ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.30 by root, Fri Nov 23 04:36:03 2007 UTC vs.
Revision 1.322 by root, Sun Oct 24 17:58:41 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 DESCRIPTION	9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
		18	ev_io stdin_watcher;
		19	ev_timer timeout_watcher;
		20
		21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
		23	static void
		24	stdin_cb (EV_P_ ev_io *w, int revents)
		25	{
		26	puts ("stdin ready");
		27	// for one-shot events, one must manually stop the watcher
		28	// with its corresponding stop function.
		29	ev_io_stop (EV_A_ w);
		30
		31	// this causes all nested ev_run's to stop iterating
		32	ev_break (EV_A_ EVBREAK_ALL);
		33	}
		34
		35	// another callback, this time for a time-out
		36	static void
		37	timeout_cb (EV_P_ ev_timer *w, int revents)
		38	{
		39	puts ("timeout");
		40	// this causes the innermost ev_run to stop iterating
		41	ev_break (EV_A_ EVBREAK_ONE);
		42	}
		43
		44	int
		45	main (void)
		46	{
		47	// use the default event loop unless you have special needs
		48	struct ev_loop *loop = EV_DEFAULT;
		49
		50	// initialise an io watcher, then start it
		51	// this one will watch for stdin to become readable
		52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		53	ev_io_start (loop, &stdin_watcher);
		54
		55	// initialise a timer watcher, then start it
		56	// simple non-repeating 5.5 second timeout
		57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		58	ev_timer_start (loop, &timeout_watcher);
		59
		60	// now wait for events to arrive
		61	ev_run (loop, 0);
		62
		63	// unloop was called, so exit
		64	return 0;
		65	}
		66
		67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
		70
		71	The newest version of this document is also available as an html-formatted
		72	web page you might find easier to navigate when reading it for the first
		73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familiarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 ABOUT LIBEV
10		84
11	Libev is an event loop: you register interest in certain events (such as a	85	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	86	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	87	these event sources and provide your program with events.
14		88
15	To do this, it must take more or less complete control over your process	89	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	90	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	91	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	93	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	94	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	95	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	96	watcher.
23		97
24	=head1 FEATURES	98	=head2 FEATURES
25		99
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	100	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	101	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	102	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		103	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		104	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		105	timers (C<ev_timer>), absolute timers with customised rescheduling
		106	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	107	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	108	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		109	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		110	limited support for fork events (C<ev_fork>).
		111
		112	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	113	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	114	for example).
33		115
34	=head1 CONVENTIONS	116	=head2 CONVENTIONS
35		117
36	Libev is very configurable. In this manual the default configuration	118	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	119	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	120	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	121	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	122	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	123	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	124	this argument.
43		125
44	=head1 TIME REPRESENTATION	126	=head2 TIME REPRESENTATION
45		127
46	Libev represents time as a single floating point number, representing the	128	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	129	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	130	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	131	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	132	too. It usually aliases to the C<double> type in C. When you need to do
		133	any calculations on it, you should treat it as some floating point value.
		134
		135	Unlike the name component C<stamp> might indicate, it is also used for
		136	time differences (e.g. delays) throughout libev.
		137
		138	=head1 ERROR HANDLING
		139
		140	Libev knows three classes of errors: operating system errors, usage errors
		141	and internal errors (bugs).
		142
		143	When libev catches an operating system error it cannot handle (for example
		144	a system call indicating a condition libev cannot fix), it calls the callback
		145	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		146	abort. The default is to print a diagnostic message and to call C<abort
		147	()>.
		148
		149	When libev detects a usage error such as a negative timer interval, then
		150	it will print a diagnostic message and abort (via the C<assert> mechanism,
		151	so C<NDEBUG> will disable this checking): these are programming errors in
		152	the libev caller and need to be fixed there.
		153
		154	Libev also has a few internal error-checking C<assert>ions, and also has
		155	extensive consistency checking code. These do not trigger under normal
		156	circumstances, as they indicate either a bug in libev or worse.
		157
51		158
52	=head1 GLOBAL FUNCTIONS	159	=head1 GLOBAL FUNCTIONS
53		160
54	These functions can be called anytime, even before initialising the	161	These functions can be called anytime, even before initialising the
55	library in any way.	162	library in any way.
…		…
58		165
59	=item ev_tstamp ev_time ()	166	=item ev_tstamp ev_time ()
60		167
61	Returns the current time as libev would use it. Please note that the	168	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	169	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	170	you actually want to know. Also interesting is the combination of
		171	C<ev_update_now> and C<ev_now>.
		172
		173	=item ev_sleep (ev_tstamp interval)
		174
		175	Sleep for the given interval: The current thread will be blocked until
		176	either it is interrupted or the given time interval has passed. Basically
		177	this is a sub-second-resolution C<sleep ()>.
64		178
65	=item int ev_version_major ()	179	=item int ev_version_major ()
66		180
67	=item int ev_version_minor ()	181	=item int ev_version_minor ()
68		182
69	You can find out the major and minor version numbers of the library	183	You can find out the major and minor ABI version numbers of the library
70	you linked against by calling the functions C<ev_version_major> and	184	you linked against by calling the functions C<ev_version_major> and
71	C<ev_version_minor>. If you want, you can compare against the global	185	C<ev_version_minor>. If you want, you can compare against the global
72	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	186	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
73	version of the library your program was compiled against.	187	version of the library your program was compiled against.
74		188
		189	These version numbers refer to the ABI version of the library, not the
		190	release version.
		191
75	Usually, it's a good idea to terminate if the major versions mismatch,	192	Usually, it's a good idea to terminate if the major versions mismatch,
76	as this indicates an incompatible change. Minor versions are usually	193	as this indicates an incompatible change. Minor versions are usually
77	compatible to older versions, so a larger minor version alone is usually	194	compatible to older versions, so a larger minor version alone is usually
78	not a problem.	195	not a problem.
79		196
		197	Example: Make sure we haven't accidentally been linked against the wrong
		198	version (note, however, that this will not detect other ABI mismatches,
		199	such as LFS or reentrancy).
		200
		201	assert (("libev version mismatch",
		202	ev_version_major () == EV_VERSION_MAJOR
		203	&& ev_version_minor () >= EV_VERSION_MINOR));
		204
		205	=item unsigned int ev_supported_backends ()
		206
		207	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		208	value) compiled into this binary of libev (independent of their
		209	availability on the system you are running on). See C<ev_default_loop> for
		210	a description of the set values.
		211
		212	Example: make sure we have the epoll method, because yeah this is cool and
		213	a must have and can we have a torrent of it please!!!11
		214
		215	assert (("sorry, no epoll, no sex",
		216	ev_supported_backends () & EVBACKEND_EPOLL));
		217
		218	=item unsigned int ev_recommended_backends ()
		219
		220	Return the set of all backends compiled into this binary of libev and
		221	also recommended for this platform, meaning it will work for most file
		222	descriptor types. This set is often smaller than the one returned by
		223	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		224	and will not be auto-detected unless you explicitly request it (assuming
		225	you know what you are doing). This is the set of backends that libev will
		226	probe for if you specify no backends explicitly.
		227
		228	=item unsigned int ev_embeddable_backends ()
		229
		230	Returns the set of backends that are embeddable in other event loops. This
		231	value is platform-specific but can include backends not available on the
		232	current system. To find which embeddable backends might be supported on
		233	the current system, you would need to look at C<ev_embeddable_backends ()
		234	& ev_supported_backends ()>, likewise for recommended ones.
		235
		236	See the description of C<ev_embed> watchers for more info.
		237
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	238	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
81		239
82	Sets the allocation function to use (the prototype is similar to the	240	Sets the allocation function to use (the prototype is similar - the
83	realloc C function, the semantics are identical). It is used to allocate	241	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
84	and free memory (no surprises here). If it returns zero when memory	242	used to allocate and free memory (no surprises here). If it returns zero
85	needs to be allocated, the library might abort or take some potentially	243	when memory needs to be allocated (C<size != 0>), the library might abort
86	destructive action. The default is your system realloc function.	244	or take some potentially destructive action.
		245
		246	Since some systems (at least OpenBSD and Darwin) fail to implement
		247	correct C<realloc> semantics, libev will use a wrapper around the system
		248	C<realloc> and C<free> functions by default.
87		249
88	You could override this function in high-availability programs to, say,	250	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	251	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	252	or even to sleep a while and retry until some memory is available.
91		253
		254	Example: Replace the libev allocator with one that waits a bit and then
		255	retries (example requires a standards-compliant C<realloc>).
		256
		257	static void *
		258	persistent_realloc (void *ptr, size_t size)
		259	{
		260	for (;;)
		261	{
		262	void *newptr = realloc (ptr, size);
		263
		264	if (newptr)
		265	return newptr;
		266
		267	sleep (60);
		268	}
		269	}
		270
		271	...
		272	ev_set_allocator (persistent_realloc);
		273
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	274	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
93		275
94	Set the callback function to call on a retryable syscall error (such	276	Set the callback function to call on a retryable system call error (such
95	as failed select, poll, epoll_wait). The message is a printable string	277	as failed select, poll, epoll_wait). The message is a printable string
96	indicating the system call or subsystem causing the problem. If this	278	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	279	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	280	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	281	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	282	(such as abort).
101		283
		284	Example: This is basically the same thing that libev does internally, too.
		285
		286	static void
		287	fatal_error (const char *msg)
		288	{
		289	perror (msg);
		290	abort ();
		291	}
		292
		293	...
		294	ev_set_syserr_cb (fatal_error);
		295
102	=back	296	=back
103		297
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	298	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
105		299
106	An event loop is described by a C<struct ev_loop *>. The library knows two	300	An event loop is described by a C<struct ev_loop *> (the C<struct> is
107	types of such loops, the I<default> loop, which supports signals and child	301	I<not> optional in this case unless libev 3 compatibility is disabled, as
108	events, and dynamically created loops which do not.	302	libev 3 had an C<ev_loop> function colliding with the struct name).
109		303
110	If you use threads, a common model is to run the default event loop	304	The library knows two types of such loops, the I<default> loop, which
111	in your main thread (or in a separate thread) and for each thread you	305	supports signals and child events, and dynamically created event loops
112	create, you also create another event loop. Libev itself does no locking	306	which do not.
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		307
117	=over 4	308	=over 4
118		309
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	310	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		311
121	This will initialise the default event loop if it hasn't been initialised	312	This returns the "default" event loop object, which is what you should
122	yet and return it. If the default loop could not be initialised, returns	313	normally use when you just need "the event loop". Event loop objects and
123	false. If it already was initialised it simply returns it (and ignores the	314	the C<flags> parameter are described in more detail in the entry for
124	flags).	315	C<ev_loop_new>.
		316
		317	If the default loop is already initialised then this function simply
		318	returns it (and ignores the flags. If that is troubling you, check
		319	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		320	flags, which should almost always be C<0>, unless the caller is also the
		321	one calling C<ev_run> or otherwise qualifies as "the main program".
125		322
126	If you don't know what event loop to use, use the one returned from this	323	If you don't know what event loop to use, use the one returned from this
127	function.	324	function (or via the C<EV_DEFAULT> macro).
		325
		326	Note that this function is I<not> thread-safe, so if you want to use it
		327	from multiple threads, you have to employ some kind of mutex (note also
		328	that this case is unlikely, as loops cannot be shared easily between
		329	threads anyway).
		330
		331	The default loop is the only loop that can handle C<ev_child> watchers,
		332	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		333	a problem for your application you can either create a dynamic loop with
		334	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		335	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		336
		337	Example: This is the most typical usage.
		338
		339	if (!ev_default_loop (0))
		340	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		341
		342	Example: Restrict libev to the select and poll backends, and do not allow
		343	environment settings to be taken into account:
		344
		345	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		346
		347	Example: Use whatever libev has to offer, but make sure that kqueue is
		348	used if available (warning, breaks stuff, best use only with your own
		349	private event loop and only if you know the OS supports your types of
		350	fds):
		351
		352	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
		353
		354	=item struct ev_loop *ev_loop_new (unsigned int flags)
		355
		356	This will create and initialise a new event loop object. If the loop
		357	could not be initialised, returns false.
		358
		359	Note that this function I<is> thread-safe, and one common way to use
		360	libev with threads is indeed to create one loop per thread, and using the
		361	default loop in the "main" or "initial" thread.
128		362
129	The flags argument can be used to specify special behaviour or specific	363	The flags argument can be used to specify special behaviour or specific
130	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	364	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
131		365
132	It supports the following flags:	366	The following flags are supported:
133		367
134	=over 4	368	=over 4
135		369
136	=item C<EVFLAG_AUTO>	370	=item C<EVFLAG_AUTO>
137		371
138	The default flags value. Use this if you have no clue (it's the right	372	The default flags value. Use this if you have no clue (it's the right
139	thing, believe me).	373	thing, believe me).
140		374
141	=item C<EVFLAG_NOENV>	375	=item C<EVFLAG_NOENV>
142		376
143	If this flag bit is ored into the flag value (or the program runs setuid	377	If this flag bit is or'ed into the flag value (or the program runs setuid
144	or setgid) then libev will I<not> look at the environment variable	378	or setgid) then libev will I<not> look at the environment variable
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	379	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	override the flags completely if it is found in the environment. This is	380	override the flags completely if it is found in the environment. This is
147	useful to try out specific backends to test their performance, or to work	381	useful to try out specific backends to test their performance, or to work
148	around bugs.	382	around bugs.
149		383
		384	=item C<EVFLAG_FORKCHECK>
		385
		386	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		387	make libev check for a fork in each iteration by enabling this flag.
		388
		389	This works by calling C<getpid ()> on every iteration of the loop,
		390	and thus this might slow down your event loop if you do a lot of loop
		391	iterations and little real work, but is usually not noticeable (on my
		392	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		393	without a system call and thus I<very> fast, but my GNU/Linux system also has
		394	C<pthread_atfork> which is even faster).
		395
		396	The big advantage of this flag is that you can forget about fork (and
		397	forget about forgetting to tell libev about forking) when you use this
		398	flag.
		399
		400	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		401	environment variable.
		402
		403	=item C<EVFLAG_NOINOTIFY>
		404
		405	When this flag is specified, then libev will not attempt to use the
		406	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		407	testing, this flag can be useful to conserve inotify file descriptors, as
		408	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		409
		410	=item C<EVFLAG_SIGNALFD>
		411
		412	When this flag is specified, then libev will attempt to use the
		413	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API
		414	delivers signals synchronously, which makes it both faster and might make
		415	it possible to get the queued signal data. It can also simplify signal
		416	handling with threads, as long as you properly block signals in your
		417	threads that are not interested in handling them.
		418
		419	Signalfd will not be used by default as this changes your signal mask, and
		420	there are a lot of shoddy libraries and programs (glib's threadpool for
		421	example) that can't properly initialise their signal masks.
		422
150	=item C<EVMETHOD_SELECT> (value 1, portable select backend)	423	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		424
152	This is your standard select(2) backend. Not I<completely> standard, as	425	This is your standard select(2) backend. Not I<completely> standard, as
153	libev tries to roll its own fd_set with no limits on the number of fds,	426	libev tries to roll its own fd_set with no limits on the number of fds,
154	but if that fails, expect a fairly low limit on the number of fds when	427	but if that fails, expect a fairly low limit on the number of fds when
155	using this backend. It doesn't scale too well (O(highest_fd)), but its usually	428	using this backend. It doesn't scale too well (O(highest_fd)), but its
156	the fastest backend for a low number of fds.	429	usually the fastest backend for a low number of (low-numbered :) fds.
157		430
		431	To get good performance out of this backend you need a high amount of
		432	parallelism (most of the file descriptors should be busy). If you are
		433	writing a server, you should C<accept ()> in a loop to accept as many
		434	connections as possible during one iteration. You might also want to have
		435	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		436	readiness notifications you get per iteration.
		437
		438	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		439	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		440	C<exceptfds> set on that platform).
		441
158	=item C<EVMETHOD_POLL> (value 2, poll backend, available everywhere except on windows)	442	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
159		443
160	And this is your standard poll(2) backend. It's more complicated than	444	And this is your standard poll(2) backend. It's more complicated
161	select, but handles sparse fds better and has no artificial limit on the	445	than select, but handles sparse fds better and has no artificial
162	number of fds you can use (except it will slow down considerably with a	446	limit on the number of fds you can use (except it will slow down
163	lot of inactive fds). It scales similarly to select, i.e. O(total_fds).	447	considerably with a lot of inactive fds). It scales similarly to select,
		448	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		449	performance tips.
164		450
		451	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		452	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
		453
165	=item C<EVMETHOD_EPOLL> (value 4, Linux)	454	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		455
		456	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		457	kernels).
166		458
167	For few fds, this backend is a bit little slower than poll and select,	459	For few fds, this backend is a bit little slower than poll and select,
168	but it scales phenomenally better. While poll and select usually scale like	460	but it scales phenomenally better. While poll and select usually scale
169	O(total_fds) where n is the total number of fds (or the highest fd), epoll scales	461	like O(total_fds) where n is the total number of fds (or the highest fd),
170	either O(1) or O(active_fds).	462	epoll scales either O(1) or O(active_fds).
171		463
		464	The epoll mechanism deserves honorable mention as the most misdesigned
		465	of the more advanced event mechanisms: mere annoyances include silently
		466	dropping file descriptors, requiring a system call per change per file
		467	descriptor (and unnecessary guessing of parameters), problems with dup and
		468	so on. The biggest issue is fork races, however - if a program forks then
		469	I<both> parent and child process have to recreate the epoll set, which can
		470	take considerable time (one syscall per file descriptor) and is of course
		471	hard to detect.
		472
		473	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		474	of course I<doesn't>, and epoll just loves to report events for totally
		475	I<different> file descriptors (even already closed ones, so one cannot
		476	even remove them from the set) than registered in the set (especially
		477	on SMP systems). Libev tries to counter these spurious notifications by
		478	employing an additional generation counter and comparing that against the
		479	events to filter out spurious ones, recreating the set when required. Last
		480	not least, it also refuses to work with some file descriptors which work
		481	perfectly fine with C<select> (files, many character devices...).
		482
172	While stopping and starting an I/O watcher in the same iteration will	483	While stopping, setting and starting an I/O watcher in the same iteration
173	result in some caching, there is still a syscall per such incident	484	will result in some caching, there is still a system call per such
174	(because the fd could point to a different file description now), so its	485	incident (because the same I<file descriptor> could point to a different
175	best to avoid that. Also, dup()ed file descriptors might not work very	486	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
176	well if you register events for both fds.	487	file descriptors might not work very well if you register events for both
		488	file descriptors.
177		489
		490	Best performance from this backend is achieved by not unregistering all
		491	watchers for a file descriptor until it has been closed, if possible,
		492	i.e. keep at least one watcher active per fd at all times. Stopping and
		493	starting a watcher (without re-setting it) also usually doesn't cause
		494	extra overhead. A fork can both result in spurious notifications as well
		495	as in libev having to destroy and recreate the epoll object, which can
		496	take considerable time and thus should be avoided.
		497
		498	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		499	faster than epoll for maybe up to a hundred file descriptors, depending on
		500	the usage. So sad.
		501
		502	While nominally embeddable in other event loops, this feature is broken in
		503	all kernel versions tested so far.
		504
		505	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		506	C<EVBACKEND_POLL>.
		507
178	=item C<EVMETHOD_KQUEUE> (value 8, most BSD clones)	508	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
179		509
180	Kqueue deserves special mention, as at the time of this writing, it	510	Kqueue deserves special mention, as at the time of this writing, it
181	was broken on all BSDs except NetBSD (usually it doesn't work with	511	was broken on all BSDs except NetBSD (usually it doesn't work reliably
182	anything but sockets and pipes, except on Darwin, where of course its	512	with anything but sockets and pipes, except on Darwin, where of course
183	completely useless). For this reason its not being "autodetected" unless	513	it's completely useless). Unlike epoll, however, whose brokenness
184	you explicitly specify the flags (i.e. you don't use EVFLAG_AUTO).	514	is by design, these kqueue bugs can (and eventually will) be fixed
		515	without API changes to existing programs. For this reason it's not being
		516	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		517	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		518	system like NetBSD.
		519
		520	You still can embed kqueue into a normal poll or select backend and use it
		521	only for sockets (after having made sure that sockets work with kqueue on
		522	the target platform). See C<ev_embed> watchers for more info.
185		523
186	It scales in the same way as the epoll backend, but the interface to the	524	It scales in the same way as the epoll backend, but the interface to the
187	kernel is more efficient (which says nothing about its actual speed, of	525	kernel is more efficient (which says nothing about its actual speed, of
188	course). While starting and stopping an I/O watcher does not cause an	526	course). While stopping, setting and starting an I/O watcher does never
189	extra syscall as with epoll, it still adds up to four event changes per	527	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
190	incident, so its best to avoid that.	528	two event changes per incident. Support for C<fork ()> is very bad (but
		529	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		530	cases
191		531
		532	This backend usually performs well under most conditions.
		533
		534	While nominally embeddable in other event loops, this doesn't work
		535	everywhere, so you might need to test for this. And since it is broken
		536	almost everywhere, you should only use it when you have a lot of sockets
		537	(for which it usually works), by embedding it into another event loop
		538	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		539	also broken on OS X)) and, did I mention it, using it only for sockets.
		540
		541	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		542	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		543	C<NOTE_EOF>.
		544
192	=item C<EVMETHOD_DEVPOLL> (value 16, Solaris 8)	545	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
193		546
194	This is not implemented yet (and might never be).	547	This is not implemented yet (and might never be, unless you send me an
		548	implementation). According to reports, C</dev/poll> only supports sockets
		549	and is not embeddable, which would limit the usefulness of this backend
		550	immensely.
195		551
196	=item C<EVMETHOD_PORT> (value 32, Solaris 10)	552	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
197		553
198	This uses the Solaris 10 port mechanism. As with everything on Solaris,	554	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
199	it's really slow, but it still scales very well (O(active_fds)).	555	it's really slow, but it still scales very well (O(active_fds)).
200		556
		557	Please note that Solaris event ports can deliver a lot of spurious
		558	notifications, so you need to use non-blocking I/O or other means to avoid
		559	blocking when no data (or space) is available.
		560
		561	While this backend scales well, it requires one system call per active
		562	file descriptor per loop iteration. For small and medium numbers of file
		563	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		564	might perform better.
		565
		566	On the positive side, with the exception of the spurious readiness
		567	notifications, this backend actually performed fully to specification
		568	in all tests and is fully embeddable, which is a rare feat among the
		569	OS-specific backends (I vastly prefer correctness over speed hacks).
		570
		571	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		572	C<EVBACKEND_POLL>.
		573
201	=item C<EVMETHOD_ALL>	574	=item C<EVBACKEND_ALL>
202		575
203	Try all backends (even potentially broken ones that wouldn't be tried	576	Try all backends (even potentially broken ones that wouldn't be tried
204	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	577	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
205	C<EVMETHOD_ALL & ~EVMETHOD_KQUEUE>.	578	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		579
		580	It is definitely not recommended to use this flag.
206		581
207	=back	582	=back
208		583
209	If one or more of these are ored into the flags value, then only these	584	If one or more of the backend flags are or'ed into the flags value,
210	backends will be tried (in the reverse order as given here). If none are	585	then only these backends will be tried (in the reverse order as listed
211	specified, most compiled-in backend will be tried, usually in reverse	586	here). If none are specified, all backends in C<ev_recommended_backends
212	order of their flag values :)	587	()> will be tried.
213		588
214	=item struct ev_loop *ev_loop_new (unsigned int flags)	589	Example: Try to create a event loop that uses epoll and nothing else.
215		590
216	Similar to C<ev_default_loop>, but always creates a new event loop that is	591	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
217	always distinct from the default loop. Unlike the default loop, it cannot	592	if (!epoller)
218	handle signal and child watchers, and attempts to do so will be greeted by	593	fatal ("no epoll found here, maybe it hides under your chair");
219	undefined behaviour (or a failed assertion if assertions are enabled).
220
221	=item ev_default_destroy ()
222
223	Destroys the default loop again (frees all memory and kernel state
224	etc.). This stops all registered event watchers (by not touching them in
225	any way whatsoever, although you cannot rely on this :).
226		594
227	=item ev_loop_destroy (loop)	595	=item ev_loop_destroy (loop)
228		596
229	Like C<ev_default_destroy>, but destroys an event loop created by an	597	Destroys an event loop object (frees all memory and kernel state
230	earlier call to C<ev_loop_new>.	598	etc.). None of the active event watchers will be stopped in the normal
		599	sense, so e.g. C<ev_is_active> might still return true. It is your
		600	responsibility to either stop all watchers cleanly yourself I<before>
		601	calling this function, or cope with the fact afterwards (which is usually
		602	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		603	for example).
