ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.16 by root, Mon Nov 12 08:47:14 2007 UTC vs.
Revision 1.262 by root, Sat Jul 25 10:14:34 2009 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_loop's to stop iterating
		32	ev_unloop (EV_A_ EVUNLOOP_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_loop to stop iterating
		41	ev_unloop (EV_A_ EVUNLOOP_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_loop (0);
		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_loop (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	Familarity 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<ev_loop *>) will not have
42	will not have this argument.	124	this argument.
43		125
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	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 (somewhere
48	the beginning of 1970, details are complicated, don't ask). This type is	130	near the beginning of 1970, details are complicated, don't ask). This
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	131	type is called C<ev_tstamp>, which is what you should use too. It usually
50	to the double type in C.	132	aliases to the C<double> type in C. When you need to do any calculations
		133	on it, you should treat it as some floating point value. Unlike the name
		134	component C<stamp> might indicate, it is also used for time differences
		135	throughout libev.
		136
		137	=head1 ERROR HANDLING
		138
		139	Libev knows three classes of errors: operating system errors, usage errors
		140	and internal errors (bugs).
		141
		142	When libev catches an operating system error it cannot handle (for example
		143	a system call indicating a condition libev cannot fix), it calls the callback
		144	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		145	abort. The default is to print a diagnostic message and to call C<abort
		146	()>.
		147
		148	When libev detects a usage error such as a negative timer interval, then
		149	it will print a diagnostic message and abort (via the C<assert> mechanism,
		150	so C<NDEBUG> will disable this checking): these are programming errors in
		151	the libev caller and need to be fixed there.
		152
		153	Libev also has a few internal error-checking C<assert>ions, and also has
		154	extensive consistency checking code. These do not trigger under normal
		155	circumstances, as they indicate either a bug in libev or worse.
		156
		157
		158	=head1 GLOBAL FUNCTIONS
		159
		160	These functions can be called anytime, even before initialising the
		161	library in any way.
51		162
52	=over 4	163	=over 4
53		164
54	=item ev_tstamp ev_time ()	165	=item ev_tstamp ev_time ()
55		166
56	Returns the current time as libev would use it.	167	Returns the current time as libev would use it. Please note that the
		168	C<ev_now> function is usually faster and also often returns the timestamp
		169	you actually want to know.
		170
		171	=item ev_sleep (ev_tstamp interval)
		172
		173	Sleep for the given interval: The current thread will be blocked until
		174	either it is interrupted or the given time interval has passed. Basically
		175	this is a sub-second-resolution C<sleep ()>.
57		176
58	=item int ev_version_major ()	177	=item int ev_version_major ()
59		178
60	=item int ev_version_minor ()	179	=item int ev_version_minor ()
61		180
62	You can find out the major and minor version numbers of the library	181	You can find out the major and minor ABI version numbers of the library
63	you linked against by calling the functions C<ev_version_major> and	182	you linked against by calling the functions C<ev_version_major> and
64	C<ev_version_minor>. If you want, you can compare against the global	183	C<ev_version_minor>. If you want, you can compare against the global
65	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	184	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
66	version of the library your program was compiled against.	185	version of the library your program was compiled against.
67		186
		187	These version numbers refer to the ABI version of the library, not the
		188	release version.
		189
68	Usually, it's a good idea to terminate if the major versions mismatch,	190	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	191	as this indicates an incompatible change. Minor versions are usually
70	compatible to older versions, so a larger minor version alone is usually	192	compatible to older versions, so a larger minor version alone is usually
71	not a problem.	193	not a problem.
72		194
		195	Example: Make sure we haven't accidentally been linked against the wrong
		196	version.
		197
		198	assert (("libev version mismatch",
		199	ev_version_major () == EV_VERSION_MAJOR
		200	&& ev_version_minor () >= EV_VERSION_MINOR));
		201
		202	=item unsigned int ev_supported_backends ()
		203
		204	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		205	value) compiled into this binary of libev (independent of their
		206	availability on the system you are running on). See C<ev_default_loop> for
		207	a description of the set values.
		208
		209	Example: make sure we have the epoll method, because yeah this is cool and
		210	a must have and can we have a torrent of it please!!!11
		211
		212	assert (("sorry, no epoll, no sex",
		213	ev_supported_backends () & EVBACKEND_EPOLL));
		214
		215	=item unsigned int ev_recommended_backends ()
		216
		217	Return the set of all backends compiled into this binary of libev and also
		218	recommended for this platform. This set is often smaller than the one
		219	returned by C<ev_supported_backends>, as for example kqueue is broken on
		220	most BSDs and will not be auto-detected unless you explicitly request it
		221	(assuming you know what you are doing). This is the set of backends that
		222	libev will probe for if you specify no backends explicitly.
		223
		224	=item unsigned int ev_embeddable_backends ()
		225
		226	Returns the set of backends that are embeddable in other event loops. This
		227	is the theoretical, all-platform, value. To find which backends
		228	might be supported on the current system, you would need to look at
		229	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		230	recommended ones.
		231
		232	See the description of C<ev_embed> watchers for more info.
		233
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	234	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
74		235
75	Sets the allocation function to use (the prototype is similar to the	236	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	237	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
77	and free memory (no surprises here). If it returns zero when memory	238	used to allocate and free memory (no surprises here). If it returns zero
78	needs to be allocated, the library might abort or take some potentially	239	when memory needs to be allocated (C<size != 0>), the library might abort
79	destructive action. The default is your system realloc function.	240	or take some potentially destructive action.
		241
		242	Since some systems (at least OpenBSD and Darwin) fail to implement
		243	correct C<realloc> semantics, libev will use a wrapper around the system
		244	C<realloc> and C<free> functions by default.
80		245
81	You could override this function in high-availability programs to, say,	246	You could override this function in high-availability programs to, say,
82	free some memory if it cannot allocate memory, to use a special allocator,	247	free some memory if it cannot allocate memory, to use a special allocator,
83	or even to sleep a while and retry until some memory is available.	248	or even to sleep a while and retry until some memory is available.
84		249
		250	Example: Replace the libev allocator with one that waits a bit and then
		251	retries (example requires a standards-compliant C<realloc>).
		252
		253	static void *
		254	persistent_realloc (void *ptr, size_t size)
		255	{
		256	for (;;)
		257	{
		258	void *newptr = realloc (ptr, size);
		259
		260	if (newptr)
		261	return newptr;
		262
		263	sleep (60);
		264	}
		265	}
		266
		267	...
		268	ev_set_allocator (persistent_realloc);
		269
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	270	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
86		271
87	Set the callback function to call on a retryable syscall error (such	272	Set the callback function to call on a retryable system call error (such
88	as failed select, poll, epoll_wait). The message is a printable string	273	as failed select, poll, epoll_wait). The message is a printable string
89	indicating the system call or subsystem causing the problem. If this	274	indicating the system call or subsystem causing the problem. If this
90	callback is set, then libev will expect it to remedy the sitution, no	275	callback is set, then libev will expect it to remedy the situation, no
91	matter what, when it returns. That is, libev will generally retry the	276	matter what, when it returns. That is, libev will generally retry the
92	requested operation, or, if the condition doesn't go away, do bad stuff	277	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	278	(such as abort).
94		279
		280	Example: This is basically the same thing that libev does internally, too.
		281
		282	static void
		283	fatal_error (const char *msg)
		284	{
		285	perror (msg);
		286	abort ();
		287	}
		288
		289	...
