
2183		3771
2184	ev_win32.c required on win32 platforms only	3772	ev_win32.c required on win32 platforms only
2185		3773
2186	ev_select.c only when select backend is enabled (which is enabled by default)	3774	ev_select.c only when select backend is enabled (which is enabled by default)
2187	ev_poll.c only when poll backend is enabled (disabled by default)	3775	ev_poll.c only when poll backend is enabled (disabled by default)
2188	ev_epoll.c only when the epoll backend is enabled (disabled by default)	3776	ev_epoll.c only when the epoll backend is enabled (disabled by default)
2189	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)	3777	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
2190	ev_port.c only when the solaris port backend is enabled (disabled by default)	3778	ev_port.c only when the solaris port backend is enabled (disabled by default)
2191		3779
2192	F<ev.c> includes the backend files directly when enabled, so you only need	3780	F<ev.c> includes the backend files directly when enabled, so you only need
2193	to compile this single file.	3781	to compile this single file.
2194		3782
2195	=head3 LIBEVENT COMPATIBILITY API	3783	=head3 LIBEVENT COMPATIBILITY API
2196		3784
2197	To include the libevent compatibility API, also include:	3785	To include the libevent compatibility API, also include:
2198		3786
2199	#include "event.c"	3787	#include "event.c"
2200		3788
2201	in the file including F<ev.c>, and:	3789	in the file including F<ev.c>, and:
2202		3790
2203	#include "event.h"	3791	#include "event.h"
2204		3792
2205	in the files that want to use the libevent API. This also includes F<ev.h>.	3793	in the files that want to use the libevent API. This also includes F<ev.h>.
2206		3794
2207	You need the following additional files for this:	3795	You need the following additional files for this:
2208		3796
2209	event.h	3797	event.h
2210	event.c	3798	event.c
2211		3799
2212	=head3 AUTOCONF SUPPORT	3800	=head3 AUTOCONF SUPPORT
2213		3801
2214	Instead of using C<EV_STANDALONE=1> and providing your config in	3802	Instead of using C<EV_STANDALONE=1> and providing your configuration in
2215	whatever way you want, you can also C<m4_include([libev.m4])> in your	3803	whatever way you want, you can also C<m4_include([libev.m4])> in your
2216	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then	3804	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
2217	include F<config.h> and configure itself accordingly.	3805	include F<config.h> and configure itself accordingly.
2218		3806
2219	For this of course you need the m4 file:	3807	For this of course you need the m4 file:
2220		3808
2221	libev.m4	3809	libev.m4
2222		3810
2223	=head2 PREPROCESSOR SYMBOLS/MACROS	3811	=head2 PREPROCESSOR SYMBOLS/MACROS
2224		3812
2225	Libev can be configured via a variety of preprocessor symbols you have to define	3813	Libev can be configured via a variety of preprocessor symbols you have to
2226	before including any of its files. The default is not to build for multiplicity	3814	define before including (or compiling) any of its files. The default in
2227	and only include the select backend.	3815	the absence of autoconf is documented for every option.
		3816
		3817	Symbols marked with "(h)" do not change the ABI, and can have different
		3818	values when compiling libev vs. including F<ev.h>, so it is permissible
		3819	to redefine them before including F<ev.h> without breaking compatibility
		3820	to a compiled library. All other symbols change the ABI, which means all
		3821	users of libev and the libev code itself must be compiled with compatible
		3822	settings.
2228		3823
2229	=over 4	3824	=over 4
2230		3825
		3826	=item EV_COMPAT3 (h)
		3827
		3828	Backwards compatibility is a major concern for libev. This is why this
		3829	release of libev comes with wrappers for the functions and symbols that
		3830	have been renamed between libev version 3 and 4.
		3831
		3832	You can disable these wrappers (to test compatibility with future
		3833	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		3834	sources. This has the additional advantage that you can drop the C<struct>
		3835	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		3836	typedef in that case.
		3837
		3838	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		3839	and in some even more future version the compatibility code will be
		3840	removed completely.
		3841
2231	=item EV_STANDALONE	3842	=item EV_STANDALONE (h)
2232		3843
2233	Must always be C<1> if you do not use autoconf configuration, which	3844	Must always be C<1> if you do not use autoconf configuration, which
2234	keeps libev from including F<config.h>, and it also defines dummy	3845	keeps libev from including F<config.h>, and it also defines dummy
2235	implementations for some libevent functions (such as logging, which is not	3846	implementations for some libevent functions (such as logging, which is not
2236	supported). It will also not define any of the structs usually found in	3847	supported). It will also not define any of the structs usually found in
2237	F<event.h> that are not directly supported by the libev core alone.	3848	F<event.h> that are not directly supported by the libev core alone.
2238		3849
		3850	In standalone mode, libev will still try to automatically deduce the
		3851	configuration, but has to be more conservative.
		3852
2239	=item EV_USE_MONOTONIC	3853	=item EV_USE_MONOTONIC
2240		3854
2241	If defined to be C<1>, libev will try to detect the availability of the	3855	If defined to be C<1>, libev will try to detect the availability of the
2242	monotonic clock option at both compiletime and runtime. Otherwise no use	3856	monotonic clock option at both compile time and runtime. Otherwise no
2243	of the monotonic clock option will be attempted. If you enable this, you	3857	use of the monotonic clock option will be attempted. If you enable this,
2244	usually have to link against librt or something similar. Enabling it when	3858	you usually have to link against librt or something similar. Enabling it
2245	the functionality isn't available is safe, though, althoguh you have	3859	when the functionality isn't available is safe, though, although you have
2246	to make sure you link against any libraries where the C<clock_gettime>	3860	to make sure you link against any libraries where the C<clock_gettime>
2247	function is hiding in (often F<-lrt>).	3861	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
2248		3862
2249	=item EV_USE_REALTIME	3863	=item EV_USE_REALTIME
2250		3864
2251	If defined to be C<1>, libev will try to detect the availability of the	3865	If defined to be C<1>, libev will try to detect the availability of the
2252	realtime clock option at compiletime (and assume its availability at	3866	real-time clock option at compile time (and assume its availability
2253	runtime if successful). Otherwise no use of the realtime clock option will	3867	at runtime if successful). Otherwise no use of the real-time clock
2254	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	3868	option will be attempted. This effectively replaces C<gettimeofday>
2255	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the	3869	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
2256	note about libraries in the description of C<EV_USE_MONOTONIC>, though.	3870	correctness. See the note about libraries in the description of
		3871	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3872	C<EV_USE_CLOCK_SYSCALL>.
		3873
		3874	=item EV_USE_CLOCK_SYSCALL
		3875
		3876	If defined to be C<1>, libev will try to use a direct syscall instead
		3877	of calling the system-provided C<clock_gettime> function. This option
		3878	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3879	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3880	programs needlessly. Using a direct syscall is slightly slower (in
		3881	theory), because no optimised vdso implementation can be used, but avoids
		3882	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3883	higher, as it simplifies linking (no need for C<-lrt>).