231		604
232	=item ev_default_fork ()	605	Note that certain global state, such as signal state (and installed signal
		606	handlers), will not be freed by this function, and related watchers (such
		607	as signal and child watchers) would need to be stopped manually.
233		608
		609	This function is normally used on loop objects allocated by
		610	C<ev_loop_new>, but it can also be used on the default loop returned by
		611	C<ev_default_loop>, in which case it is not thread-safe.
		612
		613	Note that it is not advisable to call this function on the default loop
		614	except in the rare occasion where you really need to free it's resources.
		615	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		616	and C<ev_loop_destroy>.
		617
		618	=item ev_loop_fork (loop)
		619
		620	This function sets a flag that causes subsequent C<ev_run> iterations to
234	This function reinitialises the kernel state for backends that have	621	reinitialise the kernel state for backends that have one. Despite the
235	one. Despite the name, you can call it anytime, but it makes most sense	622	name, you can call it anytime, but it makes most sense after forking, in
236	after forking, in either the parent or child process (or both, but that	623	the child process. You I<must> call it (or use C<EVFLAG_FORKCHECK>) in the
237	again makes little sense).	624	child before resuming or calling C<ev_run>.
238		625
239	You I<must> call this function in the child process after forking if and	626	Again, you I<have> to call it on I<any> loop that you want to re-use after
240	only if you want to use the event library in both processes. If you just	627	a fork, I<even if you do not plan to use the loop in the parent>. This is
241	fork+exec, you don't have to call it.	628	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		629	during fork.
		630
		631	On the other hand, you only need to call this function in the child
		632	process if and only if you want to use the event loop in the child. If
		633	you just fork+exec or create a new loop in the child, you don't have to
		634	call it at all (in fact, C<epoll> is so badly broken that it makes a
		635	difference, but libev will usually detect this case on its own and do a
		636	costly reset of the backend).
242		637
243	The function itself is quite fast and it's usually not a problem to call	638	The function itself is quite fast and it's usually not a problem to call
244	it just in case after a fork. To make this easy, the function will fit in	639	it just in case after a fork.
245	quite nicely into a call to C<pthread_atfork>:
246		640
		641	Example: Automate calling C<ev_loop_fork> on the default loop when
		642	using pthreads.
		643
		644	static void
		645	post_fork_child (void)
		646	{
		647	ev_loop_fork (EV_DEFAULT);
		648	}
		649
		650	...
247	pthread_atfork (0, 0, ev_default_fork);	651	pthread_atfork (0, 0, post_fork_child);
248		652
249	=item ev_loop_fork (loop)	653	=item int ev_is_default_loop (loop)
250		654
251	Like C<ev_default_fork>, but acts on an event loop created by	655	Returns true when the given loop is, in fact, the default loop, and false
252	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	656	otherwise.
253	after fork, and how you do this is entirely your own problem.
254		657
		658	=item unsigned int ev_iteration (loop)
		659
		660	Returns the current iteration count for the event loop, which is identical
		661	to the number of times libev did poll for new events. It starts at C<0>
		662	and happily wraps around with enough iterations.
		663
		664	This value can sometimes be useful as a generation counter of sorts (it
		665	"ticks" the number of loop iterations), as it roughly corresponds with
		666	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		667	prepare and check phases.
		668
255	=item unsigned int ev_method (loop)	669	=item unsigned int ev_depth (loop)
256		670
		671	Returns the number of times C<ev_run> was entered minus the number of
		672	times C<ev_run> was exited, in other words, the recursion depth.
		673
		674	Outside C<ev_run>, this number is zero. In a callback, this number is
		675	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		676	in which case it is higher.
		677
		678	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread
		679	etc.), doesn't count as "exit" - consider this as a hint to avoid such
		680	ungentleman-like behaviour unless it's really convenient.
		681
		682	=item unsigned int ev_backend (loop)
		683
257	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	684	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
258	use.	685	use.
259		686
260	=item ev_tstamp ev_now (loop)	687	=item ev_tstamp ev_now (loop)
261		688
262	Returns the current "event loop time", which is the time the event loop	689	Returns the current "event loop time", which is the time the event loop
263	got events and started processing them. This timestamp does not change	690	received events and started processing them. This timestamp does not
264	as long as callbacks are being processed, and this is also the base time	691	change as long as callbacks are being processed, and this is also the base
265	used for relative timers. You can treat it as the timestamp of the event	692	time used for relative timers. You can treat it as the timestamp of the
266	occuring (or more correctly, the mainloop finding out about it).	693	event occurring (or more correctly, libev finding out about it).
267		694
		695	=item ev_now_update (loop)
		696
		697	Establishes the current time by querying the kernel, updating the time
		698	returned by C<ev_now ()> in the progress. This is a costly operation and
		699	is usually done automatically within C<ev_run ()>.
		700
		701	This function is rarely useful, but when some event callback runs for a
		702	very long time without entering the event loop, updating libev's idea of
		703	the current time is a good idea.
		704
		705	See also L<The special problem of time updates> in the C<ev_timer> section.
		706
		707	=item ev_suspend (loop)
		708
		709	=item ev_resume (loop)
		710
		711	These two functions suspend and resume an event loop, for use when the
		712	loop is not used for a while and timeouts should not be processed.
		713
		714	A typical use case would be an interactive program such as a game: When
		715	the user presses C<^Z> to suspend the game and resumes it an hour later it
		716	would be best to handle timeouts as if no time had actually passed while
		717	the program was suspended. This can be achieved by calling C<ev_suspend>
		718	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		719	C<ev_resume> directly afterwards to resume timer processing.
		720
		721	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		722	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		723	will be rescheduled (that is, they will lose any events that would have
		724	occurred while suspended).
		725
		726	After calling C<ev_suspend> you B<must not> call I<any> function on the
		727	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		728	without a previous call to C<ev_suspend>.
		729
		730	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		731	event loop time (see C<ev_now_update>).
		732
268	=item ev_loop (loop, int flags)	733	=item ev_run (loop, int flags)
269		734
270	Finally, this is it, the event handler. This function usually is called	735	Finally, this is it, the event handler. This function usually is called
271	after you initialised all your watchers and you want to start handling	736	after you have initialised all your watchers and you want to start
272	events.	737	handling events. It will ask the operating system for any new events, call
		738	the watcher callbacks, an then repeat the whole process indefinitely: This
		739	is why event loops are called I<loops>.
273		740
274	If the flags argument is specified as 0, it will not return until either	741	If the flags argument is specified as C<0>, it will keep handling events
275	no event watchers are active anymore or C<ev_unloop> was called.	742	until either no event watchers are active anymore or C<ev_break> was
		743	called.
276		744
		745	Please note that an explicit C<ev_break> is usually better than
		746	relying on all watchers to be stopped when deciding when a program has
		747	finished (especially in interactive programs), but having a program
		748	that automatically loops as long as it has to and no longer by virtue
		749	of relying on its watchers stopping correctly, that is truly a thing of
		750	beauty.
		751
277	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	752	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
278	those events and any outstanding ones, but will not block your process in	753	those events and any already outstanding ones, but will not wait and
279	case there are no events and will return after one iteration of the loop.	754	block your process in case there are no events and will return after one
		755	iteration of the loop. This is sometimes useful to poll and handle new
		756	events while doing lengthy calculations, to keep the program responsive.
280		757
281	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	758	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
282	neccessary) and will handle those and any outstanding ones. It will block	759	necessary) and will handle those and any already outstanding ones. It
283	your process until at least one new event arrives, and will return after	760	will block your process until at least one new event arrives (which could
		761	be an event internal to libev itself, so there is no guarantee that a
		762	user-registered callback will be called), and will return after one
284	one iteration of the loop.	763	iteration of the loop.
285		764
286	This flags value could be used to implement alternative looping	765	This is useful if you are waiting for some external event in conjunction
287	constructs, but the C<prepare> and C<check> watchers provide a better and	766	with something not expressible using other libev watchers (i.e. "roll your
288	more generic mechanism.	767	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		768	usually a better approach for this kind of thing.
289		769
290	Here are the gory details of what ev_loop does:	770	Here are the gory details of what C<ev_run> does:
291		771
292	1. If there are no active watchers (reference count is zero), return.	772	- Increment loop depth.
		773	- Reset the ev_break status.
		774	- Before the first iteration, call any pending watchers.
		775	LOOP:
		776	- If EVFLAG_FORKCHECK was used, check for a fork.
		777	- If a fork was detected (by any means), queue and call all fork watchers.
293	2. Queue and immediately call all prepare watchers.	778	- Queue and call all prepare watchers.
		779	- If ev_break was called, goto FINISH.
294	3. If we have been forked, recreate the kernel state.	780	- If we have been forked, detach and recreate the kernel state
		781	as to not disturb the other process.
295	4. Update the kernel state with all outstanding changes.	782	- Update the kernel state with all outstanding changes.
296	5. Update the "event loop time".	783	- Update the "event loop time" (ev_now ()).
297	6. Calculate for how long to block.	784	- Calculate for how long to sleep or block, if at all
		785	(active idle watchers, EVRUN_NOWAIT or not having
		786	any active watchers at all will result in not sleeping).
		787	- Sleep if the I/O and timer collect interval say so.
		788	- Increment loop iteration counter.
298	7. Block the process, waiting for events.	789	- Block the process, waiting for any events.
		790	- Queue all outstanding I/O (fd) events.
299	8. Update the "event loop time" and do time jump handling.	791	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
300	9. Queue all outstanding timers.	792	- Queue all expired timers.
301	10. Queue all outstanding periodics.	793	- Queue all expired periodics.
302	11. If no events are pending now, queue all idle watchers.	794	- Queue all idle watchers with priority higher than that of pending events.
303	12. Queue all check watchers.	795	- Queue all check watchers.
304	13. Call all queued watchers in reverse order (i.e. check watchers first).	796	- Call all queued watchers in reverse order (i.e. check watchers first).
305	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	797	Signals and child watchers are implemented as I/O watchers, and will
306	was used, return, otherwise continue with step #1.	798	be handled here by queueing them when their watcher gets executed.
		799	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		800	were used, or there are no active watchers, goto FINISH, otherwise
		801	continue with step LOOP.
		802	FINISH:
		803	- Reset the ev_break status iff it was EVBREAK_ONE.
		804	- Decrement the loop depth.
		805	- Return.
307		806
		807	Example: Queue some jobs and then loop until no events are outstanding
		808	anymore.
		809
		810	... queue jobs here, make sure they register event watchers as long
		811	... as they still have work to do (even an idle watcher will do..)
		812	ev_run (my_loop, 0);
		813	... jobs done or somebody called unloop. yeah!
		814
308	=item ev_unloop (loop, how)	815	=item ev_break (loop, how)
309		816
310	Can be used to make a call to C<ev_loop> return early (but only after it	817	Can be used to make a call to C<ev_run> return early (but only after it
311	has processed all outstanding events). The C<how> argument must be either	818	has processed all outstanding events). The C<how> argument must be either
312	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	819	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
313	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	820	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		821
		822	This "unloop state" will be cleared when entering C<ev_run> again.
		823
		824	It is safe to call C<ev_break> from outside any C<ev_run> calls. ##TODO##
314		825
315	=item ev_ref (loop)	826	=item ev_ref (loop)
316		827
317	=item ev_unref (loop)	828	=item ev_unref (loop)
318		829
319	Ref/unref can be used to add or remove a reference count on the event	830	Ref/unref can be used to add or remove a reference count on the event
320	loop: Every watcher keeps one reference, and as long as the reference	831	loop: Every watcher keeps one reference, and as long as the reference
321	count is nonzero, C<ev_loop> will not return on its own. If you have	832	count is nonzero, C<ev_run> will not return on its own.
322	a watcher you never unregister that should not keep C<ev_loop> from	833
323	returning, ev_unref() after starting, and ev_ref() before stopping it. For	834	This is useful when you have a watcher that you never intend to
		835	unregister, but that nevertheless should not keep C<ev_run> from
		836	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		837	before stopping it.
		838
324	example, libev itself uses this for its internal signal pipe: It is not	839	As an example, libev itself uses this for its internal signal pipe: It
325	visible to the libev user and should not keep C<ev_loop> from exiting if	840	is not visible to the libev user and should not keep C<ev_run> from
326	no event watchers registered by it are active. It is also an excellent	841	exiting if no event watchers registered by it are active. It is also an
327	way to do this for generic recurring timers or from within third-party	842	excellent way to do this for generic recurring timers or from within
328	libraries. Just remember to I<unref after start> and I<ref before stop>.	843	third-party libraries. Just remember to I<unref after start> and I<ref
		844	before stop> (but only if the watcher wasn't active before, or was active
		845	before, respectively. Note also that libev might stop watchers itself
		846	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		847	in the callback).
		848
		849	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		850	running when nothing else is active.
		851
		852	ev_signal exitsig;
		853	ev_signal_init (&exitsig, sig_cb, SIGINT);
		854	ev_signal_start (loop, &exitsig);
		855	evf_unref (loop);
		856
		857	Example: For some weird reason, unregister the above signal handler again.
		858
		859	ev_ref (loop);
		860	ev_signal_stop (loop, &exitsig);
		861
		862	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		863
		864	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		865
		866	These advanced functions influence the time that libev will spend waiting
		867	for events. Both time intervals are by default C<0>, meaning that libev
		868	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		869	latency.
		870
		871	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		872	allows libev to delay invocation of I/O and timer/periodic callbacks
		873	to increase efficiency of loop iterations (or to increase power-saving
		874	opportunities).
		875
		876	The idea is that sometimes your program runs just fast enough to handle
		877	one (or very few) event(s) per loop iteration. While this makes the
		878	program responsive, it also wastes a lot of CPU time to poll for new
		879	events, especially with backends like C<select ()> which have a high
		880	overhead for the actual polling but can deliver many events at once.
		881
		882	By setting a higher I<io collect interval> you allow libev to spend more
		883	time collecting I/O events, so you can handle more events per iteration,
		884	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		885	C<ev_timer>) will be not affected. Setting this to a non-null value will
		886	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		887	sleep time ensures that libev will not poll for I/O events more often then
		888	once per this interval, on average.
		889
		890	Likewise, by setting a higher I<timeout collect interval> you allow libev
		891	to spend more time collecting timeouts, at the expense of increased
		892	latency/jitter/inexactness (the watcher callback will be called
		893	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		894	value will not introduce any overhead in libev.
		895
		896	Many (busy) programs can usually benefit by setting the I/O collect
		897	interval to a value near C<0.1> or so, which is often enough for
		898	interactive servers (of course not for games), likewise for timeouts. It
		899	usually doesn't make much sense to set it to a lower value than C<0.01>,
		900	as this approaches the timing granularity of most systems. Note that if
		901	you do transactions with the outside world and you can't increase the
		902	parallelity, then this setting will limit your transaction rate (if you
		903	need to poll once per transaction and the I/O collect interval is 0.01,
		904	then you can't do more than 100 transactions per second).
		905
		906	Setting the I<timeout collect interval> can improve the opportunity for
		907	saving power, as the program will "bundle" timer callback invocations that
		908	are "near" in time together, by delaying some, thus reducing the number of
		909	times the process sleeps and wakes up again. Another useful technique to
		910	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		911	they fire on, say, one-second boundaries only.
		912
		913	Example: we only need 0.1s timeout granularity, and we wish not to poll
		914	more often than 100 times per second:
		915
		916	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		917	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		918
		919	=item ev_invoke_pending (loop)
		920
		921	This call will simply invoke all pending watchers while resetting their
		922	pending state. Normally, C<ev_run> does this automatically when required,
		923	but when overriding the invoke callback this call comes handy. This
		924	function can be invoked from a watcher - this can be useful for example
		925	when you want to do some lengthy calculation and want to pass further
		926	event handling to another thread (you still have to make sure only one
		927	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		928
		929	=item int ev_pending_count (loop)
		930
		931	Returns the number of pending watchers - zero indicates that no watchers
		932	are pending.
		933
		934	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		935
		936	This overrides the invoke pending functionality of the loop: Instead of
		937	invoking all pending watchers when there are any, C<ev_run> will call
		938	this callback instead. This is useful, for example, when you want to
		939	invoke the actual watchers inside another context (another thread etc.).
		940
		941	If you want to reset the callback, use C<ev_invoke_pending> as new
		942	callback.
		943
		944	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		945
		946	Sometimes you want to share the same loop between multiple threads. This
		947	can be done relatively simply by putting mutex_lock/unlock calls around
		948	each call to a libev function.
		949
		950	However, C<ev_run> can run an indefinite time, so it is not feasible
		951	to wait for it to return. One way around this is to wake up the event
		952	loop via C<ev_break> and C<av_async_send>, another way is to set these
		953	I<release> and I<acquire> callbacks on the loop.
		954
		955	When set, then C<release> will be called just before the thread is
		956	suspended waiting for new events, and C<acquire> is called just
		957	afterwards.
		958
		959	Ideally, C<release> will just call your mutex_unlock function, and
		960	C<acquire> will just call the mutex_lock function again.
		961
		962	While event loop modifications are allowed between invocations of
		963	C<release> and C<acquire> (that's their only purpose after all), no
		964	modifications done will affect the event loop, i.e. adding watchers will
		965	have no effect on the set of file descriptors being watched, or the time
		966	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		967	to take note of any changes you made.
		968
		969	In theory, threads executing C<ev_run> will be async-cancel safe between
		970	invocations of C<release> and C<acquire>.
		971
		972	See also the locking example in the C<THREADS> section later in this
		973	document.
		974
		975	=item ev_set_userdata (loop, void *data)
		976
		977	=item ev_userdata (loop)
		978
		979	Set and retrieve a single C<void *> associated with a loop. When
		980	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		981	C<0.>
		982
		983	These two functions can be used to associate arbitrary data with a loop,
		984	and are intended solely for the C<invoke_pending_cb>, C<release> and
		985	C<acquire> callbacks described above, but of course can be (ab-)used for
		986	any other purpose as well.
		987
		988	=item ev_verify (loop)
		989
		990	This function only does something when C<EV_VERIFY> support has been
		991	compiled in, which is the default for non-minimal builds. It tries to go
		992	through all internal structures and checks them for validity. If anything
		993	is found to be inconsistent, it will print an error message to standard
		994	error and call C<abort ()>.
		995
		996	This can be used to catch bugs inside libev itself: under normal
		997	circumstances, this function will never abort as of course libev keeps its
		998	data structures consistent.
329		999
330	=back	1000	=back
331		1001
		1002
332	=head1 ANATOMY OF A WATCHER	1003	=head1 ANATOMY OF A WATCHER
333		1004
		1005	In the following description, uppercase C<TYPE> in names stands for the
		1006	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1007	watchers and C<ev_io_start> for I/O watchers.
		1008
334	A watcher is a structure that you create and register to record your	1009	A watcher is an opaque structure that you allocate and register to record
335	interest in some event. For instance, if you want to wait for STDIN to	1010	your interest in some event. To make a concrete example, imagine you want
336	become readable, you would create an C<ev_io> watcher for that:	1011	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1012	for that:
337		1013
338	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1014	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
339	{	1015	{
340	ev_io_stop (w);	1016	ev_io_stop (w);
341	ev_unloop (loop, EVUNLOOP_ALL);	1017	ev_break (loop, EVBREAK_ALL);
342	}	1018	}
343		1019
344	struct ev_loop *loop = ev_default_loop (0);	1020	struct ev_loop *loop = ev_default_loop (0);
		1021
345	struct ev_io stdin_watcher;	1022	ev_io stdin_watcher;
		1023
346	ev_init (&stdin_watcher, my_cb);	1024	ev_init (&stdin_watcher, my_cb);
347	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1025	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
348	ev_io_start (loop, &stdin_watcher);	1026	ev_io_start (loop, &stdin_watcher);
		1027
349	ev_loop (loop, 0);	1028	ev_run (loop, 0);
350		1029
351	As you can see, you are responsible for allocating the memory for your	1030	As you can see, you are responsible for allocating the memory for your
352	watcher structures (and it is usually a bad idea to do this on the stack,	1031	watcher structures (and it is I<usually> a bad idea to do this on the
353	although this can sometimes be quite valid).	1032	stack).
354		1033
		1034	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1035	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1036
355	Each watcher structure must be initialised by a call to C<ev_init	1037	Each watcher structure must be initialised by a call to C<ev_init (watcher
356	(watcher *, callback)>, which expects a callback to be provided. This	1038	*, callback)>, which expects a callback to be provided. This callback is
357	callback gets invoked each time the event occurs (or, in the case of io	1039	invoked each time the event occurs (or, in the case of I/O watchers, each
358	watchers, each time the event loop detects that the file descriptor given	1040	time the event loop detects that the file descriptor given is readable
359	is readable and/or writable).	1041	and/or writable).
360		1042
361	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1043	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
362	with arguments specific to this watcher type. There is also a macro	1044	macro to configure it, with arguments specific to the watcher type. There
363	to combine initialisation and setting in one call: C<< ev_<type>_init	1045	is also a macro to combine initialisation and setting in one call: C<<
364	(watcher *, callback, ...) >>.	1046	ev_TYPE_init (watcher *, callback, ...) >>.
365		1047
366	To make the watcher actually watch out for events, you have to start it	1048	To make the watcher actually watch out for events, you have to start it
367	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1049	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
368	*) >>), and you can stop watching for events at any time by calling the	1050	*) >>), and you can stop watching for events at any time by calling the
369	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1051	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
370		1052
371	As long as your watcher is active (has been started but not stopped) you	1053	As long as your watcher is active (has been started but not stopped) you
372	must not touch the values stored in it. Most specifically you must never	1054	must not touch the values stored in it. Most specifically you must never
373	reinitialise it or call its set method.	1055	reinitialise it or call its C<ev_TYPE_set> macro.
374
375	You can check whether an event is active by calling the C<ev_is_active
376	(watcher *)> macro. To see whether an event is outstanding (but the
377	callback for it has not been called yet) you can use the C<ev_is_pending
378	(watcher *)> macro.
379		1056
380	Each and every callback receives the event loop pointer as first, the	1057	Each and every callback receives the event loop pointer as first, the
381	registered watcher structure as second, and a bitset of received events as	1058	registered watcher structure as second, and a bitset of received events as
382	third argument.	1059	third argument.
383		1060
…		…
392	=item C<EV_WRITE>	1069	=item C<EV_WRITE>
393		1070
394	The file descriptor in the C<ev_io> watcher has become readable and/or	1071	The file descriptor in the C<ev_io> watcher has become readable and/or
395	writable.	1072	writable.
396		1073
397	=item C<EV_TIMEOUT>	1074	=item C<EV_TIMER>
398		1075
399	The C<ev_timer> watcher has timed out.	1076	The C<ev_timer> watcher has timed out.
400		1077
401	=item C<EV_PERIODIC>	1078	=item C<EV_PERIODIC>
402		1079
…		…
408		1085
409	=item C<EV_CHILD>	1086	=item C<EV_CHILD>
410		1087
411	The pid specified in the C<ev_child> watcher has received a status change.	1088	The pid specified in the C<ev_child> watcher has received a status change.
412		1089
		1090	=item C<EV_STAT>
		1091
		1092	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1093
413	=item C<EV_IDLE>	1094	=item C<EV_IDLE>
414		1095
415	The C<ev_idle> watcher has determined that you have nothing better to do.	1096	The C<ev_idle> watcher has determined that you have nothing better to do.
416		1097
417	=item C<EV_PREPARE>	1098	=item C<EV_PREPARE>
418		1099
419	=item C<EV_CHECK>	1100	=item C<EV_CHECK>
420		1101
421	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1102	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts
422	to gather new events, and all C<ev_check> watchers are invoked just after	1103	to gather new events, and all C<ev_check> watchers are invoked just after
423	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1104	C<ev_run> has gathered them, but before it invokes any callbacks for any
424	received events. Callbacks of both watcher types can start and stop as	1105	received events. Callbacks of both watcher types can start and stop as
425	many watchers as they want, and all of them will be taken into account	1106	many watchers as they want, and all of them will be taken into account
426	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1107	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
427	C<ev_loop> from blocking).	1108	C<ev_run> from blocking).
		1109
		1110	=item C<EV_EMBED>
		1111
		1112	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1113
		1114	=item C<EV_FORK>
		1115
		1116	The event loop has been resumed in the child process after fork (see
		1117	C<ev_fork>).
		1118
		1119	=item C<EV_ASYNC>
		1120
		1121	The given async watcher has been asynchronously notified (see C<ev_async>).
		1122
		1123	=item C<EV_CUSTOM>
		1124
		1125	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1126	by libev users to signal watchers (e.g. via C<ev_feed_event>).