		290	ev_set_syserr_cb (fatal_error);
		291
95	=back	292	=back
96		293
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	294	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
98		295
99	An event loop is described by a C<struct ev_loop *>. The library knows two	296	An event loop is described by a C<struct ev_loop *> (the C<struct>
100	types of such loops, the I<default> loop, which supports signals and child	297	is I<not> optional in this case, as there is also an C<ev_loop>
101	events, and dynamically created loops which do not.	298	I<function>).
102		299
103	If you use threads, a common model is to run the default event loop	300	The library knows two types of such loops, the I<default> loop, which
104	in your main thread (or in a separate thrad) and for each thread you	301	supports signals and child events, and dynamically created loops which do
105	create, you also create another event loop. Libev itself does no locking	302	not.
106	whatsoever, so if you mix calls to the same event loop in different
107	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because it's hideous and inefficient).
109		303
110	=over 4	304	=over 4
111		305
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	306	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		307
114	This will initialise the default event loop if it hasn't been initialised	308	This will initialise the default event loop if it hasn't been initialised
115	yet and return it. If the default loop could not be initialised, returns	309	yet and return it. If the default loop could not be initialised, returns
116	false. If it already was initialised it simply returns it (and ignores the	310	false. If it already was initialised it simply returns it (and ignores the
117	flags).	311	flags. If that is troubling you, check C<ev_backend ()> afterwards).
118		312
119	If you don't know what event loop to use, use the one returned from this	313	If you don't know what event loop to use, use the one returned from this
120	function.	314	function.
121		315
		316	Note that this function is I<not> thread-safe, so if you want to use it
		317	from multiple threads, you have to lock (note also that this is unlikely,
		318	as loops cannot be shared easily between threads anyway).
		319
		320	The default loop is the only loop that can handle C<ev_signal> and
		321	C<ev_child> watchers, and to do this, it always registers a handler
		322	for C<SIGCHLD>. If this is a problem for your application you can either
		323	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		324	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		325	C<ev_default_init>.
		326
122	The flags argument can be used to specify special behaviour or specific	327	The flags argument can be used to specify special behaviour or specific
123	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	328	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
124		329
125	It supports the following flags:	330	The following flags are supported:
126		331
127	=over 4	332	=over 4
128		333
129	=item C<EVFLAG_AUTO>	334	=item C<EVFLAG_AUTO>
130		335
131	The default flags value. Use this if you have no clue (it's the right	336	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	337	thing, believe me).
133		338
134	=item C<EVFLAG_NOENV>	339	=item C<EVFLAG_NOENV>
135		340
136	If this flag bit is ored into the flag value (or the program runs setuid	341	If this flag bit is or'ed into the flag value (or the program runs setuid
137	or setgid) then libev will I<not> look at the environment variable	342	or setgid) then libev will I<not> look at the environment variable
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	343	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	344	override the flags completely if it is found in the environment. This is
140	useful to try out specific backends to test their performance, or to work	345	useful to try out specific backends to test their performance, or to work
141	around bugs.	346	around bugs.
142		347
		348	=item C<EVFLAG_FORKCHECK>
		349
		350	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
		351	a fork, you can also make libev check for a fork in each iteration by
		352	enabling this flag.
		353
		354	This works by calling C<getpid ()> on every iteration of the loop,
		355	and thus this might slow down your event loop if you do a lot of loop
		356	iterations and little real work, but is usually not noticeable (on my
		357	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		358	without a system call and thus I<very> fast, but my GNU/Linux system also has
		359	C<pthread_atfork> which is even faster).
		360
		361	The big advantage of this flag is that you can forget about fork (and
		362	forget about forgetting to tell libev about forking) when you use this
		363	flag.
		364
		365	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		366	environment variable.
		367
		368	=item C<EVFLAG_NOINOTIFY>
		369
		370	When this flag is specified, then libev will not attempt to use the
		371	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		372	testing, this flag can be useful to conserve inotify file descriptors, as
		373	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		374
		375	=item C<EVFLAG_NOSIGNALFD>
		376
		377	When this flag is specified, then libev will not attempt to use the
		378	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This is
		379	probably only useful to work around any bugs in libev. Consequently, this
		380	flag might go away once the signalfd functionality is considered stable,
		381	so it's useful mostly in environment variables and not in program code.
		382
143	=item C<EVMETHOD_SELECT> (portable select backend)	383	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		384
		385	This is your standard select(2) backend. Not I<completely> standard, as
		386	libev tries to roll its own fd_set with no limits on the number of fds,
		387	but if that fails, expect a fairly low limit on the number of fds when
		388	using this backend. It doesn't scale too well (O(highest_fd)), but its
		389	usually the fastest backend for a low number of (low-numbered :) fds.
		390
		391	To get good performance out of this backend you need a high amount of
		392	parallelism (most of the file descriptors should be busy). If you are
		393	writing a server, you should C<accept ()> in a loop to accept as many
		394	connections as possible during one iteration. You might also want to have
		395	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		396	readiness notifications you get per iteration.
		397
		398	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		399	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		400	C<exceptfds> set on that platform).
		401
145	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	402	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		403
147	=item C<EVMETHOD_EPOLL> (linux only)	404	And this is your standard poll(2) backend. It's more complicated
		405	than select, but handles sparse fds better and has no artificial
		406	limit on the number of fds you can use (except it will slow down
		407	considerably with a lot of inactive fds). It scales similarly to select,
		408	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		409	performance tips.
148		410
149	=item C<EVMETHOD_KQUEUE> (some bsds only)	411	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		412	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
150		413
151	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	414	=item C<EVBACKEND_EPOLL> (value 4, Linux)
152		415
153	=item C<EVMETHOD_PORT> (solaris 10 only)	416	For few fds, this backend is a bit little slower than poll and select,
		417	but it scales phenomenally better. While poll and select usually scale
		418	like O(total_fds) where n is the total number of fds (or the highest fd),
		419	epoll scales either O(1) or O(active_fds).
154		420
155	If one or more of these are ored into the flags value, then only these	421	The epoll mechanism deserves honorable mention as the most misdesigned
156	backends will be tried (in the reverse order as given here). If one are	422	of the more advanced event mechanisms: mere annoyances include silently
157	specified, any backend will do.	423	dropping file descriptors, requiring a system call per change per file
		424	descriptor (and unnecessary guessing of parameters), problems with dup and
		425	so on. The biggest issue is fork races, however - if a program forks then
		426	I<both> parent and child process have to recreate the epoll set, which can
		427	take considerable time (one syscall per file descriptor) and is of course
		428	hard to detect.
		429
		430	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		431	of course I<doesn't>, and epoll just loves to report events for totally
		432	I<different> file descriptors (even already closed ones, so one cannot
		433	even remove them from the set) than registered in the set (especially
		434	on SMP systems). Libev tries to counter these spurious notifications by
		435	employing an additional generation counter and comparing that against the
		436	events to filter out spurious ones, recreating the set when required.
		437
		438	While stopping, setting and starting an I/O watcher in the same iteration
		439	will result in some caching, there is still a system call per such
		440	incident (because the same I<file descriptor> could point to a different
		441	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		442	file descriptors might not work very well if you register events for both
		443	file descriptors.
		444
		445	Best performance from this backend is achieved by not unregistering all
		446	watchers for a file descriptor until it has been closed, if possible,
		447	i.e. keep at least one watcher active per fd at all times. Stopping and
		448	starting a watcher (without re-setting it) also usually doesn't cause
		449	extra overhead. A fork can both result in spurious notifications as well
		450	as in libev having to destroy and recreate the epoll object, which can
		451	take considerable time and thus should be avoided.
		452
		453	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		454	faster than epoll for maybe up to a hundred file descriptors, depending on
		455	the usage. So sad.