		3884
		3885	=item EV_USE_NANOSLEEP
		3886
		3887	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3888	and will use it for delays. Otherwise it will use C<select ()>.
		3889
		3890	=item EV_USE_EVENTFD
		3891
		3892	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3893	available and will probe for kernel support at runtime. This will improve
		3894	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3895	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3896	2.7 or newer, otherwise disabled.
2257		3897
2258	=item EV_USE_SELECT	3898	=item EV_USE_SELECT
2259		3899
2260	If undefined or defined to be C<1>, libev will compile in support for the	3900	If undefined or defined to be C<1>, libev will compile in support for the
2261	C<select>(2) backend. No attempt at autodetection will be done: if no	3901	C<select>(2) backend. No attempt at auto-detection will be done: if no
2262	other method takes over, select will be it. Otherwise the select backend	3902	other method takes over, select will be it. Otherwise the select backend
2263	will not be compiled in.	3903	will not be compiled in.
2264		3904
2265	=item EV_SELECT_USE_FD_SET	3905	=item EV_SELECT_USE_FD_SET
2266		3906
2267	If defined to C<1>, then the select backend will use the system C<fd_set>	3907	If defined to C<1>, then the select backend will use the system C<fd_set>
2268	structure. This is useful if libev doesn't compile due to a missing	3908	structure. This is useful if libev doesn't compile due to a missing
2269	C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on	3909	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
2270	exotic systems. This usually limits the range of file descriptors to some	3910	on exotic systems. This usually limits the range of file descriptors to
2271	low limit such as 1024 or might have other limitations (winsocket only	3911	some low limit such as 1024 or might have other limitations (winsocket
2272	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might	3912	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
2273	influence the size of the C<fd_set> used.	3913	configures the maximum size of the C<fd_set>.
2274		3914
2275	=item EV_SELECT_IS_WINSOCKET	3915	=item EV_SELECT_IS_WINSOCKET
2276		3916
2277	When defined to C<1>, the select backend will assume that	3917	When defined to C<1>, the select backend will assume that
2278	select/socket/connect etc. don't understand file descriptors but	3918	select/socket/connect etc. don't understand file descriptors but
…		…
2280	be used is the winsock select). This means that it will call	3920	be used is the winsock select). This means that it will call
2281	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,	3921	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
2282	it is assumed that all these functions actually work on fds, even	3922	it is assumed that all these functions actually work on fds, even
2283	on win32. Should not be defined on non-win32 platforms.	3923	on win32. Should not be defined on non-win32 platforms.
2284		3924
		3925	=item EV_FD_TO_WIN32_HANDLE(fd)
		3926
		3927	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3928	file descriptors to socket handles. When not defining this symbol (the
		3929	default), then libev will call C<_get_osfhandle>, which is usually
		3930	correct. In some cases, programs use their own file descriptor management,
		3931	in which case they can provide this function to map fds to socket handles.
		3932
		3933	=item EV_WIN32_HANDLE_TO_FD(handle)
		3934
		3935	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3936	using the standard C<_open_osfhandle> function. For programs implementing
		3937	their own fd to handle mapping, overwriting this function makes it easier
		3938	to do so. This can be done by defining this macro to an appropriate value.
		3939
		3940	=item EV_WIN32_CLOSE_FD(fd)
		3941
		3942	If programs implement their own fd to handle mapping on win32, then this
		3943	macro can be used to override the C<close> function, useful to unregister
		3944	file descriptors again. Note that the replacement function has to close
		3945	the underlying OS handle.
		3946
2285	=item EV_USE_POLL	3947	=item EV_USE_POLL
2286		3948
2287	If defined to be C<1>, libev will compile in support for the C<poll>(2)	3949	If defined to be C<1>, libev will compile in support for the C<poll>(2)
2288	backend. Otherwise it will be enabled on non-win32 platforms. It	3950	backend. Otherwise it will be enabled on non-win32 platforms. It
2289	takes precedence over select.	3951	takes precedence over select.
2290		3952
2291	=item EV_USE_EPOLL	3953	=item EV_USE_EPOLL
2292		3954
2293	If defined to be C<1>, libev will compile in support for the Linux	3955	If defined to be C<1>, libev will compile in support for the Linux
2294	C<epoll>(7) backend. Its availability will be detected at runtime,	3956	C<epoll>(7) backend. Its availability will be detected at runtime,
2295	otherwise another method will be used as fallback. This is the	3957	otherwise another method will be used as fallback. This is the preferred
2296	preferred backend for GNU/Linux systems.	3958	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3959	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2297		3960
2298	=item EV_USE_KQUEUE	3961	=item EV_USE_KQUEUE
2299		3962
2300	If defined to be C<1>, libev will compile in support for the BSD style	3963	If defined to be C<1>, libev will compile in support for the BSD style
2301	C<kqueue>(2) backend. Its actual availability will be detected at runtime,	3964	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
…		…
2314	otherwise another method will be used as fallback. This is the preferred	3977	otherwise another method will be used as fallback. This is the preferred
2315	backend for Solaris 10 systems.	3978	backend for Solaris 10 systems.
2316		3979
2317	=item EV_USE_DEVPOLL	3980	=item EV_USE_DEVPOLL
2318		3981
2319	reserved for future expansion, works like the USE symbols above.	3982	Reserved for future expansion, works like the USE symbols above.
2320		3983
2321	=item EV_USE_INOTIFY	3984	=item EV_USE_INOTIFY
2322		3985
2323	If defined to be C<1>, libev will compile in support for the Linux inotify	3986	If defined to be C<1>, libev will compile in support for the Linux inotify
2324	interface to speed up C<ev_stat> watchers. Its actual availability will	3987	interface to speed up C<ev_stat> watchers. Its actual availability will
2325	be detected at runtime.	3988	be detected at runtime. If undefined, it will be enabled if the headers
		3989	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
2326		3990
		3991	=item EV_ATOMIC_T
		3992
		3993	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3994	access is atomic with respect to other threads or signal contexts. No such
		3995	type is easily found in the C language, so you can provide your own type
		3996	that you know is safe for your purposes. It is used both for signal handler "locking"
		3997	as well as for signal and thread safety in C<ev_async> watchers.
		3998
		3999	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4000	(from F<signal.h>), which is usually good enough on most platforms.
		4001
2327	=item EV_H	4002	=item EV_H (h)
2328		4003
2329	The name of the F<ev.h> header file used to include it. The default if	4004	The name of the F<ev.h> header file used to include it. The default if
2330	undefined is C<< <ev.h> >> in F<event.h> and C<"ev.h"> in F<ev.c>. This	4005	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2331	can be used to virtually rename the F<ev.h> header file in case of conflicts.	4006	used to virtually rename the F<ev.h> header file in case of conflicts.