428		1127
429	=item C<EV_ERROR>	1128	=item C<EV_ERROR>
430		1129
431	An unspecified error has occured, the watcher has been stopped. This might	1130	An unspecified error has occurred, the watcher has been stopped. This might
432	happen because the watcher could not be properly started because libev	1131	happen because the watcher could not be properly started because libev
433	ran out of memory, a file descriptor was found to be closed or any other	1132	ran out of memory, a file descriptor was found to be closed or any other
		1133	problem. Libev considers these application bugs.
		1134
434	problem. You best act on it by reporting the problem and somehow coping	1135	You best act on it by reporting the problem and somehow coping with the
435	with the watcher being stopped.	1136	watcher being stopped. Note that well-written programs should not receive
		1137	an error ever, so when your watcher receives it, this usually indicates a
		1138	bug in your program.
436		1139
437	Libev will usually signal a few "dummy" events together with an error,	1140	Libev will usually signal a few "dummy" events together with an error, for
438	for example it might indicate that a fd is readable or writable, and if	1141	example it might indicate that a fd is readable or writable, and if your
439	your callbacks is well-written it can just attempt the operation and cope	1142	callbacks is well-written it can just attempt the operation and cope with
440	with the error from read() or write(). This will not work in multithreaded	1143	the error from read() or write(). This will not work in multi-threaded
441	programs, though, so beware.	1144	programs, though, as the fd could already be closed and reused for another
		1145	thing, so beware.
442		1146
443	=back	1147	=back
444		1148
		1149	=head2 WATCHER STATES
		1150
		1151	There are various watcher states mentioned throughout this manual -
		1152	active, pending and so on. In this section these states and the rules to
		1153	transition between them will be described in more detail - and while these
		1154	rules might look complicated, they usually do "the right thing".
		1155
		1156	=over 4
		1157
		1158	=item initialiased
		1159
		1160	Before a watcher can be registered with the event looop it has to be
		1161	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1162	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1163
		1164	In this state it is simply some block of memory that is suitable for use
		1165	in an event loop. It can be moved around, freed, reused etc. at will.
		1166
		1167	=item started/running/active
		1168
		1169	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1170	property of the event loop, and is actively waiting for events. While in
		1171	this state it cannot be accessed (except in a few documented ways), moved,
		1172	freed or anything else - the only legal thing is to keep a pointer to it,
		1173	and call libev functions on it that are documented to work on active watchers.
		1174
		1175	=item pending
		1176
		1177	If a watcher is active and libev determines that an event it is interested
		1178	in has occurred (such as a timer expiring), it will become pending. It will
		1179	stay in this pending state until either it is stopped or its callback is
		1180	about to be invoked, so it is not normally pending inside the watcher
		1181	callback.
		1182
		1183	The watcher might or might not be active while it is pending (for example,
		1184	an expired non-repeating timer can be pending but no longer active). If it
		1185	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1186	but it is still property of the event loop at this time, so cannot be
		1187	moved, freed or reused. And if it is active the rules described in the
		1188	previous item still apply.
		1189
		1190	It is also possible to feed an event on a watcher that is not active (e.g.
		1191	via C<ev_feed_event>), in which case it becomes pending without being
		1192	active.
		1193
		1194	=item stopped
		1195
		1196	A watcher can be stopped implicitly by libev (in which case it might still
		1197	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1198	latter will clear any pending state the watcher might be in, regardless
		1199	of whether it was active or not, so stopping a watcher explicitly before
		1200	freeing it is often a good idea.
		1201
		1202	While stopped (and not pending) the watcher is essentially in the
		1203	initialised state, that is it can be reused, moved, modified in any way
		1204	you wish.
		1205
		1206	=back
		1207
		1208	=head2 GENERIC WATCHER FUNCTIONS
		1209
		1210	=over 4
		1211
		1212	=item C<ev_init> (ev_TYPE *watcher, callback)
		1213
		1214	This macro initialises the generic portion of a watcher. The contents
		1215	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1216	the generic parts of the watcher are initialised, you I<need> to call
		1217	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1218	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1219	which rolls both calls into one.
		1220
		1221	You can reinitialise a watcher at any time as long as it has been stopped
		1222	(or never started) and there are no pending events outstanding.
		1223
		1224	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1225	int revents)>.
		1226
		1227	Example: Initialise an C<ev_io> watcher in two steps.
		1228
		1229	ev_io w;
		1230	ev_init (&w, my_cb);
		1231	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1232
		1233	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1234
		1235	This macro initialises the type-specific parts of a watcher. You need to
		1236	call C<ev_init> at least once before you call this macro, but you can
		1237	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1238	macro on a watcher that is active (it can be pending, however, which is a
		1239	difference to the C<ev_init> macro).
		1240
		1241	Although some watcher types do not have type-specific arguments
		1242	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1243
		1244	See C<ev_init>, above, for an example.
		1245
		1246	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1247
		1248	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1249	calls into a single call. This is the most convenient method to initialise
		1250	a watcher. The same limitations apply, of course.
		1251
		1252	Example: Initialise and set an C<ev_io> watcher in one step.
		1253
		1254	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1255
		1256	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1257
		1258	Starts (activates) the given watcher. Only active watchers will receive
		1259	events. If the watcher is already active nothing will happen.
		1260
		1261	Example: Start the C<ev_io> watcher that is being abused as example in this
		1262	whole section.
		1263
		1264	ev_io_start (EV_DEFAULT_UC, &w);
		1265
		1266	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1267
		1268	Stops the given watcher if active, and clears the pending status (whether
		1269	the watcher was active or not).
		1270
		1271	It is possible that stopped watchers are pending - for example,
		1272	non-repeating timers are being stopped when they become pending - but
		1273	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1274	pending. If you want to free or reuse the memory used by the watcher it is
		1275	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1276
		1277	=item bool ev_is_active (ev_TYPE *watcher)
		1278
		1279	Returns a true value iff the watcher is active (i.e. it has been started
		1280	and not yet been stopped). As long as a watcher is active you must not modify
		1281	it.
		1282
		1283	=item bool ev_is_pending (ev_TYPE *watcher)
		1284
		1285	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1286	events but its callback has not yet been invoked). As long as a watcher
		1287	is pending (but not active) you must not call an init function on it (but
		1288	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1289	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1290	it).
		1291
		1292	=item callback ev_cb (ev_TYPE *watcher)
		1293
		1294	Returns the callback currently set on the watcher.
		1295
		1296	=item ev_cb_set (ev_TYPE *watcher, callback)
		1297
		1298	Change the callback. You can change the callback at virtually any time
		1299	(modulo threads).
		1300
		1301	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1302
		1303	=item int ev_priority (ev_TYPE *watcher)
		1304
		1305	Set and query the priority of the watcher. The priority is a small
		1306	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1307	(default: C<-2>). Pending watchers with higher priority will be invoked
		1308	before watchers with lower priority, but priority will not keep watchers
		1309	from being executed (except for C<ev_idle> watchers).
		1310
		1311	If you need to suppress invocation when higher priority events are pending
		1312	you need to look at C<ev_idle> watchers, which provide this functionality.
		1313
		1314	You I<must not> change the priority of a watcher as long as it is active or
		1315	pending.
		1316
		1317	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1318	fine, as long as you do not mind that the priority value you query might
		1319	or might not have been clamped to the valid range.
		1320
		1321	The default priority used by watchers when no priority has been set is
		1322	always C<0>, which is supposed to not be too high and not be too low :).
		1323
		1324	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1325	priorities.
		1326
		1327	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1328
		1329	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1330	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1331	can deal with that fact, as both are simply passed through to the
		1332	callback.
		1333
		1334	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1335
		1336	If the watcher is pending, this function clears its pending status and
		1337	returns its C<revents> bitset (as if its callback was invoked). If the
		1338	watcher isn't pending it does nothing and returns C<0>.
		1339
		1340	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1341	callback to be invoked, which can be accomplished with this function.
		1342
		1343	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1344
		1345	Feeds the given event set into the event loop, as if the specified event
		1346	had happened for the specified watcher (which must be a pointer to an
		1347	initialised but not necessarily started event watcher). Obviously you must
		1348	not free the watcher as long as it has pending events.
		1349
		1350	Stopping the watcher, letting libev invoke it, or calling
		1351	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1352	not started in the first place.
		1353
		1354	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1355	functions that do not need a watcher.
		1356
		1357	=back
		1358
		1359
445	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1360	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
446		1361
447	Each watcher has, by default, a member C<void *data> that you can change	1362	Each watcher has, by default, a member C<void *data> that you can change
448	and read at any time, libev will completely ignore it. This can be used	1363	and read at any time: libev will completely ignore it. This can be used
449	to associate arbitrary data with your watcher. If you need more data and	1364	to associate arbitrary data with your watcher. If you need more data and
450	don't want to allocate memory and store a pointer to it in that data	1365	don't want to allocate memory and store a pointer to it in that data
451	member, you can also "subclass" the watcher type and provide your own	1366	member, you can also "subclass" the watcher type and provide your own
452	data:	1367	data:
453		1368
454	struct my_io	1369	struct my_io
455	{	1370	{
456	struct ev_io io;	1371	ev_io io;
457	int otherfd;	1372	int otherfd;
458	void *somedata;	1373	void *somedata;
459	struct whatever *mostinteresting;	1374	struct whatever *mostinteresting;
460	}	1375	};
		1376
		1377	...
		1378	struct my_io w;
		1379	ev_io_init (&w.io, my_cb, fd, EV_READ);
461		1380
462	And since your callback will be called with a pointer to the watcher, you	1381	And since your callback will be called with a pointer to the watcher, you
463	can cast it back to your own type:	1382	can cast it back to your own type:
464		1383
465	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1384	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
466	{	1385	{
467	struct my_io *w = (struct my_io *)w_;	1386	struct my_io *w = (struct my_io *)w_;
468	...	1387	...
469	}	1388	}
470		1389
471	More interesting and less C-conformant ways of catsing your callback type	1390	More interesting and less C-conformant ways of casting your callback type
472	have been omitted....	1391	instead have been omitted.
		1392
		1393	Another common scenario is to use some data structure with multiple
		1394	embedded watchers:
		1395
		1396	struct my_biggy
		1397	{
		1398	int some_data;
		1399	ev_timer t1;
		1400	ev_timer t2;
		1401	}
		1402
		1403	In this case getting the pointer to C<my_biggy> is a bit more
		1404	complicated: Either you store the address of your C<my_biggy> struct
		1405	in the C<data> member of the watcher (for woozies), or you need to use
		1406	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1407	programmers):
		1408
		1409	#include <stddef.h>
		1410
		1411	static void
		1412	t1_cb (EV_P_ ev_timer *w, int revents)
		1413	{
		1414	struct my_biggy big = (struct my_biggy *)
		1415	(((char *)w) - offsetof (struct my_biggy, t1));
		1416	}
		1417
		1418	static void
		1419	t2_cb (EV_P_ ev_timer *w, int revents)
		1420	{
		1421	struct my_biggy big = (struct my_biggy *)
		1422	(((char *)w) - offsetof (struct my_biggy, t2));
		1423	}
		1424
		1425	=head2 WATCHER PRIORITY MODELS
		1426
		1427	Many event loops support I<watcher priorities>, which are usually small
		1428	integers that influence the ordering of event callback invocation
		1429	between watchers in some way, all else being equal.
		1430
		1431	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1432	description for the more technical details such as the actual priority
		1433	range.
		1434
		1435	There are two common ways how these these priorities are being interpreted
		1436	by event loops:
		1437
		1438	In the more common lock-out model, higher priorities "lock out" invocation
		1439	of lower priority watchers, which means as long as higher priority
		1440	watchers receive events, lower priority watchers are not being invoked.
		1441
		1442	The less common only-for-ordering model uses priorities solely to order
		1443	callback invocation within a single event loop iteration: Higher priority
		1444	watchers are invoked before lower priority ones, but they all get invoked
		1445	before polling for new events.
		1446
		1447	Libev uses the second (only-for-ordering) model for all its watchers
		1448	except for idle watchers (which use the lock-out model).
		1449
		1450	The rationale behind this is that implementing the lock-out model for
		1451	watchers is not well supported by most kernel interfaces, and most event
		1452	libraries will just poll for the same events again and again as long as
		1453	their callbacks have not been executed, which is very inefficient in the
		1454	common case of one high-priority watcher locking out a mass of lower
		1455	priority ones.
		1456
		1457	Static (ordering) priorities are most useful when you have two or more
		1458	watchers handling the same resource: a typical usage example is having an
		1459	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1460	timeouts. Under load, data might be received while the program handles
		1461	other jobs, but since timers normally get invoked first, the timeout
		1462	handler will be executed before checking for data. In that case, giving
		1463	the timer a lower priority than the I/O watcher ensures that I/O will be
		1464	handled first even under adverse conditions (which is usually, but not
		1465	always, what you want).
		1466
		1467	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1468	will only be executed when no same or higher priority watchers have
		1469	received events, they can be used to implement the "lock-out" model when
		1470	required.
		1471
		1472	For example, to emulate how many other event libraries handle priorities,
		1473	you can associate an C<ev_idle> watcher to each such watcher, and in
		1474	the normal watcher callback, you just start the idle watcher. The real
		1475	processing is done in the idle watcher callback. This causes libev to
		1476	continuously poll and process kernel event data for the watcher, but when
		1477	the lock-out case is known to be rare (which in turn is rare :), this is
		1478	workable.
		1479
		1480	Usually, however, the lock-out model implemented that way will perform
		1481	miserably under the type of load it was designed to handle. In that case,
		1482	it might be preferable to stop the real watcher before starting the
		1483	idle watcher, so the kernel will not have to process the event in case
		1484	the actual processing will be delayed for considerable time.
		1485
		1486	Here is an example of an I/O watcher that should run at a strictly lower
		1487	priority than the default, and which should only process data when no
		1488	other events are pending:
		1489
		1490	ev_idle idle; // actual processing watcher
		1491	ev_io io; // actual event watcher
		1492
		1493	static void
		1494	io_cb (EV_P_ ev_io *w, int revents)
		1495	{
		1496	// stop the I/O watcher, we received the event, but
		1497	// are not yet ready to handle it.
		1498	ev_io_stop (EV_A_ w);
		1499
		1500	// start the idle watcher to handle the actual event.
		1501	// it will not be executed as long as other watchers
		1502	// with the default priority are receiving events.
		1503	ev_idle_start (EV_A_ &idle);
		1504	}
		1505
		1506	static void
		1507	idle_cb (EV_P_ ev_idle *w, int revents)
		1508	{
		1509	// actual processing
		1510	read (STDIN_FILENO, ...);
		1511
		1512	// have to start the I/O watcher again, as
		1513	// we have handled the event
		1514	ev_io_start (EV_P_ &io);
		1515	}
		1516
		1517	// initialisation
		1518	ev_idle_init (&idle, idle_cb);
		1519	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1520	ev_io_start (EV_DEFAULT_ &io);
		1521
		1522	In the "real" world, it might also be beneficial to start a timer, so that
		1523	low-priority connections can not be locked out forever under load. This
		1524	enables your program to keep a lower latency for important connections
		1525	during short periods of high load, while not completely locking out less
		1526	important ones.
473		1527
474		1528
475	=head1 WATCHER TYPES	1529	=head1 WATCHER TYPES
476		1530
477	This section describes each watcher in detail, but will not repeat	1531	This section describes each watcher in detail, but will not repeat
478	information given in the last section.	1532	information given in the last section. Any initialisation/set macros,
		1533	functions and members specific to the watcher type are explained.
479		1534
		1535	Members are additionally marked with either I<[read-only]>, meaning that,
		1536	while the watcher is active, you can look at the member and expect some
		1537	sensible content, but you must not modify it (you can modify it while the
		1538	watcher is stopped to your hearts content), or I<[read-write]>, which
		1539	means you can expect it to have some sensible content while the watcher
		1540	is active, but you can also modify it. Modifying it may not do something
		1541	sensible or take immediate effect (or do anything at all), but libev will
		1542	not crash or malfunction in any way.
		1543
		1544
480	=head2 C<ev_io> - is this file descriptor readable or writable	1545	=head2 C<ev_io> - is this file descriptor readable or writable?
481		1546
482	I/O watchers check whether a file descriptor is readable or writable	1547	I/O watchers check whether a file descriptor is readable or writable
483	in each iteration of the event loop (This behaviour is called	1548	in each iteration of the event loop, or, more precisely, when reading
484	level-triggering because you keep receiving events as long as the	1549	would not block the process and writing would at least be able to write
485	condition persists. Remember you can stop the watcher if you don't want to	1550	some data. This behaviour is called level-triggering because you keep
486	act on the event and neither want to receive future events).	1551	receiving events as long as the condition persists. Remember you can stop
		1552	the watcher if you don't want to act on the event and neither want to
		1553	receive future events.
487		1554
488	In general you can register as many read and/or write event watchers per	1555	In general you can register as many read and/or write event watchers per
489	fd as you want (as long as you don't confuse yourself). Setting all file	1556	fd as you want (as long as you don't confuse yourself). Setting all file
490	descriptors to non-blocking mode is also usually a good idea (but not	1557	descriptors to non-blocking mode is also usually a good idea (but not
491	required if you know what you are doing).	1558	required if you know what you are doing).
492		1559
493	You have to be careful with dup'ed file descriptors, though. Some backends	1560	If you cannot use non-blocking mode, then force the use of a
494	(the linux epoll backend is a notable example) cannot handle dup'ed file	1561	known-to-be-good backend (at the time of this writing, this includes only
495	descriptors correctly if you register interest in two or more fds pointing	1562	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
496	to the same underlying file/socket etc. description (that is, they share	1563	descriptors for which non-blocking operation makes no sense (such as
497	the same underlying "file open").	1564	files) - libev doesn't guarantee any specific behaviour in that case.
498		1565
499	If you must do this, then force the use of a known-to-be-good backend	1566	Another thing you have to watch out for is that it is quite easy to
500	(at the time of this writing, this includes only EVMETHOD_SELECT and	1567	receive "spurious" readiness notifications, that is your callback might
501	EVMETHOD_POLL).	1568	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1569	because there is no data. Not only are some backends known to create a
		1570	lot of those (for example Solaris ports), it is very easy to get into
		1571	this situation even with a relatively standard program structure. Thus
		1572	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1573	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1574
		1575	If you cannot run the fd in non-blocking mode (for example you should
		1576	not play around with an Xlib connection), then you have to separately
		1577	re-test whether a file descriptor is really ready with a known-to-be good
		1578	interface such as poll (fortunately in our Xlib example, Xlib already
		1579	does this on its own, so its quite safe to use). Some people additionally
		1580	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1581	indefinitely.
		1582
		1583	But really, best use non-blocking mode.
		1584
		1585	=head3 The special problem of disappearing file descriptors
		1586
		1587	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1588	descriptor (either due to calling C<close> explicitly or any other means,
		1589	such as C<dup2>). The reason is that you register interest in some file
		1590	descriptor, but when it goes away, the operating system will silently drop
		1591	this interest. If another file descriptor with the same number then is
		1592	registered with libev, there is no efficient way to see that this is, in
		1593	fact, a different file descriptor.
		1594
		1595	To avoid having to explicitly tell libev about such cases, libev follows
		1596	the following policy: Each time C<ev_io_set> is being called, libev
		1597	will assume that this is potentially a new file descriptor, otherwise
		1598	it is assumed that the file descriptor stays the same. That means that
		1599	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1600	descriptor even if the file descriptor number itself did not change.
		1601
		1602	This is how one would do it normally anyway, the important point is that
		1603	the libev application should not optimise around libev but should leave
		1604	optimisations to libev.
		1605
		1606	=head3 The special problem of dup'ed file descriptors
		1607
		1608	Some backends (e.g. epoll), cannot register events for file descriptors,
		1609	but only events for the underlying file descriptions. That means when you
		1610	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1611	events for them, only one file descriptor might actually receive events.
		1612
		1613	There is no workaround possible except not registering events
		1614	for potentially C<dup ()>'ed file descriptors, or to resort to
		1615	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1616
		1617	=head3 The special problem of fork
		1618
		1619	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1620	useless behaviour. Libev fully supports fork, but needs to be told about
		1621	it in the child.
		1622
		1623	To support fork in your programs, you either have to call
		1624	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1625	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1626	C<EVBACKEND_POLL>.
		1627
		1628	=head3 The special problem of SIGPIPE
		1629
		1630	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1631	when writing to a pipe whose other end has been closed, your program gets
		1632	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1633	this is sensible behaviour, for daemons, this is usually undesirable.
		1634
		1635	So when you encounter spurious, unexplained daemon exits, make sure you
		1636	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1637	somewhere, as that would have given you a big clue).
		1638
		1639	=head3 The special problem of accept()ing when you can't
		1640
		1641	Many implementations of the POSIX C<accept> function (for example,
		1642	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1643	connection from the pending queue in all error cases.
		1644
		1645	For example, larger servers often run out of file descriptors (because
		1646	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1647	rejecting the connection, leading to libev signalling readiness on
		1648	the next iteration again (the connection still exists after all), and
		1649	typically causing the program to loop at 100% CPU usage.
		1650
		1651	Unfortunately, the set of errors that cause this issue differs between
		1652	operating systems, there is usually little the app can do to remedy the
		1653	situation, and no known thread-safe method of removing the connection to
		1654	cope with overload is known (to me).
		1655
		1656	One of the easiest ways to handle this situation is to just ignore it
		1657	- when the program encounters an overload, it will just loop until the
		1658	situation is over. While this is a form of busy waiting, no OS offers an
		1659	event-based way to handle this situation, so it's the best one can do.
		1660
		1661	A better way to handle the situation is to log any errors other than
		1662	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1663	messages, and continue as usual, which at least gives the user an idea of
		1664	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1665	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1666	usage.
		1667
		1668	If your program is single-threaded, then you could also keep a dummy file
		1669	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1670	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1671	close that fd, and create a new dummy fd. This will gracefully refuse
		1672	clients under typical overload conditions.
		1673
		1674	The last way to handle it is to simply log the error and C<exit>, as
		1675	is often done with C<malloc> failures, but this results in an easy
		1676	opportunity for a DoS attack.
		1677
		1678	=head3 Watcher-Specific Functions
502		1679
503	=over 4	1680	=over 4
504		1681
505	=item ev_io_init (ev_io *, callback, int fd, int events)	1682	=item ev_io_init (ev_io *, callback, int fd, int events)
506		1683
507	=item ev_io_set (ev_io *, int fd, int events)	1684	=item ev_io_set (ev_io *, int fd, int events)
508		1685
509	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1686	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
510	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1687	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
511	EV_WRITE> to receive the given events.	1688	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1689
		1690	=item int fd [read-only]
		1691
		1692	The file descriptor being watched.
		1693
		1694	=item int events [read-only]
		1695
		1696	The events being watched.
512		1697
513	=back	1698	=back
514		1699
		1700	=head3 Examples
		1701
		1702	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1703	readable, but only once. Since it is likely line-buffered, you could
		1704	attempt to read a whole line in the callback.
		1705
		1706	static void
		1707	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1708	{
		1709	ev_io_stop (loop, w);
		1710	.. read from stdin here (or from w->fd) and handle any I/O errors
		1711	}
		1712
		1713	...
		1714	struct ev_loop *loop = ev_default_init (0);
		1715	ev_io stdin_readable;
		1716	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1717	ev_io_start (loop, &stdin_readable);
		1718	ev_run (loop, 0);
		1719
		1720
515	=head2 C<ev_timer> - relative and optionally recurring timeouts	1721	=head2 C<ev_timer> - relative and optionally repeating timeouts
516		1722
517	Timer watchers are simple relative timers that generate an event after a	1723	Timer watchers are simple relative timers that generate an event after a
518	given time, and optionally repeating in regular intervals after that.	1724	given time, and optionally repeating in regular intervals after that.
519		1725
520	The timers are based on real time, that is, if you register an event that	1726	The timers are based on real time, that is, if you register an event that
521	times out after an hour and you reset your system clock to last years	1727	times out after an hour and you reset your system clock to January last
522	time, it will still time out after (roughly) and hour. "Roughly" because	1728	year, it will still time out after (roughly) one hour. "Roughly" because
523	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1729	detecting time jumps is hard, and some inaccuracies are unavoidable (the
524	monotonic clock option helps a lot here).	1730	monotonic clock option helps a lot here).
		1731
		1732	The callback is guaranteed to be invoked only I<after> its timeout has
		1733	passed (not I<at>, so on systems with very low-resolution clocks this
		1734	might introduce a small delay). If multiple timers become ready during the
		1735	same loop iteration then the ones with earlier time-out values are invoked
		1736	before ones of the same priority with later time-out values (but this is
		1737	no longer true when a callback calls C<ev_run> recursively).