		456
		457	While nominally embeddable in other event loops, this feature is broken in
		458	all kernel versions tested so far.
		459
		460	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		461	C<EVBACKEND_POLL>.
		462
		463	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		464
		465	Kqueue deserves special mention, as at the time of this writing, it
		466	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		467	with anything but sockets and pipes, except on Darwin, where of course
		468	it's completely useless). Unlike epoll, however, whose brokenness
		469	is by design, these kqueue bugs can (and eventually will) be fixed
		470	without API changes to existing programs. For this reason it's not being
		471	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		472	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		473	system like NetBSD.
		474
		475	You still can embed kqueue into a normal poll or select backend and use it
		476	only for sockets (after having made sure that sockets work with kqueue on
		477	the target platform). See C<ev_embed> watchers for more info.
		478
		479	It scales in the same way as the epoll backend, but the interface to the
		480	kernel is more efficient (which says nothing about its actual speed, of
		481	course). While stopping, setting and starting an I/O watcher does never
		482	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		483	two event changes per incident. Support for C<fork ()> is very bad (but
		484	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		485	cases
		486
		487	This backend usually performs well under most conditions.
		488
		489	While nominally embeddable in other event loops, this doesn't work
		490	everywhere, so you might need to test for this. And since it is broken
		491	almost everywhere, you should only use it when you have a lot of sockets
		492	(for which it usually works), by embedding it into another event loop
		493	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		494	also broken on OS X)) and, did I mention it, using it only for sockets.
		495
		496	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		497	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		498	C<NOTE_EOF>.
		499
		500	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		501
		502	This is not implemented yet (and might never be, unless you send me an
		503	implementation). According to reports, C</dev/poll> only supports sockets
		504	and is not embeddable, which would limit the usefulness of this backend
		505	immensely.
		506
		507	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		508
		509	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		510	it's really slow, but it still scales very well (O(active_fds)).
		511
		512	Please note that Solaris event ports can deliver a lot of spurious
		513	notifications, so you need to use non-blocking I/O or other means to avoid
		514	blocking when no data (or space) is available.
		515
		516	While this backend scales well, it requires one system call per active
		517	file descriptor per loop iteration. For small and medium numbers of file
		518	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		519	might perform better.
		520
		521	On the positive side, with the exception of the spurious readiness
		522	notifications, this backend actually performed fully to specification
		523	in all tests and is fully embeddable, which is a rare feat among the
		524	OS-specific backends (I vastly prefer correctness over speed hacks).
		525
		526	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		527	C<EVBACKEND_POLL>.
		528
		529	=item C<EVBACKEND_ALL>
		530
		531	Try all backends (even potentially broken ones that wouldn't be tried
		532	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		533	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		534
		535	It is definitely not recommended to use this flag.
158		536
159	=back	537	=back
		538
		539	If one or more of the backend flags are or'ed into the flags value,
		540	then only these backends will be tried (in the reverse order as listed
		541	here). If none are specified, all backends in C<ev_recommended_backends
		542	()> will be tried.
		543
		544	Example: This is the most typical usage.
		545
		546	if (!ev_default_loop (0))
		547	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		548
		549	Example: Restrict libev to the select and poll backends, and do not allow
		550	environment settings to be taken into account:
		551
		552	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		553
		554	Example: Use whatever libev has to offer, but make sure that kqueue is
		555	used if available (warning, breaks stuff, best use only with your own
		556	private event loop and only if you know the OS supports your types of
		557	fds):
		558
		559	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
160		560
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	561	=item struct ev_loop *ev_loop_new (unsigned int flags)
162		562
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	563	Similar to C<ev_default_loop>, but always creates a new event loop that is
164	always distinct from the default loop. Unlike the default loop, it cannot	564	always distinct from the default loop. Unlike the default loop, it cannot
165	handle signal and child watchers, and attempts to do so will be greeted by	565	handle signal and child watchers, and attempts to do so will be greeted by
166	undefined behaviour (or a failed assertion if assertions are enabled).	566	undefined behaviour (or a failed assertion if assertions are enabled).
167		567
		568	Note that this function I<is> thread-safe, and the recommended way to use
		569	libev with threads is indeed to create one loop per thread, and using the
		570	default loop in the "main" or "initial" thread.
		571
		572	Example: Try to create a event loop that uses epoll and nothing else.
		573
		574	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		575	if (!epoller)
		576	fatal ("no epoll found here, maybe it hides under your chair");
		577
168	=item ev_default_destroy ()	578	=item ev_default_destroy ()
169		579
170	Destroys the default loop again (frees all memory and kernel state	580	Destroys the default loop again (frees all memory and kernel state
171	etc.). This stops all registered event watchers (by not touching them in	581	etc.). None of the active event watchers will be stopped in the normal
172	any way whatsoever, although you cannot rely on this :).	582	sense, so e.g. C<ev_is_active> might still return true. It is your
		583	responsibility to either stop all watchers cleanly yourself I<before>
		584	calling this function, or cope with the fact afterwards (which is usually
		585	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		586	for example).
		587
		588	Note that certain global state, such as signal state (and installed signal
		589	handlers), will not be freed by this function, and related watchers (such
		590	as signal and child watchers) would need to be stopped manually.
		591
		592	In general it is not advisable to call this function except in the
		593	rare occasion where you really need to free e.g. the signal handling
		594	pipe fds. If you need dynamically allocated loops it is better to use
		595	C<ev_loop_new> and C<ev_loop_destroy>).
173		596
174	=item ev_loop_destroy (loop)	597	=item ev_loop_destroy (loop)
175		598
176	Like C<ev_default_destroy>, but destroys an event loop created by an	599	Like C<ev_default_destroy>, but destroys an event loop created by an
177	earlier call to C<ev_loop_new>.	600	earlier call to C<ev_loop_new>.
178		601
179	=item ev_default_fork ()	602	=item ev_default_fork ()
180		603
		604	This function sets a flag that causes subsequent C<ev_loop> iterations
181	This function reinitialises the kernel state for backends that have	605	to reinitialise the kernel state for backends that have one. Despite the
182	one. Despite the name, you can call it anytime, but it makes most sense	606	name, you can call it anytime, but it makes most sense after forking, in
183	after forking, in either the parent or child process (or both, but that	607	the child process (or both child and parent, but that again makes little
184	again makes little sense).	608	sense). You I<must> call it in the child before using any of the libev
		609	functions, and it will only take effect at the next C<ev_loop> iteration.
185		610
186	You I<must> call this function after forking if and only if you want to	611	On the other hand, you only need to call this function in the child
187	use the event library in both processes. If you just fork+exec, you don't	612	process if and only if you want to use the event library in the child. If
188	have to call it.	613	you just fork+exec, you don't have to call it at all.
189		614
190	The function itself is quite fast and it's usually not a problem to call	615	The function itself is quite fast and it's usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in	616	it just in case after a fork. To make this easy, the function will fit in
192	quite nicely into a call to C<pthread_atfork>:	617	quite nicely into a call to C<pthread_atfork>:
193		618
…		…
195		620
196	=item ev_loop_fork (loop)	621	=item ev_loop_fork (loop)
197		622
198	Like C<ev_default_fork>, but acts on an event loop created by	623	Like C<ev_default_fork>, but acts on an event loop created by
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	624	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
200	after fork, and how you do this is entirely your own problem.	625	after fork that you want to re-use in the child, and how you do this is
		626	entirely your own problem.
201		627
		628	=item int ev_is_default_loop (loop)
		629
		630	Returns true when the given loop is, in fact, the default loop, and false
		631	otherwise.