2332		4007
2333	=item EV_CONFIG_H	4008	=item EV_CONFIG_H (h)
2334		4009
2335	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override	4010	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
2336	F<ev.c>'s idea of where to find the F<config.h> file, similarly to	4011	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
2337	C<EV_H>, above.	4012	C<EV_H>, above.
2338		4013
2339	=item EV_EVENT_H	4014	=item EV_EVENT_H (h)
2340		4015
2341	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea	4016	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
2342	of how the F<event.h> header can be found.	4017	of how the F<event.h> header can be found, the default is C<"event.h">.
2343		4018
2344	=item EV_PROTOTYPES	4019	=item EV_PROTOTYPES (h)
2345		4020
2346	If defined to be C<0>, then F<ev.h> will not define any function	4021	If defined to be C<0>, then F<ev.h> will not define any function
2347	prototypes, but still define all the structs and other symbols. This is	4022	prototypes, but still define all the structs and other symbols. This is
2348	occasionally useful if you want to provide your own wrapper functions	4023	occasionally useful if you want to provide your own wrapper functions
2349	around libev functions.	4024	around libev functions.
…		…
2368	When doing priority-based operations, libev usually has to linearly search	4043	When doing priority-based operations, libev usually has to linearly search
2369	all the priorities, so having many of them (hundreds) uses a lot of space	4044	all the priorities, so having many of them (hundreds) uses a lot of space
2370	and time, so using the defaults of five priorities (-2 .. +2) is usually	4045	and time, so using the defaults of five priorities (-2 .. +2) is usually
2371	fine.	4046	fine.
2372		4047
2373	If your embedding app does not need any priorities, defining these both to	4048	If your embedding application does not need any priorities, defining these
2374	C<0> will save some memory and cpu.	4049	both to C<0> will save some memory and CPU.
2375		4050
2376	=item EV_PERIODIC_ENABLE	4051	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4052	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4053	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
2377		4054
2378	If undefined or defined to be C<1>, then periodic timers are supported. If	4055	If undefined or defined to be C<1> (and the platform supports it), then
2379	defined to be C<0>, then they are not. Disabling them saves a few kB of	4056	the respective watcher type is supported. If defined to be C<0>, then it
2380	code.	4057	is not. Disabling watcher types mainly saves code size.
2381		4058
2382	=item EV_IDLE_ENABLE	4059	=item EV_FEATURES
2383
2384	If undefined or defined to be C<1>, then idle watchers are supported. If
2385	defined to be C<0>, then they are not. Disabling them saves a few kB of
2386	code.
2387
2388	=item EV_EMBED_ENABLE
2389
2390	If undefined or defined to be C<1>, then embed watchers are supported. If
2391	defined to be C<0>, then they are not.
2392
2393	=item EV_STAT_ENABLE
2394
2395	If undefined or defined to be C<1>, then stat watchers are supported. If
2396	defined to be C<0>, then they are not.
2397
2398	=item EV_FORK_ENABLE
2399
2400	If undefined or defined to be C<1>, then fork watchers are supported. If
2401	defined to be C<0>, then they are not.
2402
2403	=item EV_MINIMAL
2404		4060
2405	If you need to shave off some kilobytes of code at the expense of some	4061	If you need to shave off some kilobytes of code at the expense of some
2406	speed, define this symbol to C<1>. Currently only used for gcc to override	4062	speed (but with the full API), you can define this symbol to request
2407	some inlining decisions, saves roughly 30% codesize of amd64.	4063	certain subsets of functionality. The default is to enable all features
		4064	that can be enabled on the platform.
		4065
		4066	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4067	with some broad features you want) and then selectively re-enable
		4068	additional parts you want, for example if you want everything minimal,
		4069	but multiple event loop support, async and child watchers and the poll
		4070	backend, use this:
		4071
		4072	#define EV_FEATURES 0
		4073	#define EV_MULTIPLICITY 1
		4074	#define EV_USE_POLL 1
		4075	#define EV_CHILD_ENABLE 1
		4076	#define EV_ASYNC_ENABLE 1
		4077
		4078	The actual value is a bitset, it can be a combination of the following
		4079	values:
		4080
		4081	=over 4
		4082
		4083	=item C<1> - faster/larger code
		4084
		4085	Use larger code to speed up some operations.
		4086
		4087	Currently this is used to override some inlining decisions (enlarging the
		4088	code size by roughly 30% on amd64).
		4089
		4090	When optimising for size, use of compiler flags such as C<-Os> with
		4091	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4092	assertions.
		4093
		4094	=item C<2> - faster/larger data structures
		4095
		4096	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4097	hash table sizes and so on. This will usually further increase code size
		4098	and can additionally have an effect on the size of data structures at
		4099	runtime.
		4100
		4101	=item C<4> - full API configuration
		4102
		4103	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4104	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4105
		4106	=item C<8> - full API
		4107
		4108	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4109	details on which parts of the API are still available without this
		4110	feature, and do not complain if this subset changes over time.
		4111
		4112	=item C<16> - enable all optional watcher types
		4113
		4114	Enables all optional watcher types. If you want to selectively enable
		4115	only some watcher types other than I/O and timers (e.g. prepare,
		4116	embed, async, child...) you can enable them manually by defining
		4117	C<EV_watchertype_ENABLE> to C<1> instead.
		4118
		4119	=item C<32> - enable all backends
		4120
		4121	This enables all backends - without this feature, you need to enable at
		4122	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4123
		4124	=item C<64> - enable OS-specific "helper" APIs
		4125
		4126	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4127	default.
		4128
		4129	=back
		4130
		4131	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4132	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4133	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4134	watchers, timers and monotonic clock support.
		4135
		4136	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4137	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4138	your program might be left out as well - a binary starting a timer and an
		4139	I/O watcher then might come out at only 5Kb.
		4140
		4141	=item EV_AVOID_STDIO
		4142
		4143	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4144	functions (printf, scanf, perror etc.). This will increase the code size
		4145	somewhat, but if your program doesn't otherwise depend on stdio and your
		4146	libc allows it, this avoids linking in the stdio library which is quite
		4147	big.
		4148
		4149	Note that error messages might become less precise when this option is
		4150	enabled.
		4151
		4152	=item EV_NSIG
		4153
		4154	The highest supported signal number, +1 (or, the number of
		4155	signals): Normally, libev tries to deduce the maximum number of signals
		4156	automatically, but sometimes this fails, in which case it can be
		4157	specified. Also, using a lower number than detected (C<32> should be
		4158	good for about any system in existence) can save some memory, as libev
		4159	statically allocates some 12-24 bytes per signal number.
2408		4160
2409	=item EV_PID_HASHSIZE	4161	=item EV_PID_HASHSIZE
2410		4162
2411	C<ev_child> watchers use a small hash table to distribute workload by	4163	C<ev_child> watchers use a small hash table to distribute workload by
2412	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more	4164	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
2413	than enough. If you need to manage thousands of children you might want to	4165	usually more than enough. If you need to manage thousands of children you
2414	increase this value (I<must> be a power of two).	4166	might want to increase this value (I<must> be a power of two).