		1738
		1739	=head3 Be smart about timeouts
		1740
		1741	Many real-world problems involve some kind of timeout, usually for error
		1742	recovery. A typical example is an HTTP request - if the other side hangs,
		1743	you want to raise some error after a while.
		1744
		1745	What follows are some ways to handle this problem, from obvious and
		1746	inefficient to smart and efficient.
		1747
		1748	In the following, a 60 second activity timeout is assumed - a timeout that
		1749	gets reset to 60 seconds each time there is activity (e.g. each time some
		1750	data or other life sign was received).
		1751
		1752	=over 4
		1753
		1754	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1755
		1756	This is the most obvious, but not the most simple way: In the beginning,
		1757	start the watcher:
		1758
		1759	ev_timer_init (timer, callback, 60., 0.);
		1760	ev_timer_start (loop, timer);
		1761
		1762	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1763	and start it again:
		1764
		1765	ev_timer_stop (loop, timer);
		1766	ev_timer_set (timer, 60., 0.);
		1767	ev_timer_start (loop, timer);
		1768
		1769	This is relatively simple to implement, but means that each time there is
		1770	some activity, libev will first have to remove the timer from its internal
		1771	data structure and then add it again. Libev tries to be fast, but it's
		1772	still not a constant-time operation.
		1773
		1774	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1775
		1776	This is the easiest way, and involves using C<ev_timer_again> instead of
		1777	C<ev_timer_start>.
		1778
		1779	To implement this, configure an C<ev_timer> with a C<repeat> value
		1780	of C<60> and then call C<ev_timer_again> at start and each time you
		1781	successfully read or write some data. If you go into an idle state where
		1782	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1783	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1784
		1785	That means you can ignore both the C<ev_timer_start> function and the
		1786	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1787	member and C<ev_timer_again>.
		1788
		1789	At start:
		1790
		1791	ev_init (timer, callback);
		1792	timer->repeat = 60.;
		1793	ev_timer_again (loop, timer);
		1794
		1795	Each time there is some activity:
		1796
		1797	ev_timer_again (loop, timer);
		1798
		1799	It is even possible to change the time-out on the fly, regardless of
		1800	whether the watcher is active or not:
		1801
		1802	timer->repeat = 30.;
		1803	ev_timer_again (loop, timer);
		1804
		1805	This is slightly more efficient then stopping/starting the timer each time
		1806	you want to modify its timeout value, as libev does not have to completely
		1807	remove and re-insert the timer from/into its internal data structure.
		1808
		1809	It is, however, even simpler than the "obvious" way to do it.
		1810
		1811	=item 3. Let the timer time out, but then re-arm it as required.
		1812
		1813	This method is more tricky, but usually most efficient: Most timeouts are
		1814	relatively long compared to the intervals between other activity - in
		1815	our example, within 60 seconds, there are usually many I/O events with
		1816	associated activity resets.
		1817
		1818	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1819	but remember the time of last activity, and check for a real timeout only
		1820	within the callback:
		1821
		1822	ev_tstamp last_activity; // time of last activity
		1823
		1824	static void
		1825	callback (EV_P_ ev_timer *w, int revents)
		1826	{
		1827	ev_tstamp now = ev_now (EV_A);
		1828	ev_tstamp timeout = last_activity + 60.;
		1829
		1830	// if last_activity + 60. is older than now, we did time out
		1831	if (timeout < now)
		1832	{
		1833	// timeout occurred, take action
		1834	}
		1835	else
		1836	{
		1837	// callback was invoked, but there was some activity, re-arm
		1838	// the watcher to fire in last_activity + 60, which is
		1839	// guaranteed to be in the future, so "again" is positive:
		1840	w->repeat = timeout - now;
		1841	ev_timer_again (EV_A_ w);
		1842	}
		1843	}
		1844
		1845	To summarise the callback: first calculate the real timeout (defined
		1846	as "60 seconds after the last activity"), then check if that time has
		1847	been reached, which means something I<did>, in fact, time out. Otherwise
		1848	the callback was invoked too early (C<timeout> is in the future), so
		1849	re-schedule the timer to fire at that future time, to see if maybe we have
		1850	a timeout then.
		1851
		1852	Note how C<ev_timer_again> is used, taking advantage of the
		1853	C<ev_timer_again> optimisation when the timer is already running.
		1854
		1855	This scheme causes more callback invocations (about one every 60 seconds
		1856	minus half the average time between activity), but virtually no calls to
		1857	libev to change the timeout.
		1858
		1859	To start the timer, simply initialise the watcher and set C<last_activity>
		1860	to the current time (meaning we just have some activity :), then call the
		1861	callback, which will "do the right thing" and start the timer:
		1862
		1863	ev_init (timer, callback);
		1864	last_activity = ev_now (loop);
		1865	callback (loop, timer, EV_TIMER);
		1866
		1867	And when there is some activity, simply store the current time in
		1868	C<last_activity>, no libev calls at all:
		1869
		1870	last_activity = ev_now (loop);
		1871
		1872	This technique is slightly more complex, but in most cases where the
		1873	time-out is unlikely to be triggered, much more efficient.
		1874
		1875	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1876	callback :) - just change the timeout and invoke the callback, which will
		1877	fix things for you.
		1878
		1879	=item 4. Wee, just use a double-linked list for your timeouts.
		1880
		1881	If there is not one request, but many thousands (millions...), all
		1882	employing some kind of timeout with the same timeout value, then one can
		1883	do even better:
		1884
		1885	When starting the timeout, calculate the timeout value and put the timeout
		1886	at the I<end> of the list.
		1887
		1888	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1889	the list is expected to fire (for example, using the technique #3).
		1890
		1891	When there is some activity, remove the timer from the list, recalculate
		1892	the timeout, append it to the end of the list again, and make sure to
		1893	update the C<ev_timer> if it was taken from the beginning of the list.
		1894
		1895	This way, one can manage an unlimited number of timeouts in O(1) time for
		1896	starting, stopping and updating the timers, at the expense of a major
		1897	complication, and having to use a constant timeout. The constant timeout
		1898	ensures that the list stays sorted.
		1899
		1900	=back
		1901
		1902	So which method the best?
		1903
		1904	Method #2 is a simple no-brain-required solution that is adequate in most
		1905	situations. Method #3 requires a bit more thinking, but handles many cases
		1906	better, and isn't very complicated either. In most case, choosing either
		1907	one is fine, with #3 being better in typical situations.
		1908
		1909	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1910	rather complicated, but extremely efficient, something that really pays
		1911	off after the first million or so of active timers, i.e. it's usually
		1912	overkill :)
		1913
		1914	=head3 The special problem of time updates
		1915
		1916	Establishing the current time is a costly operation (it usually takes at
		1917	least two system calls): EV therefore updates its idea of the current
		1918	time only before and after C<ev_run> collects new events, which causes a
		1919	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1920	lots of events in one iteration.
525		1921
526	The relative timeouts are calculated relative to the C<ev_now ()>	1922	The relative timeouts are calculated relative to the C<ev_now ()>
527	time. This is usually the right thing as this timestamp refers to the time	1923	time. This is usually the right thing as this timestamp refers to the time
528	of the event triggering whatever timeout you are modifying/starting. If	1924	of the event triggering whatever timeout you are modifying/starting. If
529	you suspect event processing to be delayed and you I<need> to base the timeout	1925	you suspect event processing to be delayed and you I<need> to base the
530	on the current time, use something like this to adjust for this:	1926	timeout on the current time, use something like this to adjust for this:
531		1927
532	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1928	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
533		1929
534	The callback is guarenteed to be invoked only when its timeout has passed,	1930	If the event loop is suspended for a long time, you can also force an
535	but if multiple timers become ready during the same loop iteration then	1931	update of the time returned by C<ev_now ()> by calling C<ev_now_update
536	order of execution is undefined.	1932	()>.
		1933
		1934	=head3 The special problems of suspended animation
		1935
		1936	When you leave the server world it is quite customary to hit machines that
		1937	can suspend/hibernate - what happens to the clocks during such a suspend?
		1938
		1939	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1940	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1941	to run until the system is suspended, but they will not advance while the
		1942	system is suspended. That means, on resume, it will be as if the program
		1943	was frozen for a few seconds, but the suspend time will not be counted
		1944	towards C<ev_timer> when a monotonic clock source is used. The real time
		1945	clock advanced as expected, but if it is used as sole clocksource, then a
		1946	long suspend would be detected as a time jump by libev, and timers would
		1947	be adjusted accordingly.
		1948
		1949	I would not be surprised to see different behaviour in different between
		1950	operating systems, OS versions or even different hardware.
		1951
		1952	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1953	time jump in the monotonic clocks and the realtime clock. If the program
		1954	is suspended for a very long time, and monotonic clock sources are in use,
		1955	then you can expect C<ev_timer>s to expire as the full suspension time
		1956	will be counted towards the timers. When no monotonic clock source is in
		1957	use, then libev will again assume a timejump and adjust accordingly.
		1958
		1959	It might be beneficial for this latter case to call C<ev_suspend>
		1960	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1961	deterministic behaviour in this case (you can do nothing against
		1962	C<SIGSTOP>).
		1963
		1964	=head3 Watcher-Specific Functions and Data Members
537		1965
538	=over 4	1966	=over 4
539		1967
540	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1968	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
541		1969
542	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1970	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
543		1971
544	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1972	Configure the timer to trigger after C<after> seconds. If C<repeat>
545	C<0.>, then it will automatically be stopped. If it is positive, then the	1973	is C<0.>, then it will automatically be stopped once the timeout is
546	timer will automatically be configured to trigger again C<repeat> seconds	1974	reached. If it is positive, then the timer will automatically be
547	later, again, and again, until stopped manually.	1975	configured to trigger again C<repeat> seconds later, again, and again,
		1976	until stopped manually.
548		1977
549	The timer itself will do a best-effort at avoiding drift, that is, if you	1978	The timer itself will do a best-effort at avoiding drift, that is, if
550	configure a timer to trigger every 10 seconds, then it will trigger at	1979	you configure a timer to trigger every 10 seconds, then it will normally
551	exactly 10 second intervals. If, however, your program cannot keep up with	1980	trigger at exactly 10 second intervals. If, however, your program cannot
552	the timer (because it takes longer than those 10 seconds to do stuff) the	1981	keep up with the timer (because it takes longer than those 10 seconds to
553	timer will not fire more than once per event loop iteration.	1982	do stuff) the timer will not fire more than once per event loop iteration.
554		1983
555	=item ev_timer_again (loop)	1984	=item ev_timer_again (loop, ev_timer *)
556		1985
557	This will act as if the timer timed out and restart it again if it is	1986	This will act as if the timer timed out and restart it again if it is
558	repeating. The exact semantics are:	1987	repeating. The exact semantics are:
559		1988
		1989	If the timer is pending, its pending status is cleared.
		1990
560	If the timer is started but nonrepeating, stop it.	1991	If the timer is started but non-repeating, stop it (as if it timed out).
561		1992
562	If the timer is repeating, either start it if necessary (with the repeat	1993	If the timer is repeating, either start it if necessary (with the
563	value), or reset the running timer to the repeat value.	1994	C<repeat> value), or reset the running timer to the C<repeat> value.
564		1995
565	This sounds a bit complicated, but here is a useful and typical	1996	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
566	example: Imagine you have a tcp connection and you want a so-called idle	1997	usage example.
567	timeout, that is, you want to be called when there have been, say, 60	1998
568	seconds of inactivity on the socket. The easiest way to do this is to	1999	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
569	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	2000
570	time you successfully read or write some data. If you go into an idle	2001	Returns the remaining time until a timer fires. If the timer is active,
571	state where you do not expect data to travel on the socket, you can stop	2002	then this time is relative to the current event loop time, otherwise it's
572	the timer, and again will automatically restart it if need be.	2003	the timeout value currently configured.
		2004
		2005	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2006	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2007	will return C<4>. When the timer expires and is restarted, it will return
		2008	roughly C<7> (likely slightly less as callback invocation takes some time,
		2009	too), and so on.
		2010
		2011	=item ev_tstamp repeat [read-write]
		2012
		2013	The current C<repeat> value. Will be used each time the watcher times out
		2014	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2015	which is also when any modifications are taken into account.
573		2016
574	=back	2017	=back
575		2018
		2019	=head3 Examples
		2020
		2021	Example: Create a timer that fires after 60 seconds.
		2022
		2023	static void
		2024	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2025	{
		2026	.. one minute over, w is actually stopped right here
		2027	}
		2028
		2029	ev_timer mytimer;
		2030	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2031	ev_timer_start (loop, &mytimer);
		2032
		2033	Example: Create a timeout timer that times out after 10 seconds of
		2034	inactivity.
		2035
		2036	static void
		2037	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2038	{
		2039	.. ten seconds without any activity
		2040	}
		2041
		2042	ev_timer mytimer;
		2043	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2044	ev_timer_again (&mytimer); /* start timer */
		2045	ev_run (loop, 0);
		2046
		2047	// and in some piece of code that gets executed on any "activity":
		2048	// reset the timeout to start ticking again at 10 seconds
		2049	ev_timer_again (&mytimer);
		2050
		2051
576	=head2 C<ev_periodic> - to cron or not to cron	2052	=head2 C<ev_periodic> - to cron or not to cron?
577		2053
578	Periodic watchers are also timers of a kind, but they are very versatile	2054	Periodic watchers are also timers of a kind, but they are very versatile
579	(and unfortunately a bit complex).	2055	(and unfortunately a bit complex).
580		2056
581	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2057	Unlike C<ev_timer>, periodic watchers are not based on real time (or
582	but on wallclock time (absolute time). You can tell a periodic watcher	2058	relative time, the physical time that passes) but on wall clock time
583	to trigger "at" some specific point in time. For example, if you tell a	2059	(absolute time, the thing you can read on your calender or clock). The
584	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2060	difference is that wall clock time can run faster or slower than real
585	+ 10.>) and then reset your system clock to the last year, then it will	2061	time, and time jumps are not uncommon (e.g. when you adjust your
586	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2062	wrist-watch).
587	roughly 10 seconds later and of course not if you reset your system time
588	again).
589		2063
590	They can also be used to implement vastly more complex timers, such as	2064	You can tell a periodic watcher to trigger after some specific point
		2065	in time: for example, if you tell a periodic watcher to trigger "in 10
		2066	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2067	not a delay) and then reset your system clock to January of the previous
		2068	year, then it will take a year or more to trigger the event (unlike an
		2069	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2070	it, as it uses a relative timeout).
		2071
		2072	C<ev_periodic> watchers can also be used to implement vastly more complex
591	triggering an event on eahc midnight, local time.	2073	timers, such as triggering an event on each "midnight, local time", or
		2074	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		2075	those cannot react to time jumps.
592		2076
593	As with timers, the callback is guarenteed to be invoked only when the	2077	As with timers, the callback is guaranteed to be invoked only when the
594	time (C<at>) has been passed, but if multiple periodic timers become ready	2078	point in time where it is supposed to trigger has passed. If multiple
595	during the same loop iteration then order of execution is undefined.	2079	timers become ready during the same loop iteration then the ones with
		2080	earlier time-out values are invoked before ones with later time-out values
		2081	(but this is no longer true when a callback calls C<ev_run> recursively).
		2082
		2083	=head3 Watcher-Specific Functions and Data Members
596		2084
597	=over 4	2085	=over 4
598		2086
599	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2087	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
600		2088
601	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2089	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
602		2090
603	Lots of arguments, lets sort it out... There are basically three modes of	2091	Lots of arguments, let's sort it out... There are basically three modes of
604	operation, and we will explain them from simplest to complex:	2092	operation, and we will explain them from simplest to most complex:
605		2093
606	=over 4	2094	=over 4
607		2095
608	=item * absolute timer (interval = reschedule_cb = 0)	2096	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
609		2097
610	In this configuration the watcher triggers an event at the wallclock time	2098	In this configuration the watcher triggers an event after the wall clock
611	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2099	time C<offset> has passed. It will not repeat and will not adjust when a
612	that is, if it is to be run at January 1st 2011 then it will run when the	2100	time jump occurs, that is, if it is to be run at January 1st 2011 then it
613	system time reaches or surpasses this time.	2101	will be stopped and invoked when the system clock reaches or surpasses
		2102	this point in time.
614		2103
615	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2104	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
616		2105
617	In this mode the watcher will always be scheduled to time out at the next	2106	In this mode the watcher will always be scheduled to time out at the next
618	C<at + N * interval> time (for some integer N) and then repeat, regardless	2107	C<offset + N * interval> time (for some integer N, which can also be
619	of any time jumps.	2108	negative) and then repeat, regardless of any time jumps. The C<offset>
		2109	argument is merely an offset into the C<interval> periods.
620		2110
621	This can be used to create timers that do not drift with respect to system	2111	This can be used to create timers that do not drift with respect to the
622	time:	2112	system clock, for example, here is an C<ev_periodic> that triggers each
		2113	hour, on the hour (with respect to UTC):
623		2114
624	ev_periodic_set (&periodic, 0., 3600., 0);	2115	ev_periodic_set (&periodic, 0., 3600., 0);
625		2116
626	This doesn't mean there will always be 3600 seconds in between triggers,	2117	This doesn't mean there will always be 3600 seconds in between triggers,
627	but only that the the callback will be called when the system time shows a	2118	but only that the callback will be called when the system time shows a
628	full hour (UTC), or more correctly, when the system time is evenly divisible	2119	full hour (UTC), or more correctly, when the system time is evenly divisible
629	by 3600.	2120	by 3600.
630		2121
631	Another way to think about it (for the mathematically inclined) is that	2122	Another way to think about it (for the mathematically inclined) is that
632	C<ev_periodic> will try to run the callback in this mode at the next possible	2123	C<ev_periodic> will try to run the callback in this mode at the next possible
633	time where C<time = at (mod interval)>, regardless of any time jumps.	2124	time where C<time = offset (mod interval)>, regardless of any time jumps.
634		2125
		2126	For numerical stability it is preferable that the C<offset> value is near
		2127	C<ev_now ()> (the current time), but there is no range requirement for
		2128	this value, and in fact is often specified as zero.
		2129
		2130	Note also that there is an upper limit to how often a timer can fire (CPU
		2131	speed for example), so if C<interval> is very small then timing stability
		2132	will of course deteriorate. Libev itself tries to be exact to be about one
		2133	millisecond (if the OS supports it and the machine is fast enough).
		2134
635	=item * manual reschedule mode (reschedule_cb = callback)	2135	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
636		2136
637	In this mode the values for C<interval> and C<at> are both being	2137	In this mode the values for C<interval> and C<offset> are both being
638	ignored. Instead, each time the periodic watcher gets scheduled, the	2138	ignored. Instead, each time the periodic watcher gets scheduled, the
639	reschedule callback will be called with the watcher as first, and the	2139	reschedule callback will be called with the watcher as first, and the
640	current time as second argument.	2140	current time as second argument.
641		2141
642	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2142	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
643	ever, or make any event loop modifications>. If you need to stop it,	2143	or make ANY other event loop modifications whatsoever, unless explicitly
644	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2144	allowed by documentation here>.
645	starting a prepare watcher).
646		2145
		2146	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2147	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2148	only event loop modification you are allowed to do).
		2149
647	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2150	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
648	ev_tstamp now)>, e.g.:	2151	*w, ev_tstamp now)>, e.g.:
649		2152
		2153	static ev_tstamp
650	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2154	my_rescheduler (ev_periodic *w, ev_tstamp now)
651	{	2155	{
652	return now + 60.;	2156	return now + 60.;
653	}	2157	}
654		2158
655	It must return the next time to trigger, based on the passed time value	2159	It must return the next time to trigger, based on the passed time value
656	(that is, the lowest time value larger than to the second argument). It	2160	(that is, the lowest time value larger than to the second argument). It
657	will usually be called just before the callback will be triggered, but	2161	will usually be called just before the callback will be triggered, but
658	might be called at other times, too.	2162	might be called at other times, too.
659		2163
660	NOTE: I<< This callback must always return a time that is later than the	2164	NOTE: I<< This callback must always return a time that is higher than or
661	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2165	equal to the passed C<now> value >>.
662		2166
663	This can be used to create very complex timers, such as a timer that	2167	This can be used to create very complex timers, such as a timer that
664	triggers on each midnight, local time. To do this, you would calculate the	2168	triggers on "next midnight, local time". To do this, you would calculate the
665	next midnight after C<now> and return the timestamp value for this. How	2169	next midnight after C<now> and return the timestamp value for this. How
666	you do this is, again, up to you (but it is not trivial, which is the main	2170	you do this is, again, up to you (but it is not trivial, which is the main
667	reason I omitted it as an example).	2171	reason I omitted it as an example).
668		2172
669	=back	2173	=back
…		…
673	Simply stops and restarts the periodic watcher again. This is only useful	2177	Simply stops and restarts the periodic watcher again. This is only useful
674	when you changed some parameters or the reschedule callback would return	2178	when you changed some parameters or the reschedule callback would return
675	a different time than the last time it was called (e.g. in a crond like	2179	a different time than the last time it was called (e.g. in a crond like
676	program when the crontabs have changed).	2180	program when the crontabs have changed).
677		2181
		2182	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2183
		2184	When active, returns the absolute time that the watcher is supposed
		2185	to trigger next. This is not the same as the C<offset> argument to
		2186	C<ev_periodic_set>, but indeed works even in interval and manual
		2187	rescheduling modes.
		2188
		2189	=item ev_tstamp offset [read-write]
		2190
		2191	When repeating, this contains the offset value, otherwise this is the
		2192	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2193	although libev might modify this value for better numerical stability).
		2194
		2195	Can be modified any time, but changes only take effect when the periodic
		2196	timer fires or C<ev_periodic_again> is being called.
		2197
		2198	=item ev_tstamp interval [read-write]
		2199
		2200	The current interval value. Can be modified any time, but changes only
		2201	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2202	called.
		2203
		2204	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2205
		2206	The current reschedule callback, or C<0>, if this functionality is
		2207	switched off. Can be changed any time, but changes only take effect when
		2208	the periodic timer fires or C<ev_periodic_again> is being called.
		2209
678	=back	2210	=back
679		2211
		2212	=head3 Examples
		2213
		2214	Example: Call a callback every hour, or, more precisely, whenever the
		2215	system time is divisible by 3600. The callback invocation times have
		2216	potentially a lot of jitter, but good long-term stability.
		2217
		2218	static void
		2219	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2220	{
		2221	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2222	}
		2223
		2224	ev_periodic hourly_tick;
		2225	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2226	ev_periodic_start (loop, &hourly_tick);
		2227
		2228	Example: The same as above, but use a reschedule callback to do it:
		2229
		2230	#include <math.h>
		2231
		2232	static ev_tstamp
		2233	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2234	{
		2235	return now + (3600. - fmod (now, 3600.));
		2236	}
		2237
		2238	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2239
		2240	Example: Call a callback every hour, starting now:
		2241
		2242	ev_periodic hourly_tick;
		2243	ev_periodic_init (&hourly_tick, clock_cb,
		2244	fmod (ev_now (loop), 3600.), 3600., 0);
		2245	ev_periodic_start (loop, &hourly_tick);
		2246
		2247
680	=head2 C<ev_signal> - signal me when a signal gets signalled	2248	=head2 C<ev_signal> - signal me when a signal gets signalled!
681		2249
682	Signal watchers will trigger an event when the process receives a specific	2250	Signal watchers will trigger an event when the process receives a specific
683	signal one or more times. Even though signals are very asynchronous, libev	2251	signal one or more times. Even though signals are very asynchronous, libev
684	will try it's best to deliver signals synchronously, i.e. as part of the	2252	will try it's best to deliver signals synchronously, i.e. as part of the
685	normal event processing, like any other event.	2253	normal event processing, like any other event.
686		2254
		2255	If you want signals to be delivered truly asynchronously, just use
		2256	C<sigaction> as you would do without libev and forget about sharing
		2257	the signal. You can even use C<ev_async> from a signal handler to
		2258	synchronously wake up an event loop.
		2259
687	You can configure as many watchers as you like per signal. Only when the	2260	You can configure as many watchers as you like for the same signal, but
		2261	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2262	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2263	C<SIGINT> in both the default loop and another loop at the same time. At
		2264	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2265
688	first watcher gets started will libev actually register a signal watcher	2266	When the first watcher gets started will libev actually register something
689	with the kernel (thus it coexists with your own signal handlers as long	2267	with the kernel (thus it coexists with your own signal handlers as long as
690	as you don't register any with libev). Similarly, when the last signal	2268	you don't register any with libev for the same signal).