		632
		633	=item unsigned int ev_loop_count (loop)
		634
		635	Returns the count of loop iterations for the loop, which is identical to
		636	the number of times libev did poll for new events. It starts at C<0> and
		637	happily wraps around with enough iterations.
		638
		639	This value can sometimes be useful as a generation counter of sorts (it
		640	"ticks" the number of loop iterations), as it roughly corresponds with
		641	C<ev_prepare> and C<ev_check> calls.
		642
		643	=item unsigned int ev_loop_depth (loop)
		644
		645	Returns the number of times C<ev_loop> was entered minus the number of
		646	times C<ev_loop> was exited, in other words, the recursion depth.
		647
		648	Outside C<ev_loop>, this number is zero. In a callback, this number is
		649	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		650	in which case it is higher.
		651
		652	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		653	etc.), doesn't count as exit.
		654
202	=item unsigned int ev_method (loop)	655	=item unsigned int ev_backend (loop)
203		656
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	657	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	658	use.
206		659
207	=item ev_tstamp ev_now (loop)	660	=item ev_tstamp ev_now (loop)
208		661
209	Returns the current "event loop time", which is the time the event loop	662	Returns the current "event loop time", which is the time the event loop
210	got events and started processing them. This timestamp does not change	663	received events and started processing them. This timestamp does not
211	as long as callbacks are being processed, and this is also the base time	664	change as long as callbacks are being processed, and this is also the base
212	used for relative timers. You can treat it as the timestamp of the event	665	time used for relative timers. You can treat it as the timestamp of the
213	occuring (or more correctly, the mainloop finding out about it).	666	event occurring (or more correctly, libev finding out about it).
		667
		668	=item ev_now_update (loop)
		669
		670	Establishes the current time by querying the kernel, updating the time
		671	returned by C<ev_now ()> in the progress. This is a costly operation and
		672	is usually done automatically within C<ev_loop ()>.
		673
		674	This function is rarely useful, but when some event callback runs for a
		675	very long time without entering the event loop, updating libev's idea of
		676	the current time is a good idea.
		677
		678	See also L<The special problem of time updates> in the C<ev_timer> section.
		679
		680	=item ev_suspend (loop)
		681
		682	=item ev_resume (loop)
		683
		684	These two functions suspend and resume a loop, for use when the loop is
		685	not used for a while and timeouts should not be processed.
		686
		687	A typical use case would be an interactive program such as a game: When
		688	the user presses C<^Z> to suspend the game and resumes it an hour later it
		689	would be best to handle timeouts as if no time had actually passed while
		690	the program was suspended. This can be achieved by calling C<ev_suspend>
		691	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		692	C<ev_resume> directly afterwards to resume timer processing.
		693
		694	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		695	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		696	will be rescheduled (that is, they will lose any events that would have
		697	occured while suspended).
		698
		699	After calling C<ev_suspend> you B<must not> call I<any> function on the
		700	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		701	without a previous call to C<ev_suspend>.
		702
		703	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		704	event loop time (see C<ev_now_update>).
214		705
215	=item ev_loop (loop, int flags)	706	=item ev_loop (loop, int flags)
216		707
217	Finally, this is it, the event handler. This function usually is called	708	Finally, this is it, the event handler. This function usually is called
218	after you initialised all your watchers and you want to start handling	709	after you initialised all your watchers and you want to start handling
219	events.	710	events.
220		711
221	If the flags argument is specified as 0, it will not return until either	712	If the flags argument is specified as C<0>, it will not return until
222	no event watchers are active anymore or C<ev_unloop> was called.	713	either no event watchers are active anymore or C<ev_unloop> was called.
		714
		715	Please note that an explicit C<ev_unloop> is usually better than
		716	relying on all watchers to be stopped when deciding when a program has
		717	finished (especially in interactive programs), but having a program
		718	that automatically loops as long as it has to and no longer by virtue
		719	of relying on its watchers stopping correctly, that is truly a thing of
		720	beauty.
223		721
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	722	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
225	those events and any outstanding ones, but will not block your process in	723	those events and any already outstanding ones, but will not block your
226	case there are no events and will return after one iteration of the loop.	724	process in case there are no events and will return after one iteration of
		725	the loop.
227		726
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	727	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
229	neccessary) and will handle those and any outstanding ones. It will block	728	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives, and will return after	729	will block your process until at least one new event arrives (which could
		730	be an event internal to libev itself, so there is no guarantee that a
		731	user-registered callback will be called), and will return after one
231	one iteration of the loop.	732	iteration of the loop.
232		733
233	This flags value could be used to implement alternative looping	734	This is useful if you are waiting for some external event in conjunction
234	constructs, but the C<prepare> and C<check> watchers provide a better and	735	with something not expressible using other libev watchers (i.e. "roll your
235	more generic mechanism.	736	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		737	usually a better approach for this kind of thing.
		738
		739	Here are the gory details of what C<ev_loop> does:
		740
		741	- Before the first iteration, call any pending watchers.
		742	* If EVFLAG_FORKCHECK was used, check for a fork.
		743	- If a fork was detected (by any means), queue and call all fork watchers.
		744	- Queue and call all prepare watchers.
		745	- If we have been forked, detach and recreate the kernel state
		746	as to not disturb the other process.
		747	- Update the kernel state with all outstanding changes.
		748	- Update the "event loop time" (ev_now ()).
		749	- Calculate for how long to sleep or block, if at all
		750	(active idle watchers, EVLOOP_NONBLOCK or not having
		751	any active watchers at all will result in not sleeping).
		752	- Sleep if the I/O and timer collect interval say so.
		753	- Block the process, waiting for any events.
		754	- Queue all outstanding I/O (fd) events.
		755	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		756	- Queue all expired timers.
		757	- Queue all expired periodics.
		758	- Unless any events are pending now, queue all idle watchers.
		759	- Queue all check watchers.
		760	- Call all queued watchers in reverse order (i.e. check watchers first).
		761	Signals and child watchers are implemented as I/O watchers, and will
		762	be handled here by queueing them when their watcher gets executed.
		763	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		764	were used, or there are no active watchers, return, otherwise
		765	continue with step *.
		766
		767	Example: Queue some jobs and then loop until no events are outstanding
		768	anymore.
		769
		770	... queue jobs here, make sure they register event watchers as long
		771	... as they still have work to do (even an idle watcher will do..)
		772	ev_loop (my_loop, 0);
		773	... jobs done or somebody called unloop. yeah!
236		774
237	=item ev_unloop (loop, how)	775	=item ev_unloop (loop, how)
238		776
239	Can be used to make a call to C<ev_loop> return early (but only after it	777	Can be used to make a call to C<ev_loop> return early (but only after it
240	has processed all outstanding events). The C<how> argument must be either	778	has processed all outstanding events). The C<how> argument must be either
241	C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or	779	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
242	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	780	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
		781
		782	This "unloop state" will be cleared when entering C<ev_loop> again.
		783
		784	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
243		785
244	=item ev_ref (loop)	786	=item ev_ref (loop)
245		787
246	=item ev_unref (loop)	788	=item ev_unref (loop)
247		789
248	Ref/unref can be used to add or remove a reference count on the event	790	Ref/unref can be used to add or remove a reference count on the event
249	loop: Every watcher keeps one reference, and as long as the reference	791	loop: Every watcher keeps one reference, and as long as the reference
250	count is nonzero, C<ev_loop> will not return on its own. If you have	792	count is nonzero, C<ev_loop> will not return on its own.
		793
251	a watcher you never unregister that should not keep C<ev_loop> from	794	If you have a watcher you never unregister that should not keep C<ev_loop>
252	returning, ev_unref() after starting, and ev_ref() before stopping it. For	795	from returning, call ev_unref() after starting, and ev_ref() before
		796	stopping it.