2415		4167
2416	=item EV_INOTIFY_HASHSIZE	4168	=item EV_INOTIFY_HASHSIZE
2417		4169
2418	C<ev_staz> watchers use a small hash table to distribute workload by	4170	C<ev_stat> watchers use a small hash table to distribute workload by
2419	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),	4171	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
2420	usually more than enough. If you need to manage thousands of C<ev_stat>	4172	disabled), usually more than enough. If you need to manage thousands of
2421	watchers you might want to increase this value (I<must> be a power of	4173	C<ev_stat> watchers you might want to increase this value (I<must> be a
2422	two).	4174	power of two).
		4175
		4176	=item EV_USE_4HEAP
		4177
		4178	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4179	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4180	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4181	faster performance with many (thousands) of watchers.
		4182
		4183	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4184	will be C<0>.
		4185
		4186	=item EV_HEAP_CACHE_AT
		4187
		4188	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4189	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4190	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4191	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4192	but avoids random read accesses on heap changes. This improves performance
		4193	noticeably with many (hundreds) of watchers.
		4194
		4195	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4196	will be C<0>.
		4197
		4198	=item EV_VERIFY
		4199
		4200	Controls how much internal verification (see C<ev_verify ()>) will
		4201	be done: If set to C<0>, no internal verification code will be compiled
		4202	in. If set to C<1>, then verification code will be compiled in, but not
		4203	called. If set to C<2>, then the internal verification code will be
		4204	called once per loop, which can slow down libev. If set to C<3>, then the
		4205	verification code will be called very frequently, which will slow down
		4206	libev considerably.
		4207
		4208	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4209	will be C<0>.
2423		4210
2424	=item EV_COMMON	4211	=item EV_COMMON
2425		4212
2426	By default, all watchers have a C<void *data> member. By redefining	4213	By default, all watchers have a C<void *data> member. By redefining
2427	this macro to a something else you can include more and other types of	4214	this macro to something else you can include more and other types of
2428	members. You have to define it each time you include one of the files,	4215	members. You have to define it each time you include one of the files,
2429	though, and it must be identical each time.	4216	though, and it must be identical each time.
2430		4217
2431	For example, the perl EV module uses something like this:	4218	For example, the perl EV module uses something like this:
2432		4219
2433	#define EV_COMMON \	4220	#define EV_COMMON \
2434	SV *self; /* contains this struct */ \	4221	SV *self; /* contains this struct */ \
2435	SV *cb_sv, *fh /* note no trailing ";" */	4222	SV *cb_sv, *fh /* note no trailing ";" */
2436		4223
2437	=item EV_CB_DECLARE (type)	4224	=item EV_CB_DECLARE (type)
2438		4225
2439	=item EV_CB_INVOKE (watcher, revents)	4226	=item EV_CB_INVOKE (watcher, revents)
2440		4227
2441	=item ev_set_cb (ev, cb)	4228	=item ev_set_cb (ev, cb)
2442		4229
2443	Can be used to change the callback member declaration in each watcher,	4230	Can be used to change the callback member declaration in each watcher,
2444	and the way callbacks are invoked and set. Must expand to a struct member	4231	and the way callbacks are invoked and set. Must expand to a struct member
2445	definition and a statement, respectively. See the F<ev.v> header file for	4232	definition and a statement, respectively. See the F<ev.h> header file for
2446	their default definitions. One possible use for overriding these is to	4233	their default definitions. One possible use for overriding these is to
2447	avoid the C<struct ev_loop *> as first argument in all cases, or to use	4234	avoid the C<struct ev_loop *> as first argument in all cases, or to use
2448	method calls instead of plain function calls in C++.	4235	method calls instead of plain function calls in C++.
2449		4236
		4237	=back
		4238
2450	=head2 EXPORTED API SYMBOLS	4239	=head2 EXPORTED API SYMBOLS
2451		4240
2452	If you need to re-export the API (e.g. via a dll) and you need a list of	4241	If you need to re-export the API (e.g. via a DLL) and you need a list of
2453	exported symbols, you can use the provided F<Symbol.*> files which list	4242	exported symbols, you can use the provided F<Symbol.*> files which list
2454	all public symbols, one per line:	4243	all public symbols, one per line:
2455		4244
2456	Symbols.ev for libev proper	4245	Symbols.ev for libev proper
2457	Symbols.event for the libevent emulation	4246	Symbols.event for the libevent emulation
2458		4247
2459	This can also be used to rename all public symbols to avoid clashes with	4248	This can also be used to rename all public symbols to avoid clashes with
2460	multiple versions of libev linked together (which is obviously bad in	4249	multiple versions of libev linked together (which is obviously bad in
2461	itself, but sometimes it is inconvinient to avoid this).	4250	itself, but sometimes it is inconvenient to avoid this).
2462		4251
2463	A sed comamnd like this will create wrapper C<#define>'s that you need to	4252	A sed command like this will create wrapper C<#define>'s that you need to
2464	include before including F<ev.h>:	4253	include before including F<ev.h>:
2465		4254
2466	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h	4255	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
2467		4256
2468	This would create a file F<wrap.h> which essentially looks like this:	4257	This would create a file F<wrap.h> which essentially looks like this:
…		…
2483	file.	4272	file.
2484		4273
2485	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file	4274	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
2486	that everybody includes and which overrides some configure choices:	4275	that everybody includes and which overrides some configure choices:
2487		4276
2488	#define EV_MINIMAL 1	4277	#define EV_FEATURES 8
2489	#define EV_USE_POLL 0	4278	#define EV_USE_SELECT 1
2490	#define EV_MULTIPLICITY 0
2491	#define EV_PERIODIC_ENABLE 0	4279	#define EV_PREPARE_ENABLE 1
		4280	#define EV_IDLE_ENABLE 1
2492	#define EV_STAT_ENABLE 0	4281	#define EV_SIGNAL_ENABLE 1
2493	#define EV_FORK_ENABLE 0	4282	#define EV_CHILD_ENABLE 1
		4283	#define EV_USE_STDEXCEPT 0
2494	#define EV_CONFIG_H <config.h>	4284	#define EV_CONFIG_H <config.h>
2495	#define EV_MINPRI 0
2496	#define EV_MAXPRI 0
2497		4285
2498	#include "ev++.h"	4286	#include "ev++.h"
2499		4287
2500	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	4288	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
2501		4289
2502	#include "ev_cpp.h"	4290	#include "ev_cpp.h"
2503	#include "ev.c"	4291	#include "ev.c"
2504		4292
		4293	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
2505		4294
		4295	=head2 THREADS AND COROUTINES
		4296
		4297	=head3 THREADS
		4298
		4299	All libev functions are reentrant and thread-safe unless explicitly
		4300	documented otherwise, but libev implements no locking itself. This means
		4301	that you can use as many loops as you want in parallel, as long as there
		4302	are no concurrent calls into any libev function with the same loop
		4303	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4304	of course): libev guarantees that different event loops share no data
		4305	structures that need any locking.