691	watcher for a signal is stopped libev will reset the signal handler to	2269
692	SIG_DFL (regardless of what it was set to before).	2270	If possible and supported, libev will install its handlers with
		2271	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2272	not be unduly interrupted. If you have a problem with system calls getting
		2273	interrupted by signals you can block all signals in an C<ev_check> watcher
		2274	and unblock them in an C<ev_prepare> watcher.
		2275
		2276	=head3 The special problem of inheritance over fork/execve/pthread_create
		2277
		2278	Both the signal mask (C<sigprocmask>) and the signal disposition
		2279	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2280	stopping it again), that is, libev might or might not block the signal,
		2281	and might or might not set or restore the installed signal handler.
		2282
		2283	While this does not matter for the signal disposition (libev never
		2284	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2285	C<execve>), this matters for the signal mask: many programs do not expect
		2286	certain signals to be blocked.
		2287
		2288	This means that before calling C<exec> (from the child) you should reset
		2289	the signal mask to whatever "default" you expect (all clear is a good
		2290	choice usually).
		2291
		2292	The simplest way to ensure that the signal mask is reset in the child is
		2293	to install a fork handler with C<pthread_atfork> that resets it. That will
		2294	catch fork calls done by libraries (such as the libc) as well.
		2295
		2296	In current versions of libev, the signal will not be blocked indefinitely
		2297	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2298	the window of opportunity for problems, it will not go away, as libev
		2299	I<has> to modify the signal mask, at least temporarily.
		2300
		2301	So I can't stress this enough: I<If you do not reset your signal mask when
		2302	you expect it to be empty, you have a race condition in your code>. This
		2303	is not a libev-specific thing, this is true for most event libraries.
		2304
		2305	=head3 Watcher-Specific Functions and Data Members
693		2306
694	=over 4	2307	=over 4
695		2308
696	=item ev_signal_init (ev_signal *, callback, int signum)	2309	=item ev_signal_init (ev_signal *, callback, int signum)
697		2310
698	=item ev_signal_set (ev_signal *, int signum)	2311	=item ev_signal_set (ev_signal *, int signum)
699		2312
700	Configures the watcher to trigger on the given signal number (usually one	2313	Configures the watcher to trigger on the given signal number (usually one
701	of the C<SIGxxx> constants).	2314	of the C<SIGxxx> constants).
702		2315
		2316	=item int signum [read-only]
		2317
		2318	The signal the watcher watches out for.
		2319
703	=back	2320	=back
704		2321
		2322	=head3 Examples
		2323
		2324	Example: Try to exit cleanly on SIGINT.
		2325
		2326	static void
		2327	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2328	{
		2329	ev_break (loop, EVBREAK_ALL);
		2330	}
		2331
		2332	ev_signal signal_watcher;
		2333	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2334	ev_signal_start (loop, &signal_watcher);
		2335
		2336
705	=head2 C<ev_child> - wait for pid status changes	2337	=head2 C<ev_child> - watch out for process status changes
706		2338
707	Child watchers trigger when your process receives a SIGCHLD in response to	2339	Child watchers trigger when your process receives a SIGCHLD in response to
708	some child status changes (most typically when a child of yours dies).	2340	some child status changes (most typically when a child of yours dies or
		2341	exits). It is permissible to install a child watcher I<after> the child
		2342	has been forked (which implies it might have already exited), as long
		2343	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2344	forking and then immediately registering a watcher for the child is fine,
		2345	but forking and registering a watcher a few event loop iterations later or
		2346	in the next callback invocation is not.
		2347
		2348	Only the default event loop is capable of handling signals, and therefore
		2349	you can only register child watchers in the default event loop.
		2350
		2351	Due to some design glitches inside libev, child watchers will always be
		2352	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2353	libev)
		2354
		2355	=head3 Process Interaction
		2356
		2357	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2358	initialised. This is necessary to guarantee proper behaviour even if the
		2359	first child watcher is started after the child exits. The occurrence
		2360	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2361	synchronously as part of the event loop processing. Libev always reaps all
		2362	children, even ones not watched.
		2363
		2364	=head3 Overriding the Built-In Processing
		2365
		2366	Libev offers no special support for overriding the built-in child
		2367	processing, but if your application collides with libev's default child
		2368	handler, you can override it easily by installing your own handler for
		2369	C<SIGCHLD> after initialising the default loop, and making sure the
		2370	default loop never gets destroyed. You are encouraged, however, to use an
		2371	event-based approach to child reaping and thus use libev's support for
		2372	that, so other libev users can use C<ev_child> watchers freely.
		2373
		2374	=head3 Stopping the Child Watcher
		2375
		2376	Currently, the child watcher never gets stopped, even when the
		2377	child terminates, so normally one needs to stop the watcher in the
		2378	callback. Future versions of libev might stop the watcher automatically
		2379	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2380	problem).
		2381
		2382	=head3 Watcher-Specific Functions and Data Members
709		2383
710	=over 4	2384	=over 4
711		2385
712	=item ev_child_init (ev_child *, callback, int pid)	2386	=item ev_child_init (ev_child *, callback, int pid, int trace)
713		2387
714	=item ev_child_set (ev_child *, int pid)	2388	=item ev_child_set (ev_child *, int pid, int trace)
715		2389
716	Configures the watcher to wait for status changes of process C<pid> (or	2390	Configures the watcher to wait for status changes of process C<pid> (or
717	I<any> process if C<pid> is specified as C<0>). The callback can look	2391	I<any> process if C<pid> is specified as C<0>). The callback can look
718	at the C<rstatus> member of the C<ev_child> watcher structure to see	2392	at the C<rstatus> member of the C<ev_child> watcher structure to see
719	the status word (use the macros from C<sys/wait.h> and see your systems	2393	the status word (use the macros from C<sys/wait.h> and see your systems
720	C<waitpid> documentation). The C<rpid> member contains the pid of the	2394	C<waitpid> documentation). The C<rpid> member contains the pid of the
721	process causing the status change.	2395	process causing the status change. C<trace> must be either C<0> (only
		2396	activate the watcher when the process terminates) or C<1> (additionally
		2397	activate the watcher when the process is stopped or continued).
		2398
		2399	=item int pid [read-only]
		2400
		2401	The process id this watcher watches out for, or C<0>, meaning any process id.
		2402
		2403	=item int rpid [read-write]
		2404
		2405	The process id that detected a status change.
		2406
		2407	=item int rstatus [read-write]
		2408
		2409	The process exit/trace status caused by C<rpid> (see your systems
		2410	C<waitpid> and C<sys/wait.h> documentation for details).
722		2411
723	=back	2412	=back
724		2413
		2414	=head3 Examples
		2415
		2416	Example: C<fork()> a new process and install a child handler to wait for
		2417	its completion.
		2418
		2419	ev_child cw;
		2420
		2421	static void
		2422	child_cb (EV_P_ ev_child *w, int revents)
		2423	{
		2424	ev_child_stop (EV_A_ w);
		2425	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2426	}
		2427
		2428	pid_t pid = fork ();
		2429
		2430	if (pid < 0)
		2431	// error
		2432	else if (pid == 0)
		2433	{
		2434	// the forked child executes here
		2435	exit (1);
		2436	}
		2437	else
		2438	{
		2439	ev_child_init (&cw, child_cb, pid, 0);
		2440	ev_child_start (EV_DEFAULT_ &cw);
		2441	}
		2442
		2443
		2444	=head2 C<ev_stat> - did the file attributes just change?
		2445
		2446	This watches a file system path for attribute changes. That is, it calls
		2447	C<stat> on that path in regular intervals (or when the OS says it changed)
		2448	and sees if it changed compared to the last time, invoking the callback if
		2449	it did.
		2450
		2451	The path does not need to exist: changing from "path exists" to "path does
		2452	not exist" is a status change like any other. The condition "path does not
		2453	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2454	C<st_nlink> field being zero (which is otherwise always forced to be at
		2455	least one) and all the other fields of the stat buffer having unspecified
		2456	contents.
		2457
		2458	The path I<must not> end in a slash or contain special components such as
		2459	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2460	your working directory changes, then the behaviour is undefined.
		2461
		2462	Since there is no portable change notification interface available, the
		2463	portable implementation simply calls C<stat(2)> regularly on the path
		2464	to see if it changed somehow. You can specify a recommended polling
		2465	interval for this case. If you specify a polling interval of C<0> (highly
		2466	recommended!) then a I<suitable, unspecified default> value will be used
		2467	(which you can expect to be around five seconds, although this might
		2468	change dynamically). Libev will also impose a minimum interval which is
		2469	currently around C<0.1>, but that's usually overkill.
		2470
		2471	This watcher type is not meant for massive numbers of stat watchers,
		2472	as even with OS-supported change notifications, this can be
		2473	resource-intensive.
		2474
		2475	At the time of this writing, the only OS-specific interface implemented
		2476	is the Linux inotify interface (implementing kqueue support is left as an
		2477	exercise for the reader. Note, however, that the author sees no way of
		2478	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2479
		2480	=head3 ABI Issues (Largefile Support)
		2481
		2482	Libev by default (unless the user overrides this) uses the default
		2483	compilation environment, which means that on systems with large file
		2484	support disabled by default, you get the 32 bit version of the stat
		2485	structure. When using the library from programs that change the ABI to
		2486	use 64 bit file offsets the programs will fail. In that case you have to
		2487	compile libev with the same flags to get binary compatibility. This is
		2488	obviously the case with any flags that change the ABI, but the problem is
		2489	most noticeably displayed with ev_stat and large file support.
		2490
		2491	The solution for this is to lobby your distribution maker to make large
		2492	file interfaces available by default (as e.g. FreeBSD does) and not
		2493	optional. Libev cannot simply switch on large file support because it has
		2494	to exchange stat structures with application programs compiled using the
		2495	default compilation environment.
		2496
		2497	=head3 Inotify and Kqueue
		2498
		2499	When C<inotify (7)> support has been compiled into libev and present at
		2500	runtime, it will be used to speed up change detection where possible. The
		2501	inotify descriptor will be created lazily when the first C<ev_stat>
		2502	watcher is being started.
		2503
		2504	Inotify presence does not change the semantics of C<ev_stat> watchers
		2505	except that changes might be detected earlier, and in some cases, to avoid
		2506	making regular C<stat> calls. Even in the presence of inotify support
		2507	there are many cases where libev has to resort to regular C<stat> polling,
		2508	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2509	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2510	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2511	xfs are fully working) libev usually gets away without polling.
		2512
		2513	There is no support for kqueue, as apparently it cannot be used to
		2514	implement this functionality, due to the requirement of having a file
		2515	descriptor open on the object at all times, and detecting renames, unlinks
		2516	etc. is difficult.
		2517
		2518	=head3 C<stat ()> is a synchronous operation
		2519
		2520	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2521	the process. The exception are C<ev_stat> watchers - those call C<stat
		2522	()>, which is a synchronous operation.
		2523
		2524	For local paths, this usually doesn't matter: unless the system is very
		2525	busy or the intervals between stat's are large, a stat call will be fast,
		2526	as the path data is usually in memory already (except when starting the
		2527	watcher).
		2528
		2529	For networked file systems, calling C<stat ()> can block an indefinite
		2530	time due to network issues, and even under good conditions, a stat call
		2531	often takes multiple milliseconds.
		2532
		2533	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2534	paths, although this is fully supported by libev.
		2535
		2536	=head3 The special problem of stat time resolution
		2537
		2538	The C<stat ()> system call only supports full-second resolution portably,
		2539	and even on systems where the resolution is higher, most file systems
		2540	still only support whole seconds.
		2541
		2542	That means that, if the time is the only thing that changes, you can
		2543	easily miss updates: on the first update, C<ev_stat> detects a change and
		2544	calls your callback, which does something. When there is another update
		2545	within the same second, C<ev_stat> will be unable to detect unless the
		2546	stat data does change in other ways (e.g. file size).
		2547
		2548	The solution to this is to delay acting on a change for slightly more
		2549	than a second (or till slightly after the next full second boundary), using
		2550	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2551	ev_timer_again (loop, w)>).
		2552
		2553	The C<.02> offset is added to work around small timing inconsistencies
		2554	of some operating systems (where the second counter of the current time
		2555	might be be delayed. One such system is the Linux kernel, where a call to
		2556	C<gettimeofday> might return a timestamp with a full second later than
		2557	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2558	update file times then there will be a small window where the kernel uses
		2559	the previous second to update file times but libev might already execute
		2560	the timer callback).
		2561
		2562	=head3 Watcher-Specific Functions and Data Members
		2563
		2564	=over 4
		2565
		2566	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2567
		2568	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2569
		2570	Configures the watcher to wait for status changes of the given
		2571	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2572	be detected and should normally be specified as C<0> to let libev choose
		2573	a suitable value. The memory pointed to by C<path> must point to the same
		2574	path for as long as the watcher is active.
		2575
		2576	The callback will receive an C<EV_STAT> event when a change was detected,
		2577	relative to the attributes at the time the watcher was started (or the
		2578	last change was detected).
		2579
		2580	=item ev_stat_stat (loop, ev_stat *)
		2581
		2582	Updates the stat buffer immediately with new values. If you change the
		2583	watched path in your callback, you could call this function to avoid
		2584	detecting this change (while introducing a race condition if you are not
		2585	the only one changing the path). Can also be useful simply to find out the
		2586	new values.
		2587
		2588	=item ev_statdata attr [read-only]
		2589
		2590	The most-recently detected attributes of the file. Although the type is
		2591	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2592	suitable for your system, but you can only rely on the POSIX-standardised
		2593	members to be present. If the C<st_nlink> member is C<0>, then there was
		2594	some error while C<stat>ing the file.
		2595
		2596	=item ev_statdata prev [read-only]
		2597
		2598	The previous attributes of the file. The callback gets invoked whenever
		2599	C<prev> != C<attr>, or, more precisely, one or more of these members
		2600	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2601	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2602
		2603	=item ev_tstamp interval [read-only]
		2604
		2605	The specified interval.
		2606
		2607	=item const char *path [read-only]
		2608
		2609	The file system path that is being watched.
		2610
		2611	=back
		2612
		2613	=head3 Examples
		2614
		2615	Example: Watch C</etc/passwd> for attribute changes.
		2616
		2617	static void
		2618	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2619	{
		2620	/* /etc/passwd changed in some way */
		2621	if (w->attr.st_nlink)
		2622	{
		2623	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2624	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2625	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2626	}
		2627	else
		2628	/* you shalt not abuse printf for puts */
		2629	puts ("wow, /etc/passwd is not there, expect problems. "
		2630	"if this is windows, they already arrived\n");
		2631	}
		2632
		2633	...
		2634	ev_stat passwd;
		2635
		2636	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2637	ev_stat_start (loop, &passwd);
		2638
		2639	Example: Like above, but additionally use a one-second delay so we do not
		2640	miss updates (however, frequent updates will delay processing, too, so
		2641	one might do the work both on C<ev_stat> callback invocation I<and> on
		2642	C<ev_timer> callback invocation).
		2643
		2644	static ev_stat passwd;
		2645	static ev_timer timer;
		2646
		2647	static void
		2648	timer_cb (EV_P_ ev_timer *w, int revents)
		2649	{
		2650	ev_timer_stop (EV_A_ w);
		2651
		2652	/* now it's one second after the most recent passwd change */
		2653	}
		2654
		2655	static void
		2656	stat_cb (EV_P_ ev_stat *w, int revents)
		2657	{
		2658	/* reset the one-second timer */
		2659	ev_timer_again (EV_A_ &timer);
		2660	}
		2661
		2662	...
		2663	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2664	ev_stat_start (loop, &passwd);
		2665	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2666
		2667
725	=head2 C<ev_idle> - when you've got nothing better to do	2668	=head2 C<ev_idle> - when you've got nothing better to do...
726		2669
727	Idle watchers trigger events when there are no other events are pending	2670	Idle watchers trigger events when no other events of the same or higher
728	(prepare, check and other idle watchers do not count). That is, as long	2671	priority are pending (prepare, check and other idle watchers do not count
729	as your process is busy handling sockets or timeouts (or even signals,	2672	as receiving "events").
730	imagine) it will not be triggered. But when your process is idle all idle	2673
731	watchers are being called again and again, once per event loop iteration -	2674	That is, as long as your process is busy handling sockets or timeouts
		2675	(or even signals, imagine) of the same or higher priority it will not be
		2676	triggered. But when your process is idle (or only lower-priority watchers
		2677	are pending), the idle watchers are being called once per event loop
732	until stopped, that is, or your process receives more events and becomes	2678	iteration - until stopped, that is, or your process receives more events
733	busy.	2679	and becomes busy again with higher priority stuff.
734		2680
735	The most noteworthy effect is that as long as any idle watchers are	2681	The most noteworthy effect is that as long as any idle watchers are
736	active, the process will not block when waiting for new events.	2682	active, the process will not block when waiting for new events.
737		2683
738	Apart from keeping your process non-blocking (which is a useful	2684	Apart from keeping your process non-blocking (which is a useful
739	effect on its own sometimes), idle watchers are a good place to do	2685	effect on its own sometimes), idle watchers are a good place to do
740	"pseudo-background processing", or delay processing stuff to after the	2686	"pseudo-background processing", or delay processing stuff to after the
741	event loop has handled all outstanding events.	2687	event loop has handled all outstanding events.
742		2688
		2689	=head3 Watcher-Specific Functions and Data Members
		2690
743	=over 4	2691	=over 4
744		2692
745	=item ev_idle_init (ev_signal *, callback)	2693	=item ev_idle_init (ev_idle *, callback)
746		2694
747	Initialises and configures the idle watcher - it has no parameters of any	2695	Initialises and configures the idle watcher - it has no parameters of any
748	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2696	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
749	believe me.	2697	believe me.
750		2698
751	=back	2699	=back
752		2700
		2701	=head3 Examples
		2702
		2703	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2704	callback, free it. Also, use no error checking, as usual.
		2705
		2706	static void
		2707	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2708	{
		2709	free (w);
		2710	// now do something you wanted to do when the program has
		2711	// no longer anything immediate to do.
		2712	}
		2713
		2714	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2715	ev_idle_init (idle_watcher, idle_cb);
		2716	ev_idle_start (loop, idle_watcher);
		2717
		2718
753	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2719	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
754		2720
755	Prepare and check watchers are usually (but not always) used in tandem:	2721	Prepare and check watchers are usually (but not always) used in pairs:
756	prepare watchers get invoked before the process blocks and check watchers	2722	prepare watchers get invoked before the process blocks and check watchers
757	afterwards.	2723	afterwards.
758		2724
		2725	You I<must not> call C<ev_run> or similar functions that enter
		2726	the current event loop from either C<ev_prepare> or C<ev_check>
		2727	watchers. Other loops than the current one are fine, however. The
		2728	rationale behind this is that you do not need to check for recursion in
		2729	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2730	C<ev_check> so if you have one watcher of each kind they will always be
		2731	called in pairs bracketing the blocking call.
		2732
759	Their main purpose is to integrate other event mechanisms into libev. This	2733	Their main purpose is to integrate other event mechanisms into libev and
760	could be used, for example, to track variable changes, implement your own	2734	their use is somewhat advanced. They could be used, for example, to track
761	watchers, integrate net-snmp or a coroutine library and lots more.	2735	variable changes, implement your own watchers, integrate net-snmp or a
		2736	coroutine library and lots more. They are also occasionally useful if
		2737	you cache some data and want to flush it before blocking (for example,
		2738	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2739	watcher).
762		2740
763	This is done by examining in each prepare call which file descriptors need	2741	This is done by examining in each prepare call which file descriptors
764	to be watched by the other library, registering C<ev_io> watchers for	2742	need to be watched by the other library, registering C<ev_io> watchers
765	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2743	for them and starting an C<ev_timer> watcher for any timeouts (many
766	provide just this functionality). Then, in the check watcher you check for	2744	libraries provide exactly this functionality). Then, in the check watcher,
767	any events that occured (by checking the pending status of all watchers	2745	you check for any events that occurred (by checking the pending status
768	and stopping them) and call back into the library. The I/O and timer	2746	of all watchers and stopping them) and call back into the library. The
769	callbacks will never actually be called (but must be valid nevertheless,	2747	I/O and timer callbacks will never actually be called (but must be valid
770	because you never know, you know?).	2748	nevertheless, because you never know, you know?).
771		2749
772	As another example, the Perl Coro module uses these hooks to integrate	2750	As another example, the Perl Coro module uses these hooks to integrate
773	coroutines into libev programs, by yielding to other active coroutines	2751	coroutines into libev programs, by yielding to other active coroutines
774	during each prepare and only letting the process block if no coroutines	2752	during each prepare and only letting the process block if no coroutines
775	are ready to run (it's actually more complicated: it only runs coroutines	2753	are ready to run (it's actually more complicated: it only runs coroutines
776	with priority higher than or equal to the event loop and one coroutine	2754	with priority higher than or equal to the event loop and one coroutine
777	of lower priority, but only once, using idle watchers to keep the event	2755	of lower priority, but only once, using idle watchers to keep the event
778	loop from blocking if lower-priority coroutines are active, thus mapping	2756	loop from blocking if lower-priority coroutines are active, thus mapping
779	low-priority coroutines to idle/background tasks).	2757	low-priority coroutines to idle/background tasks).
780		2758
		2759	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2760	priority, to ensure that they are being run before any other watchers
		2761	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2762
		2763	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2764	activate ("feed") events into libev. While libev fully supports this, they
		2765	might get executed before other C<ev_check> watchers did their job. As
		2766	C<ev_check> watchers are often used to embed other (non-libev) event
		2767	loops those other event loops might be in an unusable state until their
		2768	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2769	others).
		2770
		2771	=head3 Watcher-Specific Functions and Data Members
		2772
781	=over 4	2773	=over 4
782		2774
783	=item ev_prepare_init (ev_prepare *, callback)	2775	=item ev_prepare_init (ev_prepare *, callback)
784		2776
785	=item ev_check_init (ev_check *, callback)	2777	=item ev_check_init (ev_check *, callback)
786		2778
787	Initialises and configures the prepare or check watcher - they have no	2779	Initialises and configures the prepare or check watcher - they have no
788	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2780	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
789	macros, but using them is utterly, utterly and completely pointless.	2781	macros, but using them is utterly, utterly, utterly and completely
		2782	pointless.
790		2783
791	=back	2784	=back
792		2785
		2786	=head3 Examples
		2787
		2788	There are a number of principal ways to embed other event loops or modules
		2789	into libev. Here are some ideas on how to include libadns into libev
		2790	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2791	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2792	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2793	Glib event loop).
		2794
		2795	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2796	and in a check watcher, destroy them and call into libadns. What follows
		2797	is pseudo-code only of course. This requires you to either use a low
		2798	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2799	the callbacks for the IO/timeout watchers might not have been called yet.
		2800
		2801	static ev_io iow [nfd];
		2802	static ev_timer tw;
		2803
		2804	static void
		2805	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2806	{
		2807	}
		2808
		2809	// create io watchers for each fd and a timer before blocking
		2810	static void
		2811	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2812	{
		2813	int timeout = 3600000;
		2814	struct pollfd fds [nfd];
		2815	// actual code will need to loop here and realloc etc.
		2816	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2817
		2818	/* the callback is illegal, but won't be called as we stop during check */
		2819	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2820	ev_timer_start (loop, &tw);
		2821
		2822	// create one ev_io per pollfd
		2823	for (int i = 0; i < nfd; ++i)
		2824	{
		2825	ev_io_init (iow + i, io_cb, fds [i].fd,
		2826	((fds [i].events & POLLIN ? EV_READ : 0)
		2827	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2828
		2829	fds [i].revents = 0;
		2830	ev_io_start (loop, iow + i);
		2831	}
		2832	}
		2833
		2834	// stop all watchers after blocking
		2835	static void
		2836	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2837	{
		2838	ev_timer_stop (loop, &tw);
		2839
		2840	for (int i = 0; i < nfd; ++i)
		2841	{
		2842	// set the relevant poll flags
		2843	// could also call adns_processreadable etc. here
		2844	struct pollfd *fd = fds + i;
		2845	int revents = ev_clear_pending (iow + i);
		2846	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2847	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2848
		2849	// now stop the watcher
		2850	ev_io_stop (loop, iow + i);
		2851	}
		2852
		2853	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2854	}
		2855
		2856	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2857	in the prepare watcher and would dispose of the check watcher.
		2858
		2859	Method 3: If the module to be embedded supports explicit event
		2860	notification (libadns does), you can also make use of the actual watcher
		2861	callbacks, and only destroy/create the watchers in the prepare watcher.