		797
253	example, libev itself uses this for its internal signal pipe: It is not	798	As an example, libev itself uses this for its internal signal pipe: It
254	visible to the libev user and should not keep C<ev_loop> from exiting if	799	is not visible to the libev user and should not keep C<ev_loop> from
255	no event watchers registered by it are active. It is also an excellent	800	exiting if no event watchers registered by it are active. It is also an
256	way to do this for generic recurring timers or from within third-party	801	excellent way to do this for generic recurring timers or from within
257	libraries. Just remember to I<unref after start> and I<ref before stop>.	802	third-party libraries. Just remember to I<unref after start> and I<ref
		803	before stop> (but only if the watcher wasn't active before, or was active
		804	before, respectively. Note also that libev might stop watchers itself
		805	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		806	in the callback).
		807
		808	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		809	running when nothing else is active.
		810
		811	ev_signal exitsig;
		812	ev_signal_init (&exitsig, sig_cb, SIGINT);
		813	ev_signal_start (loop, &exitsig);
		814	evf_unref (loop);
		815
		816	Example: For some weird reason, unregister the above signal handler again.
		817
		818	ev_ref (loop);
		819	ev_signal_stop (loop, &exitsig);
		820
		821	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		822
		823	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		824
		825	These advanced functions influence the time that libev will spend waiting
		826	for events. Both time intervals are by default C<0>, meaning that libev
		827	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		828	latency.
		829
		830	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		831	allows libev to delay invocation of I/O and timer/periodic callbacks
		832	to increase efficiency of loop iterations (or to increase power-saving
		833	opportunities).
		834
		835	The idea is that sometimes your program runs just fast enough to handle
		836	one (or very few) event(s) per loop iteration. While this makes the
		837	program responsive, it also wastes a lot of CPU time to poll for new
		838	events, especially with backends like C<select ()> which have a high
		839	overhead for the actual polling but can deliver many events at once.
		840
		841	By setting a higher I<io collect interval> you allow libev to spend more
		842	time collecting I/O events, so you can handle more events per iteration,
		843	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		844	C<ev_timer>) will be not affected. Setting this to a non-null value will
		845	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		846	sleep time ensures that libev will not poll for I/O events more often then
		847	once per this interval, on average.
		848
		849	Likewise, by setting a higher I<timeout collect interval> you allow libev
		850	to spend more time collecting timeouts, at the expense of increased
		851	latency/jitter/inexactness (the watcher callback will be called
		852	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		853	value will not introduce any overhead in libev.
		854
		855	Many (busy) programs can usually benefit by setting the I/O collect
		856	interval to a value near C<0.1> or so, which is often enough for
		857	interactive servers (of course not for games), likewise for timeouts. It
		858	usually doesn't make much sense to set it to a lower value than C<0.01>,
		859	as this approaches the timing granularity of most systems. Note that if
		860	you do transactions with the outside world and you can't increase the
		861	parallelity, then this setting will limit your transaction rate (if you
		862	need to poll once per transaction and the I/O collect interval is 0.01,
		863	then you can't do more than 100 transations per second).
		864
		865	Setting the I<timeout collect interval> can improve the opportunity for
		866	saving power, as the program will "bundle" timer callback invocations that
		867	are "near" in time together, by delaying some, thus reducing the number of
		868	times the process sleeps and wakes up again. Another useful technique to
		869	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		870	they fire on, say, one-second boundaries only.
		871
		872	Example: we only need 0.1s timeout granularity, and we wish not to poll
		873	more often than 100 times per second:
		874
		875	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		876	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		877
		878	=item ev_invoke_pending (loop)
		879
		880	This call will simply invoke all pending watchers while resetting their
		881	pending state. Normally, C<ev_loop> does this automatically when required,
		882	but when overriding the invoke callback this call comes handy.
		883
		884	=item int ev_pending_count (loop)
		885
		886	Returns the number of pending watchers - zero indicates that no watchers
		887	are pending.
		888
		889	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		890
		891	This overrides the invoke pending functionality of the loop: Instead of
		892	invoking all pending watchers when there are any, C<ev_loop> will call
		893	this callback instead. This is useful, for example, when you want to
		894	invoke the actual watchers inside another context (another thread etc.).
		895
		896	If you want to reset the callback, use C<ev_invoke_pending> as new
		897	callback.
		898
		899	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		900
		901	Sometimes you want to share the same loop between multiple threads. This
		902	can be done relatively simply by putting mutex_lock/unlock calls around
		903	each call to a libev function.
		904
		905	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		906	wait for it to return. One way around this is to wake up the loop via
		907	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		908	and I<acquire> callbacks on the loop.
		909
		910	When set, then C<release> will be called just before the thread is
		911	suspended waiting for new events, and C<acquire> is called just
		912	afterwards.
		913
		914	Ideally, C<release> will just call your mutex_unlock function, and
		915	C<acquire> will just call the mutex_lock function again.
		916
		917	While event loop modifications are allowed between invocations of
		918	C<release> and C<acquire> (that's their only purpose after all), no
		919	modifications done will affect the event loop, i.e. adding watchers will
		920	have no effect on the set of file descriptors being watched, or the time
		921	waited. USe an C<ev_async> watcher to wake up C<ev_loop> when you want it
		922	to take note of any changes you made.
		923
		924	In theory, threads executing C<ev_loop> will be async-cancel safe between
		925	invocations of C<release> and C<acquire>.
		926
		927	See also the locking example in the C<THREADS> section later in this
		928	document.
		929
		930	=item ev_set_userdata (loop, void *data)
		931
		932	=item ev_userdata (loop)
		933
		934	Set and retrieve a single C<void *> associated with a loop. When
		935	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		936	C<0.>
		937
		938	These two functions can be used to associate arbitrary data with a loop,
		939	and are intended solely for the C<invoke_pending_cb>, C<release> and
		940	C<acquire> callbacks described above, but of course can be (ab-)used for
		941	any other purpose as well.
		942
		943	=item ev_loop_verify (loop)
		944
		945	This function only does something when C<EV_VERIFY> support has been
		946	compiled in, which is the default for non-minimal builds. It tries to go
		947	through all internal structures and checks them for validity. If anything
		948	is found to be inconsistent, it will print an error message to standard
		949	error and call C<abort ()>.
		950
		951	This can be used to catch bugs inside libev itself: under normal
		952	circumstances, this function will never abort as of course libev keeps its
		953	data structures consistent.
258		954
259	=back	955	=back
260		956
		957
261	=head1 ANATOMY OF A WATCHER	958	=head1 ANATOMY OF A WATCHER
		959
		960	In the following description, uppercase C<TYPE> in names stands for the
		961	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		962	watchers and C<ev_io_start> for I/O watchers.
262		963
263	A watcher is a structure that you create and register to record your	964	A watcher is a structure that you create and register to record your
264	interest in some event. For instance, if you want to wait for STDIN to	965	interest in some event. For instance, if you want to wait for STDIN to
265	become readable, you would create an C<ev_io> watcher for that:	966	become readable, you would create an C<ev_io> watcher for that:
266		967
267	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	968	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
268	{	969	{
269	ev_io_stop (w);	970	ev_io_stop (w);
270	ev_unloop (loop, EVUNLOOP_ALL);	971	ev_unloop (loop, EVUNLOOP_ALL);
271	}	972	}
272		973
273	struct ev_loop *loop = ev_default_loop (0);	974	struct ev_loop *loop = ev_default_loop (0);
		975
274	struct ev_io stdin_watcher;	976	ev_io stdin_watcher;
		977
275	ev_init (&stdin_watcher, my_cb);	978	ev_init (&stdin_watcher, my_cb);
276	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	979	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
277	ev_io_start (loop, &stdin_watcher);	980	ev_io_start (loop, &stdin_watcher);
		981
278	ev_loop (loop, 0);	982	ev_loop (loop, 0);
279		983
280	As you can see, you are responsible for allocating the memory for your	984	As you can see, you are responsible for allocating the memory for your
281	watcher structures (and it is usually a bad idea to do this on the stack,	985	watcher structures (and it is I<usually> a bad idea to do this on the
282	although this can sometimes be quite valid).	986	stack).