		4306
		4307	Or to put it differently: calls with different loop parameters can be done
		4308	concurrently from multiple threads, calls with the same loop parameter
		4309	must be done serially (but can be done from different threads, as long as
		4310	only one thread ever is inside a call at any point in time, e.g. by using
		4311	a mutex per loop).
		4312
		4313	Specifically to support threads (and signal handlers), libev implements
		4314	so-called C<ev_async> watchers, which allow some limited form of
		4315	concurrency on the same event loop, namely waking it up "from the
		4316	outside".
		4317
		4318	If you want to know which design (one loop, locking, or multiple loops
		4319	without or something else still) is best for your problem, then I cannot
		4320	help you, but here is some generic advice:
		4321
		4322	=over 4
		4323
		4324	=item * most applications have a main thread: use the default libev loop
		4325	in that thread, or create a separate thread running only the default loop.
		4326
		4327	This helps integrating other libraries or software modules that use libev
		4328	themselves and don't care/know about threading.
		4329
		4330	=item * one loop per thread is usually a good model.
		4331
		4332	Doing this is almost never wrong, sometimes a better-performance model
		4333	exists, but it is always a good start.
		4334
		4335	=item * other models exist, such as the leader/follower pattern, where one
		4336	loop is handed through multiple threads in a kind of round-robin fashion.
		4337
		4338	Choosing a model is hard - look around, learn, know that usually you can do
		4339	better than you currently do :-)
		4340
		4341	=item * often you need to talk to some other thread which blocks in the
		4342	event loop.
		4343
		4344	C<ev_async> watchers can be used to wake them up from other threads safely
		4345	(or from signal contexts...).
		4346
		4347	An example use would be to communicate signals or other events that only
		4348	work in the default loop by registering the signal watcher with the
		4349	default loop and triggering an C<ev_async> watcher from the default loop
		4350	watcher callback into the event loop interested in the signal.
		4351
		4352	=back
		4353
		4354	=head4 THREAD LOCKING EXAMPLE
		4355
		4356	Here is a fictitious example of how to run an event loop in a different
		4357	thread than where callbacks are being invoked and watchers are
		4358	created/added/removed.
		4359
		4360	For a real-world example, see the C<EV::Loop::Async> perl module,
		4361	which uses exactly this technique (which is suited for many high-level
		4362	languages).
		4363
		4364	The example uses a pthread mutex to protect the loop data, a condition
		4365	variable to wait for callback invocations, an async watcher to notify the
		4366	event loop thread and an unspecified mechanism to wake up the main thread.
		4367
		4368	First, you need to associate some data with the event loop:
		4369
		4370	typedef struct {
		4371	mutex_t lock; /* global loop lock */
		4372	ev_async async_w;
		4373	thread_t tid;
		4374	cond_t invoke_cv;
		4375	} userdata;
		4376
		4377	void prepare_loop (EV_P)
		4378	{
		4379	// for simplicity, we use a static userdata struct.
		4380	static userdata u;
		4381
		4382	ev_async_init (&u->async_w, async_cb);
		4383	ev_async_start (EV_A_ &u->async_w);
		4384
		4385	pthread_mutex_init (&u->lock, 0);
		4386	pthread_cond_init (&u->invoke_cv, 0);
		4387
		4388	// now associate this with the loop
		4389	ev_set_userdata (EV_A_ u);
		4390	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4391	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4392
		4393	// then create the thread running ev_loop
		4394	pthread_create (&u->tid, 0, l_run, EV_A);
		4395	}
		4396
		4397	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4398	solely to wake up the event loop so it takes notice of any new watchers
		4399	that might have been added:
		4400
		4401	static void
		4402	async_cb (EV_P_ ev_async *w, int revents)
		4403	{
		4404	// just used for the side effects
		4405	}
		4406
		4407	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4408	protecting the loop data, respectively.
		4409
		4410	static void
		4411	l_release (EV_P)
		4412	{
		4413	userdata *u = ev_userdata (EV_A);
		4414	pthread_mutex_unlock (&u->lock);
		4415	}
		4416
		4417	static void
		4418	l_acquire (EV_P)
		4419	{
		4420	userdata *u = ev_userdata (EV_A);
		4421	pthread_mutex_lock (&u->lock);
		4422	}
		4423
		4424	The event loop thread first acquires the mutex, and then jumps straight
		4425	into C<ev_run>:
		4426
		4427	void *
		4428	l_run (void *thr_arg)
		4429	{
		4430	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4431
		4432	l_acquire (EV_A);
		4433	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4434	ev_run (EV_A_ 0);
		4435	l_release (EV_A);
		4436
		4437	return 0;
		4438	}
		4439
		4440	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4441	signal the main thread via some unspecified mechanism (signals? pipe
		4442	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4443	have been called (in a while loop because a) spurious wakeups are possible
		4444	and b) skipping inter-thread-communication when there are no pending
		4445	watchers is very beneficial):
		4446
		4447	static void
		4448	l_invoke (EV_P)
		4449	{
		4450	userdata *u = ev_userdata (EV_A);
		4451
		4452	while (ev_pending_count (EV_A))
		4453	{
		4454	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4455	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4456	}
		4457	}
		4458
		4459	Now, whenever the main thread gets told to invoke pending watchers, it
		4460	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4461	thread to continue:
		4462
		4463	static void
		4464	real_invoke_pending (EV_P)
		4465	{
		4466	userdata *u = ev_userdata (EV_A);
		4467
		4468	pthread_mutex_lock (&u->lock);
		4469	ev_invoke_pending (EV_A);
		4470	pthread_cond_signal (&u->invoke_cv);
		4471	pthread_mutex_unlock (&u->lock);
		4472	}
		4473
		4474	Whenever you want to start/stop a watcher or do other modifications to an
		4475	event loop, you will now have to lock:
		4476
		4477	ev_timer timeout_watcher;
		4478	userdata *u = ev_userdata (EV_A);
		4479
		4480	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4481
		4482	pthread_mutex_lock (&u->lock);
		4483	ev_timer_start (EV_A_ &timeout_watcher);
		4484	ev_async_send (EV_A_ &u->async_w);
		4485	pthread_mutex_unlock (&u->lock);
		4486
		4487	Note that sending the C<ev_async> watcher is required because otherwise
		4488	an event loop currently blocking in the kernel will have no knowledge
		4489	about the newly added timer. By waking up the loop it will pick up any new
		4490	watchers in the next event loop iteration.
		4491
		4492	=head3 COROUTINES
		4493
		4494	Libev is very accommodating to coroutines ("cooperative threads"):
		4495	libev fully supports nesting calls to its functions from different
		4496	coroutines (e.g. you can call C<ev_run> on the same loop from two
		4497	different coroutines, and switch freely between both coroutines running
		4498	the loop, as long as you don't confuse yourself). The only exception is
		4499	that you must not do this from C<ev_periodic> reschedule callbacks.
		4500
		4501	Care has been taken to ensure that libev does not keep local state inside
		4502	C<ev_run>, and other calls do not usually allow for coroutine switches as
		4503	they do not call any callbacks.