		2862
		2863	static void
		2864	timer_cb (EV_P_ ev_timer *w, int revents)
		2865	{
		2866	adns_state ads = (adns_state)w->data;
		2867	update_now (EV_A);
		2868
		2869	adns_processtimeouts (ads, &tv_now);
		2870	}
		2871
		2872	static void
		2873	io_cb (EV_P_ ev_io *w, int revents)
		2874	{
		2875	adns_state ads = (adns_state)w->data;
		2876	update_now (EV_A);
		2877
		2878	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2879	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2880	}
		2881
		2882	// do not ever call adns_afterpoll
		2883
		2884	Method 4: Do not use a prepare or check watcher because the module you
		2885	want to embed is not flexible enough to support it. Instead, you can
		2886	override their poll function. The drawback with this solution is that the
		2887	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2888	this approach, effectively embedding EV as a client into the horrible
		2889	libglib event loop.
		2890
		2891	static gint
		2892	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2893	{
		2894	int got_events = 0;
		2895
		2896	for (n = 0; n < nfds; ++n)
		2897	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2898
		2899	if (timeout >= 0)
		2900	// create/start timer
		2901
		2902	// poll
		2903	ev_run (EV_A_ 0);
		2904
		2905	// stop timer again
		2906	if (timeout >= 0)
		2907	ev_timer_stop (EV_A_ &to);
		2908
		2909	// stop io watchers again - their callbacks should have set
		2910	for (n = 0; n < nfds; ++n)
		2911	ev_io_stop (EV_A_ iow [n]);
		2912
		2913	return got_events;
		2914	}
		2915
		2916
		2917	=head2 C<ev_embed> - when one backend isn't enough...
		2918
		2919	This is a rather advanced watcher type that lets you embed one event loop
		2920	into another (currently only C<ev_io> events are supported in the embedded
		2921	loop, other types of watchers might be handled in a delayed or incorrect
		2922	fashion and must not be used).
		2923
		2924	There are primarily two reasons you would want that: work around bugs and
		2925	prioritise I/O.
		2926
		2927	As an example for a bug workaround, the kqueue backend might only support
		2928	sockets on some platform, so it is unusable as generic backend, but you
		2929	still want to make use of it because you have many sockets and it scales
		2930	so nicely. In this case, you would create a kqueue-based loop and embed
		2931	it into your default loop (which might use e.g. poll). Overall operation
		2932	will be a bit slower because first libev has to call C<poll> and then
		2933	C<kevent>, but at least you can use both mechanisms for what they are
		2934	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2935
		2936	As for prioritising I/O: under rare circumstances you have the case where
		2937	some fds have to be watched and handled very quickly (with low latency),
		2938	and even priorities and idle watchers might have too much overhead. In
		2939	this case you would put all the high priority stuff in one loop and all
		2940	the rest in a second one, and embed the second one in the first.
		2941
		2942	As long as the watcher is active, the callback will be invoked every
		2943	time there might be events pending in the embedded loop. The callback
		2944	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2945	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2946	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2947	to give the embedded loop strictly lower priority for example).
		2948
		2949	You can also set the callback to C<0>, in which case the embed watcher
		2950	will automatically execute the embedded loop sweep whenever necessary.
		2951
		2952	Fork detection will be handled transparently while the C<ev_embed> watcher
		2953	is active, i.e., the embedded loop will automatically be forked when the
		2954	embedding loop forks. In other cases, the user is responsible for calling
		2955	C<ev_loop_fork> on the embedded loop.
		2956
		2957	Unfortunately, not all backends are embeddable: only the ones returned by
		2958	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2959	portable one.
		2960
		2961	So when you want to use this feature you will always have to be prepared
		2962	that you cannot get an embeddable loop. The recommended way to get around
		2963	this is to have a separate variables for your embeddable loop, try to
		2964	create it, and if that fails, use the normal loop for everything.
		2965
		2966	=head3 C<ev_embed> and fork
		2967
		2968	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2969	automatically be applied to the embedded loop as well, so no special
		2970	fork handling is required in that case. When the watcher is not running,
		2971	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2972	as applicable.
		2973
		2974	=head3 Watcher-Specific Functions and Data Members
		2975
		2976	=over 4
		2977
		2978	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2979
		2980	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2981
		2982	Configures the watcher to embed the given loop, which must be
		2983	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2984	invoked automatically, otherwise it is the responsibility of the callback
		2985	to invoke it (it will continue to be called until the sweep has been done,
		2986	if you do not want that, you need to temporarily stop the embed watcher).
		2987
		2988	=item ev_embed_sweep (loop, ev_embed *)
		2989
		2990	Make a single, non-blocking sweep over the embedded loop. This works
		2991	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		2992	appropriate way for embedded loops.
		2993
		2994	=item struct ev_loop *other [read-only]
		2995
		2996	The embedded event loop.
		2997
		2998	=back
		2999
		3000	=head3 Examples
		3001
		3002	Example: Try to get an embeddable event loop and embed it into the default
		3003	event loop. If that is not possible, use the default loop. The default
		3004	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3005	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3006	used).
		3007
		3008	struct ev_loop *loop_hi = ev_default_init (0);
		3009	struct ev_loop *loop_lo = 0;
		3010	ev_embed embed;
		3011
		3012	// see if there is a chance of getting one that works
		3013	// (remember that a flags value of 0 means autodetection)
		3014	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3015	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3016	: 0;
		3017
		3018	// if we got one, then embed it, otherwise default to loop_hi
		3019	if (loop_lo)
		3020	{
		3021	ev_embed_init (&embed, 0, loop_lo);
		3022	ev_embed_start (loop_hi, &embed);
		3023	}
		3024	else
		3025	loop_lo = loop_hi;
		3026
		3027	Example: Check if kqueue is available but not recommended and create
		3028	a kqueue backend for use with sockets (which usually work with any
		3029	kqueue implementation). Store the kqueue/socket-only event loop in
		3030	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3031
		3032	struct ev_loop *loop = ev_default_init (0);
		3033	struct ev_loop *loop_socket = 0;
		3034	ev_embed embed;
		3035
		3036	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3037	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3038	{
		3039	ev_embed_init (&embed, 0, loop_socket);
		3040	ev_embed_start (loop, &embed);
		3041	}
		3042
		3043	if (!loop_socket)
		3044	loop_socket = loop;
		3045
		3046	// now use loop_socket for all sockets, and loop for everything else
		3047
		3048
		3049	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3050
		3051	Fork watchers are called when a C<fork ()> was detected (usually because
		3052	whoever is a good citizen cared to tell libev about it by calling
		3053	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		3054	event loop blocks next and before C<ev_check> watchers are being called,
		3055	and only in the child after the fork. If whoever good citizen calling
		3056	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		3057	handlers will be invoked, too, of course.
		3058
		3059	=head3 The special problem of life after fork - how is it possible?
		3060
		3061	Most uses of C<fork()> consist of forking, then some simple calls to set
		3062	up/change the process environment, followed by a call to C<exec()>. This
		3063	sequence should be handled by libev without any problems.
		3064
		3065	This changes when the application actually wants to do event handling
		3066	in the child, or both parent in child, in effect "continuing" after the
		3067	fork.
		3068
		3069	The default mode of operation (for libev, with application help to detect
		3070	forks) is to duplicate all the state in the child, as would be expected
		3071	when I<either> the parent I<or> the child process continues.
		3072
		3073	When both processes want to continue using libev, then this is usually the
		3074	wrong result. In that case, usually one process (typically the parent) is
		3075	supposed to continue with all watchers in place as before, while the other
		3076	process typically wants to start fresh, i.e. without any active watchers.
		3077
		3078	The cleanest and most efficient way to achieve that with libev is to
		3079	simply create a new event loop, which of course will be "empty", and
		3080	use that for new watchers. This has the advantage of not touching more
		3081	memory than necessary, and thus avoiding the copy-on-write, and the
		3082	disadvantage of having to use multiple event loops (which do not support
		3083	signal watchers).
		3084
		3085	When this is not possible, or you want to use the default loop for
		3086	other reasons, then in the process that wants to start "fresh", call
		3087	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3088	Destroying the default loop will "orphan" (not stop) all registered
		3089	watchers, so you have to be careful not to execute code that modifies
		3090	those watchers. Note also that in that case, you have to re-register any
		3091	signal watchers.
		3092
		3093	=head3 Watcher-Specific Functions and Data Members
		3094
		3095	=over 4
		3096
		3097	=item ev_fork_init (ev_signal *, callback)
		3098
		3099	Initialises and configures the fork watcher - it has no parameters of any
		3100	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3101	believe me.
		3102
		3103	=back
		3104
		3105
		3106	=head2 C<ev_async> - how to wake up an event loop
		3107
		3108	In general, you cannot use an C<ev_run> from multiple threads or other
		3109	asynchronous sources such as signal handlers (as opposed to multiple event
		3110	loops - those are of course safe to use in different threads).
		3111
		3112	Sometimes, however, you need to wake up an event loop you do not control,
		3113	for example because it belongs to another thread. This is what C<ev_async>
		3114	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3115	it by calling C<ev_async_send>, which is thread- and signal safe.
		3116
		3117	This functionality is very similar to C<ev_signal> watchers, as signals,
		3118	too, are asynchronous in nature, and signals, too, will be compressed
		3119	(i.e. the number of callback invocations may be less than the number of
		3120	C<ev_async_sent> calls).
		3121
		3122	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3123	just the default loop.
		3124
		3125	=head3 Queueing
		3126
		3127	C<ev_async> does not support queueing of data in any way. The reason
		3128	is that the author does not know of a simple (or any) algorithm for a
		3129	multiple-writer-single-reader queue that works in all cases and doesn't
		3130	need elaborate support such as pthreads or unportable memory access
		3131	semantics.
		3132
		3133	That means that if you want to queue data, you have to provide your own
		3134	queue. But at least I can tell you how to implement locking around your
		3135	queue:
		3136
		3137	=over 4
		3138
		3139	=item queueing from a signal handler context
		3140
		3141	To implement race-free queueing, you simply add to the queue in the signal
		3142	handler but you block the signal handler in the watcher callback. Here is
		3143	an example that does that for some fictitious SIGUSR1 handler:
		3144
		3145	static ev_async mysig;
		3146
		3147	static void
		3148	sigusr1_handler (void)
		3149	{
		3150	sometype data;
		3151
		3152	// no locking etc.
		3153	queue_put (data);
		3154	ev_async_send (EV_DEFAULT_ &mysig);
		3155	}
		3156
		3157	static void
		3158	mysig_cb (EV_P_ ev_async *w, int revents)
		3159	{
		3160	sometype data;
		3161	sigset_t block, prev;
		3162
		3163	sigemptyset (&block);
		3164	sigaddset (&block, SIGUSR1);
		3165	sigprocmask (SIG_BLOCK, &block, &prev);
		3166
		3167	while (queue_get (&data))
		3168	process (data);
		3169
		3170	if (sigismember (&prev, SIGUSR1)
		3171	sigprocmask (SIG_UNBLOCK, &block, 0);
		3172	}
		3173
		3174	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3175	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3176	either...).
		3177
		3178	=item queueing from a thread context
		3179
		3180	The strategy for threads is different, as you cannot (easily) block
		3181	threads but you can easily preempt them, so to queue safely you need to
		3182	employ a traditional mutex lock, such as in this pthread example:
		3183
		3184	static ev_async mysig;
		3185	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3186
		3187	static void
		3188	otherthread (void)
		3189	{
		3190	// only need to lock the actual queueing operation
		3191	pthread_mutex_lock (&mymutex);
		3192	queue_put (data);
		3193	pthread_mutex_unlock (&mymutex);
		3194
		3195	ev_async_send (EV_DEFAULT_ &mysig);
		3196	}
		3197
		3198	static void
		3199	mysig_cb (EV_P_ ev_async *w, int revents)
		3200	{
		3201	pthread_mutex_lock (&mymutex);
		3202
		3203	while (queue_get (&data))
		3204	process (data);
		3205
		3206	pthread_mutex_unlock (&mymutex);
		3207	}
		3208
		3209	=back
		3210
		3211
		3212	=head3 Watcher-Specific Functions and Data Members
		3213
		3214	=over 4
		3215
		3216	=item ev_async_init (ev_async *, callback)
		3217
		3218	Initialises and configures the async watcher - it has no parameters of any
		3219	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3220	trust me.
		3221
		3222	=item ev_async_send (loop, ev_async *)
		3223
		3224	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3225	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3226	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3227	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3228	section below on what exactly this means).
		3229
		3230	Note that, as with other watchers in libev, multiple events might get
		3231	compressed into a single callback invocation (another way to look at this
		3232	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3233	reset when the event loop detects that).
		3234
		3235	This call incurs the overhead of a system call only once per event loop
		3236	iteration, so while the overhead might be noticeable, it doesn't apply to
		3237	repeated calls to C<ev_async_send> for the same event loop.
		3238
		3239	=item bool = ev_async_pending (ev_async *)
		3240
		3241	Returns a non-zero value when C<ev_async_send> has been called on the
		3242	watcher but the event has not yet been processed (or even noted) by the
		3243	event loop.
		3244
		3245	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3246	the loop iterates next and checks for the watcher to have become active,
		3247	it will reset the flag again. C<ev_async_pending> can be used to very
		3248	quickly check whether invoking the loop might be a good idea.
		3249
		3250	Not that this does I<not> check whether the watcher itself is pending,
		3251	only whether it has been requested to make this watcher pending: there
		3252	is a time window between the event loop checking and resetting the async
		3253	notification, and the callback being invoked.
		3254
		3255	=back
		3256
		3257
793	=head1 OTHER FUNCTIONS	3258	=head1 OTHER FUNCTIONS
794		3259
795	There are some other functions of possible interest. Described. Here. Now.	3260	There are some other functions of possible interest. Described. Here. Now.
796		3261
797	=over 4	3262	=over 4
798		3263
799	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3264	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
800		3265
801	This function combines a simple timer and an I/O watcher, calls your	3266	This function combines a simple timer and an I/O watcher, calls your
802	callback on whichever event happens first and automatically stop both	3267	callback on whichever event happens first and automatically stops both
803	watchers. This is useful if you want to wait for a single event on an fd	3268	watchers. This is useful if you want to wait for a single event on an fd
804	or timeout without having to allocate/configure/start/stop/free one or	3269	or timeout without having to allocate/configure/start/stop/free one or
805	more watchers yourself.	3270	more watchers yourself.
806		3271
807	If C<fd> is less than 0, then no I/O watcher will be started and events	3272	If C<fd> is less than 0, then no I/O watcher will be started and the
808	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3273	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
809	C<events> set will be craeted and started.	3274	the given C<fd> and C<events> set will be created and started.
810		3275
811	If C<timeout> is less than 0, then no timeout watcher will be	3276	If C<timeout> is less than 0, then no timeout watcher will be
812	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3277	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
813	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3278	repeat = 0) will be started. C<0> is a valid timeout.
814	dubious value.
815		3279
816	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3280	The callback has the type C<void (*cb)(int revents, void *arg)> and is
817	passed an C<revents> set like normal event callbacks (a combination of	3281	passed an C<revents> set like normal event callbacks (a combination of
818	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3282	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
819	value passed to C<ev_once>:	3283	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3284	a timeout and an io event at the same time - you probably should give io
		3285	events precedence.
820		3286
		3287	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3288
821	static void stdin_ready (int revents, void *arg)	3289	static void stdin_ready (int revents, void *arg)
822	{	3290	{
823	if (revents & EV_TIMEOUT)
824	/* doh, nothing entered */;
825	else if (revents & EV_READ)	3291	if (revents & EV_READ)
826	/* stdin might have data for us, joy! */;	3292	/* stdin might have data for us, joy! */;
		3293	else if (revents & EV_TIMER)
		3294	/* doh, nothing entered */;
827	}	3295	}
828		3296
829	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3297	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
830
831	=item ev_feed_event (loop, watcher, int events)
832
833	Feeds the given event set into the event loop, as if the specified event
834	had happened for the specified watcher (which must be a pointer to an
835	initialised but not necessarily started event watcher).
836		3298
837	=item ev_feed_fd_event (loop, int fd, int revents)	3299	=item ev_feed_fd_event (loop, int fd, int revents)
838		3300
839	Feed an event on the given fd, as if a file descriptor backend detected	3301	Feed an event on the given fd, as if a file descriptor backend detected
840	the given events it.	3302	the given events it.
841		3303
842	=item ev_feed_signal_event (loop, int signum)	3304	=item ev_feed_signal_event (loop, int signum)
843		3305
844	Feed an event as if the given signal occured (loop must be the default loop!).	3306	Feed an event as if the given signal occurred (C<loop> must be the default
		3307	loop!).
845		3308
846	=back	3309	=back
		3310
847		3311
848	=head1 LIBEVENT EMULATION	3312	=head1 LIBEVENT EMULATION
849		3313
850	Libev offers a compatibility emulation layer for libevent. It cannot	3314	Libev offers a compatibility emulation layer for libevent. It cannot
851	emulate the internals of libevent, so here are some usage hints:	3315	emulate the internals of libevent, so here are some usage hints:
…		…
863		3327
864	=item * Priorities are not currently supported. Initialising priorities	3328	=item * Priorities are not currently supported. Initialising priorities
865	will fail and all watchers will have the same priority, even though there	3329	will fail and all watchers will have the same priority, even though there
866	is an ev_pri field.	3330	is an ev_pri field.
867		3331
		3332	=item * In libevent, the last base created gets the signals, in libev, the
		3333	first base created (== the default loop) gets the signals.
		3334
868	=item * Other members are not supported.	3335	=item * Other members are not supported.
869		3336
870	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3337	=item * The libev emulation is I<not> ABI compatible to libevent, you need
871	to use the libev header file and library.	3338	to use the libev header file and library.
872		3339
873	=back	3340	=back
874		3341
875	=head1 C++ SUPPORT	3342	=head1 C++ SUPPORT
876		3343
877	TBD.	3344	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3345	you to use some convenience methods to start/stop watchers and also change
		3346	the callback model to a model using method callbacks on objects.
		3347
		3348	To use it,
		3349
		3350	#include <ev++.h>
		3351
		3352	This automatically includes F<ev.h> and puts all of its definitions (many
		3353	of them macros) into the global namespace. All C++ specific things are
		3354	put into the C<ev> namespace. It should support all the same embedding
		3355	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3356
		3357	Care has been taken to keep the overhead low. The only data member the C++
		3358	classes add (compared to plain C-style watchers) is the event loop pointer
		3359	that the watcher is associated with (or no additional members at all if
		3360	you disable C<EV_MULTIPLICITY> when embedding libev).
		3361
		3362	Currently, functions, and static and non-static member functions can be
		3363	used as callbacks. Other types should be easy to add as long as they only
		3364	need one additional pointer for context. If you need support for other
		3365	types of functors please contact the author (preferably after implementing
		3366	it).
		3367
		3368	Here is a list of things available in the C<ev> namespace:
		3369
		3370	=over 4
		3371
		3372	=item C<ev::READ>, C<ev::WRITE> etc.
		3373
		3374	These are just enum values with the same values as the C<EV_READ> etc.
		3375	macros from F<ev.h>.
		3376
		3377	=item C<ev::tstamp>, C<ev::now>
		3378
		3379	Aliases to the same types/functions as with the C<ev_> prefix.
		3380
		3381	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3382
		3383	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3384	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3385	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3386	defines by many implementations.
		3387
		3388	All of those classes have these methods:
		3389
		3390	=over 4
		3391
		3392	=item ev::TYPE::TYPE ()
		3393
		3394	=item ev::TYPE::TYPE (loop)
		3395
		3396	=item ev::TYPE::~TYPE
		3397
		3398	The constructor (optionally) takes an event loop to associate the watcher
		3399	with. If it is omitted, it will use C<EV_DEFAULT>.
		3400
		3401	The constructor calls C<ev_init> for you, which means you have to call the
		3402	C<set> method before starting it.
		3403
		3404	It will not set a callback, however: You have to call the templated C<set>
		3405	method to set a callback before you can start the watcher.
		3406
		3407	(The reason why you have to use a method is a limitation in C++ which does
		3408	not allow explicit template arguments for constructors).
		3409
		3410	The destructor automatically stops the watcher if it is active.
		3411
		3412	=item w->set<class, &class::method> (object *)
		3413
		3414	This method sets the callback method to call. The method has to have a
		3415	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3416	first argument and the C<revents> as second. The object must be given as
		3417	parameter and is stored in the C<data> member of the watcher.
		3418
		3419	This method synthesizes efficient thunking code to call your method from
		3420	the C callback that libev requires. If your compiler can inline your
		3421	callback (i.e. it is visible to it at the place of the C<set> call and
		3422	your compiler is good :), then the method will be fully inlined into the
		3423	thunking function, making it as fast as a direct C callback.
		3424
		3425	Example: simple class declaration and watcher initialisation
		3426
		3427	struct myclass
		3428	{
		3429	void io_cb (ev::io &w, int revents) { }
		3430	}
		3431
		3432	myclass obj;
		3433	ev::io iow;
		3434	iow.set <myclass, &myclass::io_cb> (&obj);
		3435
		3436	=item w->set (object *)
		3437
		3438	This is a variation of a method callback - leaving out the method to call
		3439	will default the method to C<operator ()>, which makes it possible to use
		3440	functor objects without having to manually specify the C<operator ()> all
		3441	the time. Incidentally, you can then also leave out the template argument
		3442	list.
		3443
		3444	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3445	int revents)>.
		3446
		3447	See the method-C<set> above for more details.
		3448
		3449	Example: use a functor object as callback.
		3450
		3451	struct myfunctor
		3452	{
		3453	void operator() (ev::io &w, int revents)
		3454	{
		3455	...
		3456	}
		3457	}
		3458
		3459	myfunctor f;
		3460
		3461	ev::io w;
		3462	w.set (&f);
		3463
		3464	=item w->set<function> (void *data = 0)
		3465
		3466	Also sets a callback, but uses a static method or plain function as
		3467	callback. The optional C<data> argument will be stored in the watcher's
		3468	C<data> member and is free for you to use.
		3469
		3470	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3471
		3472	See the method-C<set> above for more details.
		3473
		3474	Example: Use a plain function as callback.
		3475
		3476	static void io_cb (ev::io &w, int revents) { }
		3477	iow.set <io_cb> ();
		3478
		3479	=item w->set (loop)
		3480
		3481	Associates a different C<struct ev_loop> with this watcher. You can only
		3482	do this when the watcher is inactive (and not pending either).
		3483
		3484	=item w->set ([arguments])
		3485
		3486	Basically the same as C<ev_TYPE_set>, with the same arguments. Either this
		3487	method or a suitable start method must be called at least once. Unlike the
		3488	C counterpart, an active watcher gets automatically stopped and restarted
		3489	when reconfiguring it with this method.
		3490
		3491	=item w->start ()
		3492
		3493	Starts the watcher. Note that there is no C<loop> argument, as the
		3494	constructor already stores the event loop.
		3495
		3496	=item w->start ([arguments])
		3497
		3498	Instead of calling C<set> and C<start> methods separately, it is often
		3499	convenient to wrap them in one call. Uses the same type of arguments as
		3500	the configure C<set> method of the watcher.
		3501
		3502	=item w->stop ()
		3503
		3504	Stops the watcher if it is active. Again, no C<loop> argument.
		3505
		3506	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3507
		3508	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3509	C<ev_TYPE_again> function.
		3510
		3511	=item w->sweep () (C<ev::embed> only)
		3512
		3513	Invokes C<ev_embed_sweep>.
		3514
		3515	=item w->update () (C<ev::stat> only)
		3516
		3517	Invokes C<ev_stat_stat>.
		3518
		3519	=back
		3520
		3521	=back
		3522
		3523	Example: Define a class with two I/O and idle watchers, start the I/O
		3524	watchers in the constructor.
		3525
		3526	class myclass
		3527	{
		3528	ev::io io ; void io_cb (ev::io &w, int revents);
		3529	ev::io2 io2 ; void io2_cb (ev::io &w, int revents);
		3530	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3531
		3532	myclass (int fd)
		3533	{
		3534	io .set <myclass, &myclass::io_cb > (this);
		3535	io2 .set <myclass, &myclass::io2_cb > (this);
		3536	idle.set <myclass, &myclass::idle_cb> (this);
		3537
		3538	io.set (fd, ev::WRITE); // configure the watcher
		3539	io.start (); // start it whenever convenient
		3540
		3541	io2.start (fd, ev::READ); // set + start in one call
		3542	}
		3543	};
		3544
		3545
		3546	=head1 OTHER LANGUAGE BINDINGS
		3547
		3548	Libev does not offer other language bindings itself, but bindings for a
		3549	number of languages exist in the form of third-party packages. If you know
		3550	any interesting language binding in addition to the ones listed here, drop
		3551	me a note.