		987
		988	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		989	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
283		990
284	Each watcher structure must be initialised by a call to C<ev_init	991	Each watcher structure must be initialised by a call to C<ev_init
285	(watcher *, callback)>, which expects a callback to be provided. This	992	(watcher *, callback)>, which expects a callback to be provided. This
286	callback gets invoked each time the event occurs (or, in the case of io	993	callback gets invoked each time the event occurs (or, in the case of I/O
287	watchers, each time the event loop detects that the file descriptor given	994	watchers, each time the event loop detects that the file descriptor given
288	is readable and/or writable).	995	is readable and/or writable).
289		996
290	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	997	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
291	with arguments specific to this watcher type. There is also a macro	998	macro to configure it, with arguments specific to the watcher type. There
292	to combine initialisation and setting in one call: C<< ev_<type>_init	999	is also a macro to combine initialisation and setting in one call: C<<
293	(watcher *, callback, ...) >>.	1000	ev_TYPE_init (watcher *, callback, ...) >>.
294		1001
295	To make the watcher actually watch out for events, you have to start it	1002	To make the watcher actually watch out for events, you have to start it
296	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1003	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
297	*) >>), and you can stop watching for events at any time by calling the	1004	*) >>), and you can stop watching for events at any time by calling the
298	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1005	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
299		1006
300	As long as your watcher is active (has been started but not stopped) you	1007	As long as your watcher is active (has been started but not stopped) you
301	must not touch the values stored in it. Most specifically you must never	1008	must not touch the values stored in it. Most specifically you must never
302	reinitialise it or call its set method.	1009	reinitialise it or call its C<ev_TYPE_set> macro.
303
304	You can check whether an event is active by calling the C<ev_is_active
305	(watcher *)> macro. To see whether an event is outstanding (but the
306	callback for it has not been called yet) you can use the C<ev_is_pending
307	(watcher *)> macro.
308		1010
309	Each and every callback receives the event loop pointer as first, the	1011	Each and every callback receives the event loop pointer as first, the
310	registered watcher structure as second, and a bitset of received events as	1012	registered watcher structure as second, and a bitset of received events as
311	third argument.	1013	third argument.
312		1014
…		…
336	The signal specified in the C<ev_signal> watcher has been received by a thread.	1038	The signal specified in the C<ev_signal> watcher has been received by a thread.
337		1039
338	=item C<EV_CHILD>	1040	=item C<EV_CHILD>
339		1041
340	The pid specified in the C<ev_child> watcher has received a status change.	1042	The pid specified in the C<ev_child> watcher has received a status change.
		1043
		1044	=item C<EV_STAT>
		1045
		1046	The path specified in the C<ev_stat> watcher changed its attributes somehow.
341		1047
342	=item C<EV_IDLE>	1048	=item C<EV_IDLE>
343		1049
344	The C<ev_idle> watcher has determined that you have nothing better to do.	1050	The C<ev_idle> watcher has determined that you have nothing better to do.
345		1051
…		…
353	received events. Callbacks of both watcher types can start and stop as	1059	received events. Callbacks of both watcher types can start and stop as
354	many watchers as they want, and all of them will be taken into account	1060	many watchers as they want, and all of them will be taken into account
355	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1061	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
356	C<ev_loop> from blocking).	1062	C<ev_loop> from blocking).
357		1063
		1064	=item C<EV_EMBED>
		1065
		1066	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1067
		1068	=item C<EV_FORK>
		1069
		1070	The event loop has been resumed in the child process after fork (see
		1071	C<ev_fork>).
		1072
		1073	=item C<EV_ASYNC>
		1074
		1075	The given async watcher has been asynchronously notified (see C<ev_async>).
		1076
		1077	=item C<EV_CUSTOM>
		1078
		1079	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1080	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1081
358	=item C<EV_ERROR>	1082	=item C<EV_ERROR>
359		1083
360	An unspecified error has occured, the watcher has been stopped. This might	1084	An unspecified error has occurred, the watcher has been stopped. This might
361	happen because the watcher could not be properly started because libev	1085	happen because the watcher could not be properly started because libev
362	ran out of memory, a file descriptor was found to be closed or any other	1086	ran out of memory, a file descriptor was found to be closed or any other
		1087	problem. Libev considers these application bugs.
		1088
363	problem. You best act on it by reporting the problem and somehow coping	1089	You best act on it by reporting the problem and somehow coping with the
364	with the watcher being stopped.	1090	watcher being stopped. Note that well-written programs should not receive
		1091	an error ever, so when your watcher receives it, this usually indicates a
		1092	bug in your program.
365		1093
366	Libev will usually signal a few "dummy" events together with an error,	1094	Libev will usually signal a few "dummy" events together with an error, for
367	for example it might indicate that a fd is readable or writable, and if	1095	example it might indicate that a fd is readable or writable, and if your
368	your callbacks is well-written it can just attempt the operation and cope	1096	callbacks is well-written it can just attempt the operation and cope with
369	with the error from read() or write(). This will not work in multithreaded	1097	the error from read() or write(). This will not work in multi-threaded
370	programs, though, so beware.	1098	programs, though, as the fd could already be closed and reused for another
		1099	thing, so beware.
371		1100
372	=back	1101	=back
373		1102
		1103	=head2 GENERIC WATCHER FUNCTIONS
		1104
		1105	=over 4
		1106
		1107	=item C<ev_init> (ev_TYPE *watcher, callback)
		1108
		1109	This macro initialises the generic portion of a watcher. The contents
		1110	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1111	the generic parts of the watcher are initialised, you I<need> to call
		1112	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1113	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1114	which rolls both calls into one.
		1115
		1116	You can reinitialise a watcher at any time as long as it has been stopped
		1117	(or never started) and there are no pending events outstanding.
		1118
		1119	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1120	int revents)>.
		1121
		1122	Example: Initialise an C<ev_io> watcher in two steps.
		1123
		1124	ev_io w;
		1125	ev_init (&w, my_cb);
		1126	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1127
		1128	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		1129
		1130	This macro initialises the type-specific parts of a watcher. You need to
		1131	call C<ev_init> at least once before you call this macro, but you can
		1132	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1133	macro on a watcher that is active (it can be pending, however, which is a
		1134	difference to the C<ev_init> macro).
		1135
		1136	Although some watcher types do not have type-specific arguments
		1137	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1138
		1139	See C<ev_init>, above, for an example.
		1140
		1141	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1142
		1143	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1144	calls into a single call. This is the most convenient method to initialise
		1145	a watcher. The same limitations apply, of course.
		1146
		1147	Example: Initialise and set an C<ev_io> watcher in one step.
		1148
		1149	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1150
		1151	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		1152
		1153	Starts (activates) the given watcher. Only active watchers will receive
		1154	events. If the watcher is already active nothing will happen.
		1155
		1156	Example: Start the C<ev_io> watcher that is being abused as example in this
		1157	whole section.