		4504
		4505	=head2 COMPILER WARNINGS
		4506
		4507	Depending on your compiler and compiler settings, you might get no or a
		4508	lot of warnings when compiling libev code. Some people are apparently
		4509	scared by this.
		4510
		4511	However, these are unavoidable for many reasons. For one, each compiler
		4512	has different warnings, and each user has different tastes regarding
		4513	warning options. "Warn-free" code therefore cannot be a goal except when
		4514	targeting a specific compiler and compiler-version.
		4515
		4516	Another reason is that some compiler warnings require elaborate
		4517	workarounds, or other changes to the code that make it less clear and less
		4518	maintainable.
		4519
		4520	And of course, some compiler warnings are just plain stupid, or simply
		4521	wrong (because they don't actually warn about the condition their message
		4522	seems to warn about). For example, certain older gcc versions had some
		4523	warnings that resulted in an extreme number of false positives. These have
		4524	been fixed, but some people still insist on making code warn-free with
		4525	such buggy versions.
		4526
		4527	While libev is written to generate as few warnings as possible,
		4528	"warn-free" code is not a goal, and it is recommended not to build libev
		4529	with any compiler warnings enabled unless you are prepared to cope with
		4530	them (e.g. by ignoring them). Remember that warnings are just that:
		4531	warnings, not errors, or proof of bugs.
		4532
		4533
		4534	=head2 VALGRIND
		4535
		4536	Valgrind has a special section here because it is a popular tool that is
		4537	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4538
		4539	If you think you found a bug (memory leak, uninitialised data access etc.)
		4540	in libev, then check twice: If valgrind reports something like:
		4541
		4542	==2274== definitely lost: 0 bytes in 0 blocks.
		4543	==2274== possibly lost: 0 bytes in 0 blocks.
		4544	==2274== still reachable: 256 bytes in 1 blocks.
		4545
		4546	Then there is no memory leak, just as memory accounted to global variables
		4547	is not a memleak - the memory is still being referenced, and didn't leak.
		4548
		4549	Similarly, under some circumstances, valgrind might report kernel bugs
		4550	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4551	although an acceptable workaround has been found here), or it might be
		4552	confused.
		4553
		4554	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4555	make it into some kind of religion.
		4556
		4557	If you are unsure about something, feel free to contact the mailing list
		4558	with the full valgrind report and an explanation on why you think this
		4559	is a bug in libev (best check the archives, too :). However, don't be
		4560	annoyed when you get a brisk "this is no bug" answer and take the chance
		4561	of learning how to interpret valgrind properly.
		4562
		4563	If you need, for some reason, empty reports from valgrind for your project
		4564	I suggest using suppression lists.
		4565
		4566
		4567	=head1 PORTABILITY NOTES
		4568
		4569	=head2 GNU/LINUX 32 BIT LIMITATIONS
		4570
		4571	GNU/Linux is the only common platform that supports 64 bit file/large file
		4572	interfaces but I<disables> them by default.
		4573
		4574	That means that libev compiled in the default environment doesn't support
		4575	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		4576
		4577	Unfortunately, many programs try to work around this GNU/Linux issue
		4578	by enabling the large file API, which makes them incompatible with the
		4579	standard libev compiled for their system.
		4580
		4581	Likewise, libev cannot enable the large file API itself as this would
		4582	suddenly make it incompatible to the default compile time environment,
		4583	i.e. all programs not using special compile switches.
		4584
		4585	=head2 OS/X AND DARWIN BUGS
		4586
		4587	The whole thing is a bug if you ask me - basically any system interface
		4588	you touch is broken, whether it is locales, poll, kqueue or even the
		4589	OpenGL drivers.
		4590
		4591	=head3 C<kqueue> is buggy
		4592
		4593	The kqueue syscall is broken in all known versions - most versions support
		4594	only sockets, many support pipes.
		4595
		4596	Libev tries to work around this by not using C<kqueue> by default on this
		4597	rotten platform, but of course you can still ask for it when creating a
		4598	loop - embedding a socket-only kqueue loop into a select-based one is
		4599	probably going to work well.
		4600
		4601	=head3 C<poll> is buggy
		4602
		4603	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4604	implementation by something calling C<kqueue> internally around the 10.5.6
		4605	release, so now C<kqueue> I<and> C<poll> are broken.
		4606
		4607	Libev tries to work around this by not using C<poll> by default on
		4608	this rotten platform, but of course you can still ask for it when creating
		4609	a loop.
		4610
		4611	=head3 C<select> is buggy
		4612
		4613	All that's left is C<select>, and of course Apple found a way to fuck this
		4614	one up as well: On OS/X, C<select> actively limits the number of file
		4615	descriptors you can pass in to 1024 - your program suddenly crashes when
		4616	you use more.
		4617
		4618	There is an undocumented "workaround" for this - defining
		4619	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4620	work on OS/X.
		4621
		4622	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4623
		4624	=head3 C<errno> reentrancy
		4625
		4626	The default compile environment on Solaris is unfortunately so
		4627	thread-unsafe that you can't even use components/libraries compiled
		4628	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		4629	defined by default. A valid, if stupid, implementation choice.
		4630
		4631	If you want to use libev in threaded environments you have to make sure
		4632	it's compiled with C<_REENTRANT> defined.
		4633
		4634	=head3 Event port backend
		4635
		4636	The scalable event interface for Solaris is called "event
		4637	ports". Unfortunately, this mechanism is very buggy in all major
		4638	releases. If you run into high CPU usage, your program freezes or you get
		4639	a large number of spurious wakeups, make sure you have all the relevant
		4640	and latest kernel patches applied. No, I don't know which ones, but there
		4641	are multiple ones to apply, and afterwards, event ports actually work
		4642	great.
		4643
		4644	If you can't get it to work, you can try running the program by setting
		4645	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4646	C<select> backends.
		4647
		4648	=head2 AIX POLL BUG
		4649
		4650	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4651	this by trying to avoid the poll backend altogether (i.e. it's not even
		4652	compiled in), which normally isn't a big problem as C<select> works fine
		4653	with large bitsets on AIX, and AIX is dead anyway.
		4654
		4655	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4656
		4657	=head3 General issues
		4658
		4659	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4660	requires, and its I/O model is fundamentally incompatible with the POSIX
		4661	model. Libev still offers limited functionality on this platform in
		4662	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4663	descriptors. This only applies when using Win32 natively, not when using
		4664	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4665	as every compielr comes with a slightly differently broken/incompatible
		4666	environment.
		4667
		4668	Lifting these limitations would basically require the full
		4669	re-implementation of the I/O system. If you are into this kind of thing,
		4670	then note that glib does exactly that for you in a very portable way (note
		4671	also that glib is the slowest event library known to man).
		4672
		4673	There is no supported compilation method available on windows except
		4674	embedding it into other applications.
		4675
		4676	Sensible signal handling is officially unsupported by Microsoft - libev
		4677	tries its best, but under most conditions, signals will simply not work.