		3552
		3553	=over 4
		3554
		3555	=item Perl
		3556
		3557	The EV module implements the full libev API and is actually used to test
		3558	libev. EV is developed together with libev. Apart from the EV core module,
		3559	there are additional modules that implement libev-compatible interfaces
		3560	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3561	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3562	and C<EV::Glib>).
		3563
		3564	It can be found and installed via CPAN, its homepage is at
		3565	L<http://software.schmorp.de/pkg/EV>.
		3566
		3567	=item Python
		3568
		3569	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3570	seems to be quite complete and well-documented.
		3571
		3572	=item Ruby
		3573
		3574	Tony Arcieri has written a ruby extension that offers access to a subset
		3575	of the libev API and adds file handle abstractions, asynchronous DNS and
		3576	more on top of it. It can be found via gem servers. Its homepage is at
		3577	L<http://rev.rubyforge.org/>.
		3578
		3579	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3580	makes rev work even on mingw.
		3581
		3582	=item Haskell
		3583
		3584	A haskell binding to libev is available at
		3585	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3586
		3587	=item D
		3588
		3589	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3590	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3591
		3592	=item Ocaml
		3593
		3594	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3595	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3596
		3597	=item Lua
		3598
		3599	Brian Maher has written a partial interface to libev for lua (at the
		3600	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3601	L<http://github.com/brimworks/lua-ev>.
		3602
		3603	=back
		3604
		3605
		3606	=head1 MACRO MAGIC
		3607
		3608	Libev can be compiled with a variety of options, the most fundamental
		3609	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3610	functions and callbacks have an initial C<struct ev_loop *> argument.
		3611
		3612	To make it easier to write programs that cope with either variant, the
		3613	following macros are defined:
		3614
		3615	=over 4
		3616
		3617	=item C<EV_A>, C<EV_A_>
		3618
		3619	This provides the loop I<argument> for functions, if one is required ("ev
		3620	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3621	C<EV_A_> is used when other arguments are following. Example:
		3622
		3623	ev_unref (EV_A);
		3624	ev_timer_add (EV_A_ watcher);
		3625	ev_run (EV_A_ 0);
		3626
		3627	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3628	which is often provided by the following macro.
		3629
		3630	=item C<EV_P>, C<EV_P_>
		3631
		3632	This provides the loop I<parameter> for functions, if one is required ("ev
		3633	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3634	C<EV_P_> is used when other parameters are following. Example:
		3635
		3636	// this is how ev_unref is being declared
		3637	static void ev_unref (EV_P);
		3638
		3639	// this is how you can declare your typical callback
		3640	static void cb (EV_P_ ev_timer *w, int revents)
		3641
		3642	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3643	suitable for use with C<EV_A>.
		3644
		3645	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3646
		3647	Similar to the other two macros, this gives you the value of the default
		3648	loop, if multiple loops are supported ("ev loop default").
		3649
		3650	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3651
		3652	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3653	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3654	is undefined when the default loop has not been initialised by a previous
		3655	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3656
		3657	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3658	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3659
		3660	=back
		3661
		3662	Example: Declare and initialise a check watcher, utilising the above
		3663	macros so it will work regardless of whether multiple loops are supported
		3664	or not.
		3665
		3666	static void
		3667	check_cb (EV_P_ ev_timer *w, int revents)
		3668	{
		3669	ev_check_stop (EV_A_ w);
		3670	}
		3671
		3672	ev_check check;
		3673	ev_check_init (&check, check_cb);
		3674	ev_check_start (EV_DEFAULT_ &check);
		3675	ev_run (EV_DEFAULT_ 0);
		3676
		3677	=head1 EMBEDDING
		3678
		3679	Libev can (and often is) directly embedded into host
		3680	applications. Examples of applications that embed it include the Deliantra
		3681	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3682	and rxvt-unicode.
		3683
		3684	The goal is to enable you to just copy the necessary files into your
		3685	source directory without having to change even a single line in them, so
		3686	you can easily upgrade by simply copying (or having a checked-out copy of
		3687	libev somewhere in your source tree).
		3688
		3689	=head2 FILESETS
		3690
		3691	Depending on what features you need you need to include one or more sets of files
		3692	in your application.
		3693
		3694	=head3 CORE EVENT LOOP
		3695
		3696	To include only the libev core (all the C<ev_*> functions), with manual
		3697	configuration (no autoconf):
		3698
		3699	#define EV_STANDALONE 1
		3700	#include "ev.c"
		3701
		3702	This will automatically include F<ev.h>, too, and should be done in a
		3703	single C source file only to provide the function implementations. To use
		3704	it, do the same for F<ev.h> in all files wishing to use this API (best
		3705	done by writing a wrapper around F<ev.h> that you can include instead and
		3706	where you can put other configuration options):
		3707
		3708	#define EV_STANDALONE 1
		3709	#include "ev.h"
		3710
		3711	Both header files and implementation files can be compiled with a C++
		3712	compiler (at least, that's a stated goal, and breakage will be treated
		3713	as a bug).
		3714
		3715	You need the following files in your source tree, or in a directory
		3716	in your include path (e.g. in libev/ when using -Ilibev):
		3717
		3718	ev.h
		3719	ev.c
		3720	ev_vars.h
		3721	ev_wrap.h
		3722
		3723	ev_win32.c required on win32 platforms only
		3724
		3725	ev_select.c only when select backend is enabled (which is enabled by default)
		3726	ev_poll.c only when poll backend is enabled (disabled by default)
		3727	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3728	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3729	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3730
		3731	F<ev.c> includes the backend files directly when enabled, so you only need
		3732	to compile this single file.
		3733
		3734	=head3 LIBEVENT COMPATIBILITY API
		3735
		3736	To include the libevent compatibility API, also include:
		3737
		3738	#include "event.c"
		3739
		3740	in the file including F<ev.c>, and:
		3741
		3742	#include "event.h"
		3743
		3744	in the files that want to use the libevent API. This also includes F<ev.h>.
		3745
		3746	You need the following additional files for this:
		3747
		3748	event.h
		3749	event.c
		3750
		3751	=head3 AUTOCONF SUPPORT
		3752
		3753	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3754	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3755	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3756	include F<config.h> and configure itself accordingly.
		3757
		3758	For this of course you need the m4 file:
		3759
		3760	libev.m4
		3761
		3762	=head2 PREPROCESSOR SYMBOLS/MACROS
		3763
		3764	Libev can be configured via a variety of preprocessor symbols you have to
		3765	define before including (or compiling) any of its files. The default in
		3766	the absence of autoconf is documented for every option.
		3767
		3768	Symbols marked with "(h)" do not change the ABI, and can have different
		3769	values when compiling libev vs. including F<ev.h>, so it is permissible
		3770	to redefine them before including F<ev.h> without breaking compatibility
		3771	to a compiled library. All other symbols change the ABI, which means all
		3772	users of libev and the libev code itself must be compiled with compatible
		3773	settings.
		3774
		3775	=over 4
		3776
		3777	=item EV_COMPAT3 (h)
		3778
		3779	Backwards compatibility is a major concern for libev. This is why this
		3780	release of libev comes with wrappers for the functions and symbols that
		3781	have been renamed between libev version 3 and 4.
		3782
		3783	You can disable these wrappers (to test compatibility with future
		3784	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		3785	sources. This has the additional advantage that you can drop the C<struct>
		3786	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		3787	typedef in that case.
		3788
		3789	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		3790	and in some even more future version the compatibility code will be
		3791	removed completely.
		3792
		3793	=item EV_STANDALONE (h)
		3794
		3795	Must always be C<1> if you do not use autoconf configuration, which
		3796	keeps libev from including F<config.h>, and it also defines dummy
		3797	implementations for some libevent functions (such as logging, which is not
		3798	supported). It will also not define any of the structs usually found in
		3799	F<event.h> that are not directly supported by the libev core alone.
		3800
		3801	In standalone mode, libev will still try to automatically deduce the
		3802	configuration, but has to be more conservative.
		3803
		3804	=item EV_USE_MONOTONIC
		3805
		3806	If defined to be C<1>, libev will try to detect the availability of the
		3807	monotonic clock option at both compile time and runtime. Otherwise no
		3808	use of the monotonic clock option will be attempted. If you enable this,
		3809	you usually have to link against librt or something similar. Enabling it
		3810	when the functionality isn't available is safe, though, although you have
		3811	to make sure you link against any libraries where the C<clock_gettime>
		3812	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3813
		3814	=item EV_USE_REALTIME
		3815
		3816	If defined to be C<1>, libev will try to detect the availability of the
		3817	real-time clock option at compile time (and assume its availability
		3818	at runtime if successful). Otherwise no use of the real-time clock
		3819	option will be attempted. This effectively replaces C<gettimeofday>
		3820	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3821	correctness. See the note about libraries in the description of
		3822	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3823	C<EV_USE_CLOCK_SYSCALL>.
		3824
		3825	=item EV_USE_CLOCK_SYSCALL
		3826
		3827	If defined to be C<1>, libev will try to use a direct syscall instead
		3828	of calling the system-provided C<clock_gettime> function. This option
		3829	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3830	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3831	programs needlessly. Using a direct syscall is slightly slower (in
		3832	theory), because no optimised vdso implementation can be used, but avoids
		3833	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3834	higher, as it simplifies linking (no need for C<-lrt>).
		3835
		3836	=item EV_USE_NANOSLEEP
		3837
		3838	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3839	and will use it for delays. Otherwise it will use C<select ()>.
		3840
		3841	=item EV_USE_EVENTFD
		3842
		3843	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3844	available and will probe for kernel support at runtime. This will improve
		3845	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3846	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3847	2.7 or newer, otherwise disabled.
		3848
		3849	=item EV_USE_SELECT
		3850
		3851	If undefined or defined to be C<1>, libev will compile in support for the
		3852	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3853	other method takes over, select will be it. Otherwise the select backend
		3854	will not be compiled in.
		3855
		3856	=item EV_SELECT_USE_FD_SET
		3857
		3858	If defined to C<1>, then the select backend will use the system C<fd_set>
		3859	structure. This is useful if libev doesn't compile due to a missing
		3860	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3861	on exotic systems. This usually limits the range of file descriptors to
		3862	some low limit such as 1024 or might have other limitations (winsocket
		3863	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3864	configures the maximum size of the C<fd_set>.
		3865
		3866	=item EV_SELECT_IS_WINSOCKET
		3867
		3868	When defined to C<1>, the select backend will assume that
		3869	select/socket/connect etc. don't understand file descriptors but
		3870	wants osf handles on win32 (this is the case when the select to
		3871	be used is the winsock select). This means that it will call
		3872	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3873	it is assumed that all these functions actually work on fds, even
		3874	on win32. Should not be defined on non-win32 platforms.
		3875
		3876	=item EV_FD_TO_WIN32_HANDLE(fd)
		3877
		3878	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3879	file descriptors to socket handles. When not defining this symbol (the
		3880	default), then libev will call C<_get_osfhandle>, which is usually
		3881	correct. In some cases, programs use their own file descriptor management,
		3882	in which case they can provide this function to map fds to socket handles.
		3883
		3884	=item EV_WIN32_HANDLE_TO_FD(handle)
		3885
		3886	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3887	using the standard C<_open_osfhandle> function. For programs implementing
		3888	their own fd to handle mapping, overwriting this function makes it easier
		3889	to do so. This can be done by defining this macro to an appropriate value.
		3890
		3891	=item EV_WIN32_CLOSE_FD(fd)
		3892
		3893	If programs implement their own fd to handle mapping on win32, then this
		3894	macro can be used to override the C<close> function, useful to unregister
		3895	file descriptors again. Note that the replacement function has to close
		3896	the underlying OS handle.
		3897
		3898	=item EV_USE_POLL
		3899
		3900	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3901	backend. Otherwise it will be enabled on non-win32 platforms. It
		3902	takes precedence over select.
		3903
		3904	=item EV_USE_EPOLL
		3905
		3906	If defined to be C<1>, libev will compile in support for the Linux
		3907	C<epoll>(7) backend. Its availability will be detected at runtime,
		3908	otherwise another method will be used as fallback. This is the preferred
		3909	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3910	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3911
		3912	=item EV_USE_KQUEUE
		3913
		3914	If defined to be C<1>, libev will compile in support for the BSD style
		3915	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3916	otherwise another method will be used as fallback. This is the preferred
		3917	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3918	supports some types of fds correctly (the only platform we found that
		3919	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3920	not be used unless explicitly requested. The best way to use it is to find
		3921	out whether kqueue supports your type of fd properly and use an embedded
		3922	kqueue loop.
		3923
		3924	=item EV_USE_PORT
		3925
		3926	If defined to be C<1>, libev will compile in support for the Solaris
		3927	10 port style backend. Its availability will be detected at runtime,
		3928	otherwise another method will be used as fallback. This is the preferred
		3929	backend for Solaris 10 systems.
		3930
		3931	=item EV_USE_DEVPOLL
		3932
		3933	Reserved for future expansion, works like the USE symbols above.
		3934
		3935	=item EV_USE_INOTIFY
		3936
		3937	If defined to be C<1>, libev will compile in support for the Linux inotify
		3938	interface to speed up C<ev_stat> watchers. Its actual availability will
		3939	be detected at runtime. If undefined, it will be enabled if the headers
		3940	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3941
		3942	=item EV_ATOMIC_T
		3943
		3944	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3945	access is atomic with respect to other threads or signal contexts. No such
		3946	type is easily found in the C language, so you can provide your own type
		3947	that you know is safe for your purposes. It is used both for signal handler "locking"
		3948	as well as for signal and thread safety in C<ev_async> watchers.
		3949
		3950	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3951	(from F<signal.h>), which is usually good enough on most platforms.
		3952
		3953	=item EV_H (h)
		3954
		3955	The name of the F<ev.h> header file used to include it. The default if
		3956	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3957	used to virtually rename the F<ev.h> header file in case of conflicts.
		3958
		3959	=item EV_CONFIG_H (h)
		3960
		3961	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3962	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3963	C<EV_H>, above.
		3964
		3965	=item EV_EVENT_H (h)
		3966
		3967	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3968	of how the F<event.h> header can be found, the default is C<"event.h">.
		3969
		3970	=item EV_PROTOTYPES (h)
		3971
		3972	If defined to be C<0>, then F<ev.h> will not define any function
		3973	prototypes, but still define all the structs and other symbols. This is
		3974	occasionally useful if you want to provide your own wrapper functions
		3975	around libev functions.
		3976
		3977	=item EV_MULTIPLICITY
		3978
		3979	If undefined or defined to C<1>, then all event-loop-specific functions
		3980	will have the C<struct ev_loop *> as first argument, and you can create
		3981	additional independent event loops. Otherwise there will be no support
		3982	for multiple event loops and there is no first event loop pointer
		3983	argument. Instead, all functions act on the single default loop.
		3984
		3985	=item EV_MINPRI
		3986
		3987	=item EV_MAXPRI
		3988
		3989	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3990	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3991	provide for more priorities by overriding those symbols (usually defined
		3992	to be C<-2> and C<2>, respectively).
		3993
		3994	When doing priority-based operations, libev usually has to linearly search
		3995	all the priorities, so having many of them (hundreds) uses a lot of space
		3996	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3997	fine.
		3998
		3999	If your embedding application does not need any priorities, defining these
		4000	both to C<0> will save some memory and CPU.
		4001
		4002	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4003	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4004	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4005
		4006	If undefined or defined to be C<1> (and the platform supports it), then
		4007	the respective watcher type is supported. If defined to be C<0>, then it
		4008	is not. Disabling watcher types mainly saves code size.
		4009
		4010	=item EV_FEATURES
		4011
		4012	If you need to shave off some kilobytes of code at the expense of some
		4013	speed (but with the full API), you can define this symbol to request
		4014	certain subsets of functionality. The default is to enable all features
		4015	that can be enabled on the platform.
		4016
		4017	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4018	with some broad features you want) and then selectively re-enable
		4019	additional parts you want, for example if you want everything minimal,
		4020	but multiple event loop support, async and child watchers and the poll
		4021	backend, use this:
		4022
		4023	#define EV_FEATURES 0
		4024	#define EV_MULTIPLICITY 1
		4025	#define EV_USE_POLL 1
		4026	#define EV_CHILD_ENABLE 1
		4027	#define EV_ASYNC_ENABLE 1
		4028
		4029	The actual value is a bitset, it can be a combination of the following
		4030	values:
		4031
		4032	=over 4
		4033
		4034	=item C<1> - faster/larger code
		4035
		4036	Use larger code to speed up some operations.
		4037
		4038	Currently this is used to override some inlining decisions (enlarging the
		4039	code size by roughly 30% on amd64).
		4040
		4041	When optimising for size, use of compiler flags such as C<-Os> with
		4042	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4043	assertions.
		4044
		4045	=item C<2> - faster/larger data structures
		4046
		4047	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4048	hash table sizes and so on. This will usually further increase code size
		4049	and can additionally have an effect on the size of data structures at
		4050	runtime.
		4051
		4052	=item C<4> - full API configuration
		4053
		4054	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4055	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4056
		4057	=item C<8> - full API
		4058
		4059	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4060	details on which parts of the API are still available without this
		4061	feature, and do not complain if this subset changes over time.
		4062
		4063	=item C<16> - enable all optional watcher types
		4064
		4065	Enables all optional watcher types. If you want to selectively enable
		4066	only some watcher types other than I/O and timers (e.g. prepare,
		4067	embed, async, child...) you can enable them manually by defining
		4068	C<EV_watchertype_ENABLE> to C<1> instead.
		4069
		4070	=item C<32> - enable all backends
		4071
		4072	This enables all backends - without this feature, you need to enable at
		4073	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4074
		4075	=item C<64> - enable OS-specific "helper" APIs
		4076
		4077	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4078	default.
		4079
		4080	=back
		4081
		4082	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4083	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4084	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4085	watchers, timers and monotonic clock support.
		4086
		4087	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4088	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4089	your program might be left out as well - a binary starting a timer and an
		4090	I/O watcher then might come out at only 5Kb.
		4091
		4092	=item EV_AVOID_STDIO
		4093
		4094	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4095	functions (printf, scanf, perror etc.). This will increase the code size
		4096	somewhat, but if your program doesn't otherwise depend on stdio and your
		4097	libc allows it, this avoids linking in the stdio library which is quite
		4098	big.
		4099
		4100	Note that error messages might become less precise when this option is
		4101	enabled.
		4102
		4103	=item EV_NSIG
		4104
		4105	The highest supported signal number, +1 (or, the number of
		4106	signals): Normally, libev tries to deduce the maximum number of signals
		4107	automatically, but sometimes this fails, in which case it can be
		4108	specified. Also, using a lower number than detected (C<32> should be
		4109	good for about any system in existence) can save some memory, as libev
		4110	statically allocates some 12-24 bytes per signal number.
		4111
		4112	=item EV_PID_HASHSIZE
		4113
		4114	C<ev_child> watchers use a small hash table to distribute workload by
		4115	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4116	usually more than enough. If you need to manage thousands of children you
		4117	might want to increase this value (I<must> be a power of two).
		4118
		4119	=item EV_INOTIFY_HASHSIZE
		4120
		4121	C<ev_stat> watchers use a small hash table to distribute workload by
		4122	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4123	disabled), usually more than enough. If you need to manage thousands of
		4124	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4125	power of two).
		4126
		4127	=item EV_USE_4HEAP
		4128
		4129	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4130	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4131	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4132	faster performance with many (thousands) of watchers.
		4133
		4134	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4135	will be C<0>.
		4136
		4137	=item EV_HEAP_CACHE_AT
		4138
		4139	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4140	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4141	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4142	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4143	but avoids random read accesses on heap changes. This improves performance
		4144	noticeably with many (hundreds) of watchers.
		4145
		4146	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4147	will be C<0>.
		4148
		4149	=item EV_VERIFY
		4150
		4151	Controls how much internal verification (see C<ev_verify ()>) will
		4152	be done: If set to C<0>, no internal verification code will be compiled
		4153	in. If set to C<1>, then verification code will be compiled in, but not
		4154	called. If set to C<2>, then the internal verification code will be
		4155	called once per loop, which can slow down libev. If set to C<3>, then the
		4156	verification code will be called very frequently, which will slow down
		4157	libev considerably.
		4158
		4159	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4160	will be C<0>.
		4161
		4162	=item EV_COMMON
		4163
		4164	By default, all watchers have a C<void *data> member. By redefining
		4165	this macro to something else you can include more and other types of
		4166	members. You have to define it each time you include one of the files,
		4167	though, and it must be identical each time.
		4168
		4169	For example, the perl EV module uses something like this:
		4170
		4171	#define EV_COMMON \
		4172	SV *self; /* contains this struct */ \
		4173	SV *cb_sv, *fh /* note no trailing ";" */
		4174
		4175	=item EV_CB_DECLARE (type)
		4176
		4177	=item EV_CB_INVOKE (watcher, revents)
		4178
		4179	=item ev_set_cb (ev, cb)
		4180
		4181	Can be used to change the callback member declaration in each watcher,
		4182	and the way callbacks are invoked and set. Must expand to a struct member
		4183	definition and a statement, respectively. See the F<ev.h> header file for
		4184	their default definitions. One possible use for overriding these is to
		4185	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4186	method calls instead of plain function calls in C++.
		4187
		4188	=back
		4189
		4190	=head2 EXPORTED API SYMBOLS
		4191
		4192	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4193	exported symbols, you can use the provided F<Symbol.*> files which list
		4194	all public symbols, one per line:
		4195
		4196	Symbols.ev for libev proper
		4197	Symbols.event for the libevent emulation
		4198
		4199	This can also be used to rename all public symbols to avoid clashes with
		4200	multiple versions of libev linked together (which is obviously bad in
		4201	itself, but sometimes it is inconvenient to avoid this).
		4202
		4203	A sed command like this will create wrapper C<#define>'s that you need to
		4204	include before including F<ev.h>:
		4205
		4206	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4207
		4208	This would create a file F<wrap.h> which essentially looks like this:
		4209
		4210	#define ev_backend myprefix_ev_backend
		4211	#define ev_check_start myprefix_ev_check_start
		4212	#define ev_check_stop myprefix_ev_check_stop
		4213	...
		4214
		4215	=head2 EXAMPLES
		4216
		4217	For a real-world example of a program the includes libev
		4218	verbatim, you can have a look at the EV perl module
		4219	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4220	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4221	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4222	will be compiled. It is pretty complex because it provides its own header
		4223	file.
		4224
		4225	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4226	that everybody includes and which overrides some configure choices:
		4227
		4228	#define EV_FEATURES 8
		4229	#define EV_USE_SELECT 1
		4230	#define EV_PREPARE_ENABLE 1
		4231	#define EV_IDLE_ENABLE 1
		4232	#define EV_SIGNAL_ENABLE 1
		4233	#define EV_CHILD_ENABLE 1
		4234	#define EV_USE_STDEXCEPT 0
		4235	#define EV_CONFIG_H <config.h>
		4236
		4237	#include "ev++.h"
		4238
		4239	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4240
		4241	#include "ev_cpp.h"
		4242	#include "ev.c"
		4243
		4244	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4245
		4246	=head2 THREADS AND COROUTINES
		4247
		4248	=head3 THREADS
		4249
		4250	All libev functions are reentrant and thread-safe unless explicitly
		4251	documented otherwise, but libev implements no locking itself. This means
		4252	that you can use as many loops as you want in parallel, as long as there
		4253	are no concurrent calls into any libev function with the same loop
		4254	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4255	of course): libev guarantees that different event loops share no data
		4256	structures that need any locking.
		4257
		4258	Or to put it differently: calls with different loop parameters can be done
		4259	concurrently from multiple threads, calls with the same loop parameter
		4260	must be done serially (but can be done from different threads, as long as
		4261	only one thread ever is inside a call at any point in time, e.g. by using
		4262	a mutex per loop).
		4263
		4264	Specifically to support threads (and signal handlers), libev implements
		4265	so-called C<ev_async> watchers, which allow some limited form of
		4266	concurrency on the same event loop, namely waking it up "from the
		4267	outside".
		4268
		4269	If you want to know which design (one loop, locking, or multiple loops
		4270	without or something else still) is best for your problem, then I cannot
		4271	help you, but here is some generic advice:
		4272
		4273	=over 4
		4274
		4275	=item * most applications have a main thread: use the default libev loop
		4276	in that thread, or create a separate thread running only the default loop.