		1158
		1159	ev_io_start (EV_DEFAULT_UC, &w);
		1160
		1161	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		1162
		1163	Stops the given watcher if active, and clears the pending status (whether
		1164	the watcher was active or not).
		1165
		1166	It is possible that stopped watchers are pending - for example,
		1167	non-repeating timers are being stopped when they become pending - but
		1168	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1169	pending. If you want to free or reuse the memory used by the watcher it is
		1170	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1171
		1172	=item bool ev_is_active (ev_TYPE *watcher)
		1173
		1174	Returns a true value iff the watcher is active (i.e. it has been started
		1175	and not yet been stopped). As long as a watcher is active you must not modify
		1176	it.
		1177
		1178	=item bool ev_is_pending (ev_TYPE *watcher)
		1179
		1180	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1181	events but its callback has not yet been invoked). As long as a watcher
		1182	is pending (but not active) you must not call an init function on it (but
		1183	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1184	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1185	it).
		1186
		1187	=item callback ev_cb (ev_TYPE *watcher)
		1188
		1189	Returns the callback currently set on the watcher.
		1190
		1191	=item ev_cb_set (ev_TYPE *watcher, callback)
		1192
		1193	Change the callback. You can change the callback at virtually any time
		1194	(modulo threads).
		1195
		1196	=item ev_set_priority (ev_TYPE *watcher, priority)
		1197
		1198	=item int ev_priority (ev_TYPE *watcher)
		1199
		1200	Set and query the priority of the watcher. The priority is a small
		1201	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1202	(default: C<-2>). Pending watchers with higher priority will be invoked
		1203	before watchers with lower priority, but priority will not keep watchers
		1204	from being executed (except for C<ev_idle> watchers).
		1205
		1206	If you need to suppress invocation when higher priority events are pending
		1207	you need to look at C<ev_idle> watchers, which provide this functionality.
		1208
		1209	You I<must not> change the priority of a watcher as long as it is active or
		1210	pending.
		1211
		1212	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1213	fine, as long as you do not mind that the priority value you query might
		1214	or might not have been clamped to the valid range.
		1215
		1216	The default priority used by watchers when no priority has been set is
		1217	always C<0>, which is supposed to not be too high and not be too low :).
		1218
		1219	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1220	priorities.
		1221
		1222	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1223
		1224	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1225	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1226	can deal with that fact, as both are simply passed through to the
		1227	callback.
		1228
		1229	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1230
		1231	If the watcher is pending, this function clears its pending status and
		1232	returns its C<revents> bitset (as if its callback was invoked). If the
		1233	watcher isn't pending it does nothing and returns C<0>.
		1234
		1235	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1236	callback to be invoked, which can be accomplished with this function.
		1237
		1238	=back
		1239
		1240
374	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1241	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375		1242
376	Each watcher has, by default, a member C<void *data> that you can change	1243	Each watcher has, by default, a member C<void *data> that you can change
377	and read at any time, libev will completely ignore it. This can be used	1244	and read at any time: libev will completely ignore it. This can be used
378	to associate arbitrary data with your watcher. If you need more data and	1245	to associate arbitrary data with your watcher. If you need more data and
379	don't want to allocate memory and store a pointer to it in that data	1246	don't want to allocate memory and store a pointer to it in that data
380	member, you can also "subclass" the watcher type and provide your own	1247	member, you can also "subclass" the watcher type and provide your own
381	data:	1248	data:
382		1249
383	struct my_io	1250	struct my_io
384	{	1251	{
385	struct ev_io io;	1252	ev_io io;
386	int otherfd;	1253	int otherfd;
387	void *somedata;	1254	void *somedata;
388	struct whatever *mostinteresting;	1255	struct whatever *mostinteresting;
389	}	1256	};
		1257
		1258	...
		1259	struct my_io w;
		1260	ev_io_init (&w.io, my_cb, fd, EV_READ);
390		1261
391	And since your callback will be called with a pointer to the watcher, you	1262	And since your callback will be called with a pointer to the watcher, you
392	can cast it back to your own type:	1263	can cast it back to your own type:
393		1264
394	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1265	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
395	{	1266	{
396	struct my_io *w = (struct my_io *)w_;	1267	struct my_io *w = (struct my_io *)w_;
397	...	1268	...
398	}	1269	}
399		1270
400	More interesting and less C-conformant ways of catsing your callback type	1271	More interesting and less C-conformant ways of casting your callback type
401	have been omitted....	1272	instead have been omitted.
		1273
		1274	Another common scenario is to use some data structure with multiple
		1275	embedded watchers:
		1276
		1277	struct my_biggy
		1278	{
		1279	int some_data;
		1280	ev_timer t1;
		1281	ev_timer t2;
		1282	}
		1283
		1284	In this case getting the pointer to C<my_biggy> is a bit more
		1285	complicated: Either you store the address of your C<my_biggy> struct
		1286	in the C<data> member of the watcher (for woozies), or you need to use
		1287	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1288	programmers):
		1289
		1290	#include <stddef.h>
		1291
		1292	static void
		1293	t1_cb (EV_P_ ev_timer *w, int revents)
		1294	{
		1295	struct my_biggy big = (struct my_biggy *)
		1296	(((char *)w) - offsetof (struct my_biggy, t1));
		1297	}
		1298
		1299	static void
		1300	t2_cb (EV_P_ ev_timer *w, int revents)
		1301	{
		1302	struct my_biggy big = (struct my_biggy *)
		1303	(((char *)w) - offsetof (struct my_biggy, t2));
		1304	}
		1305
		1306	=head2 WATCHER PRIORITY MODELS
		1307
		1308	Many event loops support I<watcher priorities>, which are usually small
		1309	integers that influence the ordering of event callback invocation
		1310	between watchers in some way, all else being equal.
		1311
		1312	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1313	description for the more technical details such as the actual priority
		1314	range.
		1315
		1316	There are two common ways how these these priorities are being interpreted
		1317	by event loops:
		1318
		1319	In the more common lock-out model, higher priorities "lock out" invocation
		1320	of lower priority watchers, which means as long as higher priority
		1321	watchers receive events, lower priority watchers are not being invoked.
		1322
		1323	The less common only-for-ordering model uses priorities solely to order
		1324	callback invocation within a single event loop iteration: Higher priority
		1325	watchers are invoked before lower priority ones, but they all get invoked
		1326	before polling for new events.
		1327
		1328	Libev uses the second (only-for-ordering) model for all its watchers
		1329	except for idle watchers (which use the lock-out model).
		1330
		1331	The rationale behind this is that implementing the lock-out model for
		1332	watchers is not well supported by most kernel interfaces, and most event
		1333	libraries will just poll for the same events again and again as long as
		1334	their callbacks have not been executed, which is very inefficient in the
		1335	common case of one high-priority watcher locking out a mass of lower
		1336	priority ones.
		1337
		1338	Static (ordering) priorities are most useful when you have two or more
		1339	watchers handling the same resource: a typical usage example is having an
		1340	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1341	timeouts. Under load, data might be received while the program handles
		1342	other jobs, but since timers normally get invoked first, the timeout
		1343	handler will be executed before checking for data. In that case, giving
		1344	the timer a lower priority than the I/O watcher ensures that I/O will be
		1345	handled first even under adverse conditions (which is usually, but not
		1346	always, what you want).
		1347
		1348	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1349	will only be executed when no same or higher priority watchers have
		1350	received events, they can be used to implement the "lock-out" model when
		1351	required.