		4678
		4679	Not a libev limitation but worth mentioning: windows apparently doesn't
		4680	accept large writes: instead of resulting in a partial write, windows will
		4681	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4682	so make sure you only write small amounts into your sockets (less than a
		4683	megabyte seems safe, but this apparently depends on the amount of memory
		4684	available).
		4685
		4686	Due to the many, low, and arbitrary limits on the win32 platform and
		4687	the abysmal performance of winsockets, using a large number of sockets
		4688	is not recommended (and not reasonable). If your program needs to use
		4689	more than a hundred or so sockets, then likely it needs to use a totally
		4690	different implementation for windows, as libev offers the POSIX readiness
		4691	notification model, which cannot be implemented efficiently on windows
		4692	(due to Microsoft monopoly games).
		4693
		4694	A typical way to use libev under windows is to embed it (see the embedding
		4695	section for details) and use the following F<evwrap.h> header file instead
		4696	of F<ev.h>:
		4697
		4698	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4699	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4700
		4701	#include "ev.h"
		4702
		4703	And compile the following F<evwrap.c> file into your project (make sure
		4704	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4705
		4706	#include "evwrap.h"
		4707	#include "ev.c"
		4708
		4709	=head3 The winsocket C<select> function
		4710
		4711	The winsocket C<select> function doesn't follow POSIX in that it
		4712	requires socket I<handles> and not socket I<file descriptors> (it is
		4713	also extremely buggy). This makes select very inefficient, and also
		4714	requires a mapping from file descriptors to socket handles (the Microsoft
		4715	C runtime provides the function C<_open_osfhandle> for this). See the
		4716	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4717	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4718
		4719	The configuration for a "naked" win32 using the Microsoft runtime
		4720	libraries and raw winsocket select is:
		4721
		4722	#define EV_USE_SELECT 1
		4723	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4724
		4725	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4726	complexity in the O(n²) range when using win32.
		4727
		4728	=head3 Limited number of file descriptors
		4729
		4730	Windows has numerous arbitrary (and low) limits on things.
		4731
		4732	Early versions of winsocket's select only supported waiting for a maximum
		4733	of C<64> handles (probably owning to the fact that all windows kernels
		4734	can only wait for C<64> things at the same time internally; Microsoft
		4735	recommends spawning a chain of threads and wait for 63 handles and the
		4736	previous thread in each. Sounds great!).
		4737
		4738	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4739	to some high number (e.g. C<2048>) before compiling the winsocket select
		4740	call (which might be in libev or elsewhere, for example, perl and many
		4741	other interpreters do their own select emulation on windows).
		4742
		4743	Another limit is the number of file descriptors in the Microsoft runtime
		4744	libraries, which by default is C<64> (there must be a hidden I<64>
		4745	fetish or something like this inside Microsoft). You can increase this
		4746	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4747	(another arbitrary limit), but is broken in many versions of the Microsoft
		4748	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4749	(depending on windows version and/or the phase of the moon). To get more,
		4750	you need to wrap all I/O functions and provide your own fd management, but
		4751	the cost of calling select (O(n²)) will likely make this unworkable.
		4752
		4753	=head2 PORTABILITY REQUIREMENTS
		4754
		4755	In addition to a working ISO-C implementation and of course the
		4756	backend-specific APIs, libev relies on a few additional extensions:
		4757
		4758	=over 4
		4759
		4760	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4761	calling conventions regardless of C<ev_watcher_type *>.
		4762
		4763	Libev assumes not only that all watcher pointers have the same internal
		4764	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4765	assumes that the same (machine) code can be used to call any watcher
		4766	callback: The watcher callbacks have different type signatures, but libev
		4767	calls them using an C<ev_watcher *> internally.
		4768
		4769	=item pointer accesses must be thread-atomic
		4770
		4771	Accessing a pointer value must be atomic, it must both be readable and
		4772	writable in one piece - this is the case on all current architectures.
		4773
		4774	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4775
		4776	The type C<sig_atomic_t volatile> (or whatever is defined as
		4777	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4778	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4779	believed to be sufficiently portable.
		4780
		4781	=item C<sigprocmask> must work in a threaded environment
		4782
		4783	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4784	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4785	pthread implementations will either allow C<sigprocmask> in the "main
		4786	thread" or will block signals process-wide, both behaviours would
		4787	be compatible with libev. Interaction between C<sigprocmask> and
		4788	C<pthread_sigmask> could complicate things, however.
		4789
		4790	The most portable way to handle signals is to block signals in all threads
		4791	except the initial one, and run the default loop in the initial thread as
		4792	well.
		4793
		4794	=item C<long> must be large enough for common memory allocation sizes
		4795
		4796	To improve portability and simplify its API, libev uses C<long> internally
		4797	instead of C<size_t> when allocating its data structures. On non-POSIX
		4798	systems (Microsoft...) this might be unexpectedly low, but is still at
		4799	least 31 bits everywhere, which is enough for hundreds of millions of
		4800	watchers.
		4801
		4802	=item C<double> must hold a time value in seconds with enough accuracy
		4803
		4804	The type C<double> is used to represent timestamps. It is required to
		4805	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		4806	good enough for at least into the year 4000 with millisecond accuracy
		4807	(the design goal for libev). This requirement is overfulfilled by
		4808	implementations using IEEE 754, which is basically all existing ones. With
		4809	IEEE 754 doubles, you get microsecond accuracy until at least 2200.
		4810
		4811	=back
		4812
		4813	If you know of other additional requirements drop me a note.
		4814
		4815
2506	=head1 COMPLEXITIES	4816	=head1 ALGORITHMIC COMPLEXITIES
2507		4817
2508	In this section the complexities of (many of) the algorithms used inside	4818	In this section the complexities of (many of) the algorithms used inside
2509	libev will be explained. For complexity discussions about backends see the	4819	libev will be documented. For complexity discussions about backends see
2510	documentation for C<ev_default_init>.	4820	the documentation for C<ev_default_init>.
2511		4821
2512	All of the following are about amortised time: If an array needs to be	4822	All of the following are about amortised time: If an array needs to be
2513	extended, libev needs to realloc and move the whole array, but this	4823	extended, libev needs to realloc and move the whole array, but this
2514	happens asymptotically never with higher number of elements, so O(1) might	4824	happens asymptotically rarer with higher number of elements, so O(1) might
2515	mean it might do a lengthy realloc operation in rare cases, but on average	4825	mean that libev does a lengthy realloc operation in rare cases, but on
2516	it is much faster and asymptotically approaches constant time.	4826	average it is much faster and asymptotically approaches constant time.
2517		4827
2518	=over 4	4828	=over 4
2519		4829
2520	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	4830	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
2521		4831
2522	This means that, when you have a watcher that triggers in one hour and	4832	This means that, when you have a watcher that triggers in one hour and
2523	there are 100 watchers that would trigger before that then inserting will	4833	there are 100 watchers that would trigger before that, then inserting will
2524	have to skip those 100 watchers.	4834	have to skip roughly seven (C<ld 100>) of these watchers.