		4277
		4278	This helps integrating other libraries or software modules that use libev
		4279	themselves and don't care/know about threading.
		4280
		4281	=item * one loop per thread is usually a good model.
		4282
		4283	Doing this is almost never wrong, sometimes a better-performance model
		4284	exists, but it is always a good start.
		4285
		4286	=item * other models exist, such as the leader/follower pattern, where one
		4287	loop is handed through multiple threads in a kind of round-robin fashion.
		4288
		4289	Choosing a model is hard - look around, learn, know that usually you can do
		4290	better than you currently do :-)
		4291
		4292	=item * often you need to talk to some other thread which blocks in the
		4293	event loop.
		4294
		4295	C<ev_async> watchers can be used to wake them up from other threads safely
		4296	(or from signal contexts...).
		4297
		4298	An example use would be to communicate signals or other events that only
		4299	work in the default loop by registering the signal watcher with the
		4300	default loop and triggering an C<ev_async> watcher from the default loop
		4301	watcher callback into the event loop interested in the signal.
		4302
		4303	=back
		4304
		4305	=head4 THREAD LOCKING EXAMPLE
		4306
		4307	Here is a fictitious example of how to run an event loop in a different
		4308	thread than where callbacks are being invoked and watchers are
		4309	created/added/removed.
		4310
		4311	For a real-world example, see the C<EV::Loop::Async> perl module,
		4312	which uses exactly this technique (which is suited for many high-level
		4313	languages).
		4314
		4315	The example uses a pthread mutex to protect the loop data, a condition
		4316	variable to wait for callback invocations, an async watcher to notify the
		4317	event loop thread and an unspecified mechanism to wake up the main thread.
		4318
		4319	First, you need to associate some data with the event loop:
		4320
		4321	typedef struct {
		4322	mutex_t lock; /* global loop lock */
		4323	ev_async async_w;
		4324	thread_t tid;
		4325	cond_t invoke_cv;
		4326	} userdata;
		4327
		4328	void prepare_loop (EV_P)
		4329	{
		4330	// for simplicity, we use a static userdata struct.
		4331	static userdata u;
		4332
		4333	ev_async_init (&u->async_w, async_cb);
		4334	ev_async_start (EV_A_ &u->async_w);
		4335
		4336	pthread_mutex_init (&u->lock, 0);
		4337	pthread_cond_init (&u->invoke_cv, 0);
		4338
		4339	// now associate this with the loop
		4340	ev_set_userdata (EV_A_ u);
		4341	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4342	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4343
		4344	// then create the thread running ev_loop
		4345	pthread_create (&u->tid, 0, l_run, EV_A);
		4346	}
		4347
		4348	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4349	solely to wake up the event loop so it takes notice of any new watchers
		4350	that might have been added:
		4351
		4352	static void
		4353	async_cb (EV_P_ ev_async *w, int revents)
		4354	{
		4355	// just used for the side effects
		4356	}
		4357
		4358	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4359	protecting the loop data, respectively.
		4360
		4361	static void
		4362	l_release (EV_P)
		4363	{
		4364	userdata *u = ev_userdata (EV_A);
		4365	pthread_mutex_unlock (&u->lock);
		4366	}
		4367
		4368	static void
		4369	l_acquire (EV_P)
		4370	{
		4371	userdata *u = ev_userdata (EV_A);
		4372	pthread_mutex_lock (&u->lock);
		4373	}
		4374
		4375	The event loop thread first acquires the mutex, and then jumps straight
		4376	into C<ev_run>:
		4377
		4378	void *
		4379	l_run (void *thr_arg)
		4380	{
		4381	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4382
		4383	l_acquire (EV_A);
		4384	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4385	ev_run (EV_A_ 0);
		4386	l_release (EV_A);
		4387
		4388	return 0;
		4389	}
		4390
		4391	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4392	signal the main thread via some unspecified mechanism (signals? pipe
		4393	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4394	have been called (in a while loop because a) spurious wakeups are possible
		4395	and b) skipping inter-thread-communication when there are no pending
		4396	watchers is very beneficial):
		4397
		4398	static void
		4399	l_invoke (EV_P)
		4400	{
		4401	userdata *u = ev_userdata (EV_A);
		4402
		4403	while (ev_pending_count (EV_A))
		4404	{
		4405	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4406	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4407	}
		4408	}
		4409
		4410	Now, whenever the main thread gets told to invoke pending watchers, it
		4411	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4412	thread to continue:
		4413
		4414	static void
		4415	real_invoke_pending (EV_P)
		4416	{
		4417	userdata *u = ev_userdata (EV_A);
		4418
		4419	pthread_mutex_lock (&u->lock);
		4420	ev_invoke_pending (EV_A);
		4421	pthread_cond_signal (&u->invoke_cv);
		4422	pthread_mutex_unlock (&u->lock);
		4423	}
		4424
		4425	Whenever you want to start/stop a watcher or do other modifications to an
		4426	event loop, you will now have to lock:
		4427
		4428	ev_timer timeout_watcher;
		4429	userdata *u = ev_userdata (EV_A);
		4430
		4431	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4432
		4433	pthread_mutex_lock (&u->lock);
		4434	ev_timer_start (EV_A_ &timeout_watcher);
		4435	ev_async_send (EV_A_ &u->async_w);
		4436	pthread_mutex_unlock (&u->lock);
		4437
		4438	Note that sending the C<ev_async> watcher is required because otherwise
		4439	an event loop currently blocking in the kernel will have no knowledge
		4440	about the newly added timer. By waking up the loop it will pick up any new
		4441	watchers in the next event loop iteration.
		4442
		4443	=head3 COROUTINES
		4444
		4445	Libev is very accommodating to coroutines ("cooperative threads"):
		4446	libev fully supports nesting calls to its functions from different
		4447	coroutines (e.g. you can call C<ev_run> on the same loop from two
		4448	different coroutines, and switch freely between both coroutines running
		4449	the loop, as long as you don't confuse yourself). The only exception is
		4450	that you must not do this from C<ev_periodic> reschedule callbacks.
		4451
		4452	Care has been taken to ensure that libev does not keep local state inside
		4453	C<ev_run>, and other calls do not usually allow for coroutine switches as
		4454	they do not call any callbacks.
		4455
		4456	=head2 COMPILER WARNINGS
		4457
		4458	Depending on your compiler and compiler settings, you might get no or a
		4459	lot of warnings when compiling libev code. Some people are apparently
		4460	scared by this.
		4461
		4462	However, these are unavoidable for many reasons. For one, each compiler
		4463	has different warnings, and each user has different tastes regarding
		4464	warning options. "Warn-free" code therefore cannot be a goal except when
		4465	targeting a specific compiler and compiler-version.
		4466
		4467	Another reason is that some compiler warnings require elaborate
		4468	workarounds, or other changes to the code that make it less clear and less
		4469	maintainable.
		4470
		4471	And of course, some compiler warnings are just plain stupid, or simply
		4472	wrong (because they don't actually warn about the condition their message
		4473	seems to warn about). For example, certain older gcc versions had some
		4474	warnings that resulted in an extreme number of false positives. These have
		4475	been fixed, but some people still insist on making code warn-free with
		4476	such buggy versions.
		4477
		4478	While libev is written to generate as few warnings as possible,
		4479	"warn-free" code is not a goal, and it is recommended not to build libev
		4480	with any compiler warnings enabled unless you are prepared to cope with
		4481	them (e.g. by ignoring them). Remember that warnings are just that:
		4482	warnings, not errors, or proof of bugs.
		4483
		4484
		4485	=head2 VALGRIND
		4486
		4487	Valgrind has a special section here because it is a popular tool that is
		4488	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4489
		4490	If you think you found a bug (memory leak, uninitialised data access etc.)
		4491	in libev, then check twice: If valgrind reports something like:
		4492
		4493	==2274== definitely lost: 0 bytes in 0 blocks.
		4494	==2274== possibly lost: 0 bytes in 0 blocks.
		4495	==2274== still reachable: 256 bytes in 1 blocks.
		4496
		4497	Then there is no memory leak, just as memory accounted to global variables
		4498	is not a memleak - the memory is still being referenced, and didn't leak.
		4499
		4500	Similarly, under some circumstances, valgrind might report kernel bugs
		4501	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4502	although an acceptable workaround has been found here), or it might be
		4503	confused.
		4504
		4505	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4506	make it into some kind of religion.
		4507
		4508	If you are unsure about something, feel free to contact the mailing list
		4509	with the full valgrind report and an explanation on why you think this
		4510	is a bug in libev (best check the archives, too :). However, don't be
		4511	annoyed when you get a brisk "this is no bug" answer and take the chance
		4512	of learning how to interpret valgrind properly.
		4513
		4514	If you need, for some reason, empty reports from valgrind for your project
		4515	I suggest using suppression lists.
		4516
		4517
		4518	=head1 PORTABILITY NOTES
		4519
		4520	=head2 GNU/LINUX 32 BIT LIMITATIONS
		4521
		4522	GNU/Linux is the only common platform that supports 64 bit file/large file
		4523	interfaces but I<disables> them by default.
		4524
		4525	That means that libev compiled in the default environment doesn't support
		4526	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		4527
		4528	Unfortunately, many programs try to work around this GNU/Linux issue
		4529	by enabling the large file API, which makes them incompatible with the
		4530	standard libev compiled for their system.
		4531
		4532	Likewise, libev cannot enable the large file API itself as this would
		4533	suddenly make it incompatible to the default compile time environment,
		4534	i.e. all programs not using special compile switches.
		4535
		4536	=head2 OS/X AND DARWIN BUGS
		4537
		4538	The whole thing is a bug if you ask me - basically any system interface
		4539	you touch is broken, whether it is locales, poll, kqueue or even the
		4540	OpenGL drivers.
		4541
		4542	=head3 C<kqueue> is buggy
		4543
		4544	The kqueue syscall is broken in all known versions - most versions support
		4545	only sockets, many support pipes.
		4546
		4547	Libev tries to work around this by not using C<kqueue> by default on this
		4548	rotten platform, but of course you can still ask for it when creating a
		4549	loop - embedding a socket-only kqueue loop into a select-based one is
		4550	probably going to work well.
		4551
		4552	=head3 C<poll> is buggy
		4553
		4554	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		4555	implementation by something calling C<kqueue> internally around the 10.5.6
		4556	release, so now C<kqueue> I<and> C<poll> are broken.
		4557
		4558	Libev tries to work around this by not using C<poll> by default on
		4559	this rotten platform, but of course you can still ask for it when creating
		4560	a loop.
		4561
		4562	=head3 C<select> is buggy
		4563
		4564	All that's left is C<select>, and of course Apple found a way to fuck this
		4565	one up as well: On OS/X, C<select> actively limits the number of file
		4566	descriptors you can pass in to 1024 - your program suddenly crashes when
		4567	you use more.
		4568
		4569	There is an undocumented "workaround" for this - defining
		4570	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		4571	work on OS/X.
		4572
		4573	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		4574
		4575	=head3 C<errno> reentrancy
		4576
		4577	The default compile environment on Solaris is unfortunately so
		4578	thread-unsafe that you can't even use components/libraries compiled
		4579	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		4580	defined by default. A valid, if stupid, implementation choice.
		4581
		4582	If you want to use libev in threaded environments you have to make sure
		4583	it's compiled with C<_REENTRANT> defined.
		4584
		4585	=head3 Event port backend
		4586
		4587	The scalable event interface for Solaris is called "event
		4588	ports". Unfortunately, this mechanism is very buggy in all major
		4589	releases. If you run into high CPU usage, your program freezes or you get
		4590	a large number of spurious wakeups, make sure you have all the relevant
		4591	and latest kernel patches applied. No, I don't know which ones, but there
		4592	are multiple ones to apply, and afterwards, event ports actually work
		4593	great.
		4594
		4595	If you can't get it to work, you can try running the program by setting
		4596	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		4597	C<select> backends.
		4598
		4599	=head2 AIX POLL BUG
		4600
		4601	AIX unfortunately has a broken C<poll.h> header. Libev works around
		4602	this by trying to avoid the poll backend altogether (i.e. it's not even
		4603	compiled in), which normally isn't a big problem as C<select> works fine
		4604	with large bitsets on AIX, and AIX is dead anyway.
		4605
		4606	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4607
		4608	=head3 General issues
		4609
		4610	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4611	requires, and its I/O model is fundamentally incompatible with the POSIX
		4612	model. Libev still offers limited functionality on this platform in
		4613	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4614	descriptors. This only applies when using Win32 natively, not when using
		4615	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		4616	as every compielr comes with a slightly differently broken/incompatible
		4617	environment.
		4618
		4619	Lifting these limitations would basically require the full
		4620	re-implementation of the I/O system. If you are into this kind of thing,
		4621	then note that glib does exactly that for you in a very portable way (note
		4622	also that glib is the slowest event library known to man).
		4623
		4624	There is no supported compilation method available on windows except
		4625	embedding it into other applications.
		4626
		4627	Sensible signal handling is officially unsupported by Microsoft - libev
		4628	tries its best, but under most conditions, signals will simply not work.
		4629
		4630	Not a libev limitation but worth mentioning: windows apparently doesn't
		4631	accept large writes: instead of resulting in a partial write, windows will
		4632	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4633	so make sure you only write small amounts into your sockets (less than a
		4634	megabyte seems safe, but this apparently depends on the amount of memory
		4635	available).
		4636
		4637	Due to the many, low, and arbitrary limits on the win32 platform and
		4638	the abysmal performance of winsockets, using a large number of sockets
		4639	is not recommended (and not reasonable). If your program needs to use
		4640	more than a hundred or so sockets, then likely it needs to use a totally
		4641	different implementation for windows, as libev offers the POSIX readiness
		4642	notification model, which cannot be implemented efficiently on windows
		4643	(due to Microsoft monopoly games).
		4644
		4645	A typical way to use libev under windows is to embed it (see the embedding
		4646	section for details) and use the following F<evwrap.h> header file instead
		4647	of F<ev.h>:
		4648
		4649	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4650	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4651
		4652	#include "ev.h"
		4653
		4654	And compile the following F<evwrap.c> file into your project (make sure
		4655	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4656
		4657	#include "evwrap.h"
		4658	#include "ev.c"
		4659
		4660	=head3 The winsocket C<select> function
		4661
		4662	The winsocket C<select> function doesn't follow POSIX in that it
		4663	requires socket I<handles> and not socket I<file descriptors> (it is
		4664	also extremely buggy). This makes select very inefficient, and also
		4665	requires a mapping from file descriptors to socket handles (the Microsoft
		4666	C runtime provides the function C<_open_osfhandle> for this). See the
		4667	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4668	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4669
		4670	The configuration for a "naked" win32 using the Microsoft runtime
		4671	libraries and raw winsocket select is:
		4672
		4673	#define EV_USE_SELECT 1
		4674	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4675
		4676	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4677	complexity in the O(n²) range when using win32.
		4678
		4679	=head3 Limited number of file descriptors
		4680
		4681	Windows has numerous arbitrary (and low) limits on things.
		4682
		4683	Early versions of winsocket's select only supported waiting for a maximum
		4684	of C<64> handles (probably owning to the fact that all windows kernels
		4685	can only wait for C<64> things at the same time internally; Microsoft
		4686	recommends spawning a chain of threads and wait for 63 handles and the
		4687	previous thread in each. Sounds great!).
		4688
		4689	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4690	to some high number (e.g. C<2048>) before compiling the winsocket select
		4691	call (which might be in libev or elsewhere, for example, perl and many
		4692	other interpreters do their own select emulation on windows).
		4693
		4694	Another limit is the number of file descriptors in the Microsoft runtime
		4695	libraries, which by default is C<64> (there must be a hidden I<64>
		4696	fetish or something like this inside Microsoft). You can increase this
		4697	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4698	(another arbitrary limit), but is broken in many versions of the Microsoft
		4699	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4700	(depending on windows version and/or the phase of the moon). To get more,
		4701	you need to wrap all I/O functions and provide your own fd management, but
		4702	the cost of calling select (O(n²)) will likely make this unworkable.
		4703
		4704	=head2 PORTABILITY REQUIREMENTS
		4705
		4706	In addition to a working ISO-C implementation and of course the
		4707	backend-specific APIs, libev relies on a few additional extensions:
		4708
		4709	=over 4
		4710
		4711	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4712	calling conventions regardless of C<ev_watcher_type *>.
		4713
		4714	Libev assumes not only that all watcher pointers have the same internal
		4715	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4716	assumes that the same (machine) code can be used to call any watcher
		4717	callback: The watcher callbacks have different type signatures, but libev
		4718	calls them using an C<ev_watcher *> internally.
		4719
		4720	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4721
		4722	The type C<sig_atomic_t volatile> (or whatever is defined as
		4723	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4724	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4725	believed to be sufficiently portable.
		4726
		4727	=item C<sigprocmask> must work in a threaded environment
		4728
		4729	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4730	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4731	pthread implementations will either allow C<sigprocmask> in the "main
		4732	thread" or will block signals process-wide, both behaviours would
		4733	be compatible with libev. Interaction between C<sigprocmask> and
		4734	C<pthread_sigmask> could complicate things, however.
		4735
		4736	The most portable way to handle signals is to block signals in all threads
		4737	except the initial one, and run the default loop in the initial thread as
		4738	well.
		4739
		4740	=item C<long> must be large enough for common memory allocation sizes
		4741
		4742	To improve portability and simplify its API, libev uses C<long> internally
		4743	instead of C<size_t> when allocating its data structures. On non-POSIX
		4744	systems (Microsoft...) this might be unexpectedly low, but is still at
		4745	least 31 bits everywhere, which is enough for hundreds of millions of
		4746	watchers.
		4747
		4748	=item C<double> must hold a time value in seconds with enough accuracy
		4749
		4750	The type C<double> is used to represent timestamps. It is required to
		4751	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		4752	good enough for at least into the year 4000 with millisecond accuracy
		4753	(the design goal for libev). This requirement is overfulfilled by
		4754	implementations using IEEE 754, which is basically all existing ones. With
		4755	IEEE 754 doubles, you get microsecond accuracy until at least 2200.
		4756
		4757	=back
		4758
		4759	If you know of other additional requirements drop me a note.
		4760
		4761
		4762	=head1 ALGORITHMIC COMPLEXITIES
		4763
		4764	In this section the complexities of (many of) the algorithms used inside
		4765	libev will be documented. For complexity discussions about backends see
		4766	the documentation for C<ev_default_init>.
		4767
		4768	All of the following are about amortised time: If an array needs to be
		4769	extended, libev needs to realloc and move the whole array, but this
		4770	happens asymptotically rarer with higher number of elements, so O(1) might
		4771	mean that libev does a lengthy realloc operation in rare cases, but on
		4772	average it is much faster and asymptotically approaches constant time.
		4773
		4774	=over 4
		4775
		4776	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4777
		4778	This means that, when you have a watcher that triggers in one hour and
		4779	there are 100 watchers that would trigger before that, then inserting will
		4780	have to skip roughly seven (C<ld 100>) of these watchers.
		4781
		4782	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4783
		4784	That means that changing a timer costs less than removing/adding them,
		4785	as only the relative motion in the event queue has to be paid for.
		4786
		4787	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4788
		4789	These just add the watcher into an array or at the head of a list.
		4790
		4791	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4792
		4793	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4794
		4795	These watchers are stored in lists, so they need to be walked to find the
		4796	correct watcher to remove. The lists are usually short (you don't usually
		4797	have many watchers waiting for the same fd or signal: one is typical, two
		4798	is rare).
		4799
		4800	=item Finding the next timer in each loop iteration: O(1)
		4801
		4802	By virtue of using a binary or 4-heap, the next timer is always found at a
		4803	fixed position in the storage array.
		4804
		4805	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4806
		4807	A change means an I/O watcher gets started or stopped, which requires
		4808	libev to recalculate its status (and possibly tell the kernel, depending
		4809	on backend and whether C<ev_io_set> was used).
		4810
		4811	=item Activating one watcher (putting it into the pending state): O(1)
		4812
		4813	=item Priority handling: O(number_of_priorities)
		4814
		4815	Priorities are implemented by allocating some space for each
		4816	priority. When doing priority-based operations, libev usually has to
		4817	linearly search all the priorities, but starting/stopping and activating
		4818	watchers becomes O(1) with respect to priority handling.
		4819
		4820	=item Sending an ev_async: O(1)
		4821
		4822	=item Processing ev_async_send: O(number_of_async_watchers)
		4823
		4824	=item Processing signals: O(max_signal_number)
		4825
		4826	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4827	calls in the current loop iteration. Checking for async and signal events
		4828	involves iterating over all running async watchers or all signal numbers.
		4829
		4830	=back
		4831
		4832
		4833	=head1 PORTING FROM LIBEV 3.X TO 4.X
		4834
		4835	The major version 4 introduced some minor incompatible changes to the API.
		4836
		4837	At the moment, the C<ev.h> header file tries to implement superficial
		4838	compatibility, so most programs should still compile. Those might be
		4839	removed in later versions of libev, so better update early than late.
		4840
		4841	=over 4
		4842
		4843	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		4844
		4845	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		4846
		4847	ev_loop_destroy (EV_DEFAULT);
		4848	ev_loop_fork (EV_DEFAULT);
		4849
		4850	=item function/symbol renames
		4851
		4852	A number of functions and symbols have been renamed:
		4853
		4854	ev_loop => ev_run
		4855	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		4856	EVLOOP_ONESHOT => EVRUN_ONCE
		4857
		4858	ev_unloop => ev_break
		4859	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		4860	EVUNLOOP_ONE => EVBREAK_ONE
		4861	EVUNLOOP_ALL => EVBREAK_ALL
		4862
		4863	EV_TIMEOUT => EV_TIMER
		4864
		4865	ev_loop_count => ev_iteration
		4866	ev_loop_depth => ev_depth
		4867	ev_loop_verify => ev_verify
		4868
		4869	Most functions working on C<struct ev_loop> objects don't have an
		4870	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		4871	associated constants have been renamed to not collide with the C<struct
		4872	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		4873	as all other watcher types. Note that C<ev_loop_fork> is still called
		4874	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		4875	typedef.
		4876
		4877	=item C<EV_COMPAT3> backwards compatibility mechanism
		4878
		4879	The backward compatibility mechanism can be controlled by
		4880	C<EV_COMPAT3>. See L<PREPROCESSOR SYMBOLS/MACROS> in the L<EMBEDDING>
		4881	section.
		4882
		4883	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		4884
		4885	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		4886	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		4887	and work, but the library code will of course be larger.
		4888
		4889	=back
		4890
		4891
		4892	=head1 GLOSSARY
		4893
		4894	=over 4
		4895
		4896	=item active
		4897
		4898	A watcher is active as long as it has been started and not yet stopped.
		4899	See L<WATCHER STATES> for details.
		4900
		4901	=item application
		4902
		4903	In this document, an application is whatever is using libev.
		4904
		4905	=item backend
		4906
		4907	The part of the code dealing with the operating system interfaces.
		4908
		4909	=item callback
		4910
		4911	The address of a function that is called when some event has been
		4912	detected. Callbacks are being passed the event loop, the watcher that
		4913	received the event, and the actual event bitset.
		4914
		4915	=item callback/watcher invocation
		4916
		4917	The act of calling the callback associated with a watcher.
		4918
		4919	=item event
		4920
		4921	A change of state of some external event, such as data now being available
		4922	for reading on a file descriptor, time having passed or simply not having
		4923	any other events happening anymore.
		4924
		4925	In libev, events are represented as single bits (such as C<EV_READ> or
		4926	C<EV_TIMER>).
		4927
		4928	=item event library
		4929
		4930	A software package implementing an event model and loop.
		4931
		4932	=item event loop
		4933
		4934	An entity that handles and processes external events and converts them
		4935	into callback invocations.
		4936
		4937	=item event model
		4938
		4939	The model used to describe how an event loop handles and processes
		4940	watchers and events.
		4941
		4942	=item pending
		4943
		4944	A watcher is pending as soon as the corresponding event has been
		4945	detected. See L<WATCHER STATES> for details.
		4946
		4947	=item real time
		4948
		4949	The physical time that is observed. It is apparently strictly monotonic :)
		4950
		4951	=item wall-clock time
		4952
		4953	The time and date as shown on clocks. Unlike real time, it can actually
		4954	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4955	clock.
		4956
		4957	=item watcher
		4958
		4959	A data structure that describes interest in certain events. Watchers need
		4960	to be started (attached to an event loop) before they can receive events.
		4961
		4962	=back
878		4963
879	=head1 AUTHOR	4964	=head1 AUTHOR
880		4965
881	Marc Lehmann <libev@schmorp.de>.	4966	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
882		4967

Diff Legend

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