2525		4835
2526	=item Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)	4836	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2527		4837
2528	That means that for changing a timer costs less than removing/adding them	4838	That means that changing a timer costs less than removing/adding them,
2529	as only the relative motion in the event queue has to be paid for.	4839	as only the relative motion in the event queue has to be paid for.
2530		4840
2531	=item Starting io/check/prepare/idle/signal/child watchers: O(1)	4841	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
2532		4842
2533	These just add the watcher into an array or at the head of a list.	4843	These just add the watcher into an array or at the head of a list.
		4844
2534	=item Stopping check/prepare/idle watchers: O(1)	4845	=item Stopping check/prepare/idle/fork/async watchers: O(1)
2535		4846
2536	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	4847	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2537		4848
2538	These watchers are stored in lists then need to be walked to find the	4849	These watchers are stored in lists, so they need to be walked to find the
2539	correct watcher to remove. The lists are usually short (you don't usually	4850	correct watcher to remove. The lists are usually short (you don't usually
2540	have many watchers waiting for the same fd or signal).	4851	have many watchers waiting for the same fd or signal: one is typical, two
		4852	is rare).
2541		4853
2542	=item Finding the next timer per loop iteration: O(1)	4854	=item Finding the next timer in each loop iteration: O(1)
		4855
		4856	By virtue of using a binary or 4-heap, the next timer is always found at a
		4857	fixed position in the storage array.
2543		4858
2544	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)	4859	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
2545		4860
2546	A change means an I/O watcher gets started or stopped, which requires	4861	A change means an I/O watcher gets started or stopped, which requires
2547	libev to recalculate its status (and possibly tell the kernel).	4862	libev to recalculate its status (and possibly tell the kernel, depending
		4863	on backend and whether C<ev_io_set> was used).
2548		4864
2549	=item Activating one watcher: O(1)	4865	=item Activating one watcher (putting it into the pending state): O(1)
2550		4866
2551	=item Priority handling: O(number_of_priorities)	4867	=item Priority handling: O(number_of_priorities)
2552		4868
2553	Priorities are implemented by allocating some space for each	4869	Priorities are implemented by allocating some space for each
2554	priority. When doing priority-based operations, libev usually has to	4870	priority. When doing priority-based operations, libev usually has to
2555	linearly search all the priorities.	4871	linearly search all the priorities, but starting/stopping and activating
		4872	watchers becomes O(1) with respect to priority handling.
		4873
		4874	=item Sending an ev_async: O(1)
		4875
		4876	=item Processing ev_async_send: O(number_of_async_watchers)
		4877
		4878	=item Processing signals: O(max_signal_number)
		4879
		4880	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4881	calls in the current loop iteration. Checking for async and signal events
		4882	involves iterating over all running async watchers or all signal numbers.
2556		4883
2557	=back	4884	=back
2558		4885
2559		4886
		4887	=head1 PORTING FROM LIBEV 3.X TO 4.X
		4888
		4889	The major version 4 introduced some incompatible changes to the API.
		4890
		4891	At the moment, the C<ev.h> header file provides compatibility definitions
		4892	for all changes, so most programs should still compile. The compatibility
		4893	layer might be removed in later versions of libev, so better update to the
		4894	new API early than late.
		4895
		4896	=over 4
		4897
		4898	=item C<EV_COMPAT3> backwards compatibility mechanism
		4899
		4900	The backward compatibility mechanism can be controlled by
		4901	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>
		4902	section.
		4903
		4904	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		4905
		4906	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		4907
		4908	ev_loop_destroy (EV_DEFAULT_UC);
		4909	ev_loop_fork (EV_DEFAULT);
		4910
		4911	=item function/symbol renames
		4912
		4913	A number of functions and symbols have been renamed:
		4914
		4915	ev_loop => ev_run
		4916	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		4917	EVLOOP_ONESHOT => EVRUN_ONCE
		4918
		4919	ev_unloop => ev_break
		4920	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		4921	EVUNLOOP_ONE => EVBREAK_ONE
		4922	EVUNLOOP_ALL => EVBREAK_ALL
		4923
		4924	EV_TIMEOUT => EV_TIMER
		4925
		4926	ev_loop_count => ev_iteration
		4927	ev_loop_depth => ev_depth
		4928	ev_loop_verify => ev_verify
		4929
		4930	Most functions working on C<struct ev_loop> objects don't have an
		4931	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		4932	associated constants have been renamed to not collide with the C<struct
		4933	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		4934	as all other watcher types. Note that C<ev_loop_fork> is still called
		4935	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		4936	typedef.
		4937
		4938	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		4939
		4940	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		4941	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		4942	and work, but the library code will of course be larger.
		4943
		4944	=back
		4945
		4946
		4947	=head1 GLOSSARY
		4948
		4949	=over 4
		4950
		4951	=item active
		4952
		4953	A watcher is active as long as it has been started and not yet stopped.
		4954	See L<WATCHER STATES> for details.
		4955
		4956	=item application
		4957
		4958	In this document, an application is whatever is using libev.
		4959
		4960	=item backend
		4961
		4962	The part of the code dealing with the operating system interfaces.
		4963
		4964	=item callback
		4965
		4966	The address of a function that is called when some event has been
		4967	detected. Callbacks are being passed the event loop, the watcher that
		4968	received the event, and the actual event bitset.
		4969
		4970	=item callback/watcher invocation
		4971
		4972	The act of calling the callback associated with a watcher.
		4973
		4974	=item event
		4975
		4976	A change of state of some external event, such as data now being available
		4977	for reading on a file descriptor, time having passed or simply not having
		4978	any other events happening anymore.
		4979
		4980	In libev, events are represented as single bits (such as C<EV_READ> or
		4981	C<EV_TIMER>).
		4982
		4983	=item event library
		4984
		4985	A software package implementing an event model and loop.
		4986
		4987	=item event loop
		4988
		4989	An entity that handles and processes external events and converts them
		4990	into callback invocations.
		4991
		4992	=item event model
		4993
		4994	The model used to describe how an event loop handles and processes
		4995	watchers and events.
		4996
		4997	=item pending
		4998
		4999	A watcher is pending as soon as the corresponding event has been
		5000	detected. See L<WATCHER STATES> for details.
		5001
		5002	=item real time
		5003
		5004	The physical time that is observed. It is apparently strictly monotonic :)
		5005
		5006	=item wall-clock time
		5007
		5008	The time and date as shown on clocks. Unlike real time, it can actually
		5009	be wrong and jump forwards and backwards, e.g. when the you adjust your
		5010	clock.
		5011
		5012	=item watcher
		5013
		5014	A data structure that describes interest in certain events. Watchers need
		5015	to be started (attached to an event loop) before they can receive events.
		5016
		5017	=back
		5018
2560	=head1 AUTHOR	5019	=head1 AUTHOR
2561		5020
2562	Marc Lehmann <libev@schmorp.de>.	5021	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5022	Magnusson and Emanuele Giaquinta.
2563		5023

Diff Legend

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