		1352
		1353	For example, to emulate how many other event libraries handle priorities,
		1354	you can associate an C<ev_idle> watcher to each such watcher, and in
		1355	the normal watcher callback, you just start the idle watcher. The real
		1356	processing is done in the idle watcher callback. This causes libev to
		1357	continously poll and process kernel event data for the watcher, but when
		1358	the lock-out case is known to be rare (which in turn is rare :), this is
		1359	workable.
		1360
		1361	Usually, however, the lock-out model implemented that way will perform
		1362	miserably under the type of load it was designed to handle. In that case,
		1363	it might be preferable to stop the real watcher before starting the
		1364	idle watcher, so the kernel will not have to process the event in case
		1365	the actual processing will be delayed for considerable time.
		1366
		1367	Here is an example of an I/O watcher that should run at a strictly lower
		1368	priority than the default, and which should only process data when no
		1369	other events are pending:
		1370
		1371	ev_idle idle; // actual processing watcher
		1372	ev_io io; // actual event watcher
		1373
		1374	static void
		1375	io_cb (EV_P_ ev_io *w, int revents)
		1376	{
		1377	// stop the I/O watcher, we received the event, but
		1378	// are not yet ready to handle it.
		1379	ev_io_stop (EV_A_ w);
		1380
		1381	// start the idle watcher to ahndle the actual event.
		1382	// it will not be executed as long as other watchers
		1383	// with the default priority are receiving events.
		1384	ev_idle_start (EV_A_ &idle);
		1385	}
		1386
		1387	static void
		1388	idle_cb (EV_P_ ev_idle *w, int revents)
		1389	{
		1390	// actual processing
		1391	read (STDIN_FILENO, ...);
		1392
		1393	// have to start the I/O watcher again, as
		1394	// we have handled the event
		1395	ev_io_start (EV_P_ &io);
		1396	}
		1397
		1398	// initialisation
		1399	ev_idle_init (&idle, idle_cb);
		1400	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1401	ev_io_start (EV_DEFAULT_ &io);
		1402
		1403	In the "real" world, it might also be beneficial to start a timer, so that
		1404	low-priority connections can not be locked out forever under load. This
		1405	enables your program to keep a lower latency for important connections
		1406	during short periods of high load, while not completely locking out less
		1407	important ones.
402		1408
403		1409
404	=head1 WATCHER TYPES	1410	=head1 WATCHER TYPES
405		1411
406	This section describes each watcher in detail, but will not repeat	1412	This section describes each watcher in detail, but will not repeat
407	information given in the last section.	1413	information given in the last section. Any initialisation/set macros,
		1414	functions and members specific to the watcher type are explained.
408		1415
		1416	Members are additionally marked with either I<[read-only]>, meaning that,
		1417	while the watcher is active, you can look at the member and expect some
		1418	sensible content, but you must not modify it (you can modify it while the
		1419	watcher is stopped to your hearts content), or I<[read-write]>, which
		1420	means you can expect it to have some sensible content while the watcher
		1421	is active, but you can also modify it. Modifying it may not do something
		1422	sensible or take immediate effect (or do anything at all), but libev will
		1423	not crash or malfunction in any way.
		1424
		1425
409	=head2 C<ev_io> - is this file descriptor readable or writable	1426	=head2 C<ev_io> - is this file descriptor readable or writable?
410		1427
411	I/O watchers check whether a file descriptor is readable or writable	1428	I/O watchers check whether a file descriptor is readable or writable
412	in each iteration of the event loop (This behaviour is called	1429	in each iteration of the event loop, or, more precisely, when reading
413	level-triggering because you keep receiving events as long as the	1430	would not block the process and writing would at least be able to write
414	condition persists. Remember you can stop the watcher if you don't want to	1431	some data. This behaviour is called level-triggering because you keep
415	act on the event and neither want to receive future events).	1432	receiving events as long as the condition persists. Remember you can stop
		1433	the watcher if you don't want to act on the event and neither want to
		1434	receive future events.
416		1435
417	In general you can register as many read and/or write event watchers oer	1436	In general you can register as many read and/or write event watchers per
418	fd as you want (as long as you don't confuse yourself). Setting all file	1437	fd as you want (as long as you don't confuse yourself). Setting all file
419	descriptors to non-blocking mode is also usually a good idea (but not	1438	descriptors to non-blocking mode is also usually a good idea (but not
420	required if you know what you are doing).	1439	required if you know what you are doing).
421		1440
422	You have to be careful with dup'ed file descriptors, though. Some backends	1441	If you cannot use non-blocking mode, then force the use of a
423	(the linux epoll backend is a notable example) cannot handle dup'ed file	1442	known-to-be-good backend (at the time of this writing, this includes only
424	descriptors correctly if you register interest in two or more fds pointing	1443	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
425	to the same file/socket etc. description.	1444	descriptors for which non-blocking operation makes no sense (such as
		1445	files) - libev doesn't guarentee any specific behaviour in that case.
426		1446
427	If you must do this, then force the use of a known-to-be-good backend	1447	Another thing you have to watch out for is that it is quite easy to
428	(at the time of this writing, this includes only EVMETHOD_SELECT and	1448	receive "spurious" readiness notifications, that is your callback might
429	EVMETHOD_POLL).	1449	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1450	because there is no data. Not only are some backends known to create a
		1451	lot of those (for example Solaris ports), it is very easy to get into
		1452	this situation even with a relatively standard program structure. Thus
		1453	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1454	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1455
		1456	If you cannot run the fd in non-blocking mode (for example you should
		1457	not play around with an Xlib connection), then you have to separately
		1458	re-test whether a file descriptor is really ready with a known-to-be good
		1459	interface such as poll (fortunately in our Xlib example, Xlib already
		1460	does this on its own, so its quite safe to use). Some people additionally
		1461	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1462	indefinitely.
		1463
		1464	But really, best use non-blocking mode.
		1465
		1466	=head3 The special problem of disappearing file descriptors
		1467
		1468	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1469	descriptor (either due to calling C<close> explicitly or any other means,
		1470	such as C<dup2>). The reason is that you register interest in some file
		1471	descriptor, but when it goes away, the operating system will silently drop
		1472	this interest. If another file descriptor with the same number then is
		1473	registered with libev, there is no efficient way to see that this is, in
		1474	fact, a different file descriptor.
		1475
		1476	To avoid having to explicitly tell libev about such cases, libev follows
		1477	the following policy: Each time C<ev_io_set> is being called, libev
		1478	will assume that this is potentially a new file descriptor, otherwise
		1479	it is assumed that the file descriptor stays the same. That means that
		1480	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1481	descriptor even if the file descriptor number itself did not change.
		1482
		1483	This is how one would do it normally anyway, the important point is that
		1484	the libev application should not optimise around libev but should leave
		1485	optimisations to libev.
		1486
		1487	=head3 The special problem of dup'ed file descriptors
		1488
		1489	Some backends (e.g. epoll), cannot register events for file descriptors,
		1490	but only events for the underlying file descriptions. That means when you
		1491	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1492	events for them, only one file descriptor might actually receive events.
		1493
		1494	There is no workaround possible except not registering events
		1495	for potentially C<dup ()>'ed file descriptors, or to resort to
		1496	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1497
		1498	=head3 The special problem of fork
		1499
		1500	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1501	useless behaviour. Libev fully supports fork, but needs to be told about
		1502	it in the child.
		1503
		1504	To support fork in your programs, you either have to call
		1505	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1506	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1507	C<EVBACKEND_POLL>.
		1508
		1509	=head3 The special problem of SIGPIPE
		1510
		1511	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1512	when writing to a pipe whose other end has been closed, your program gets
		1513	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1514	this is sensible behaviour, for daemons, this is usually undesirable.
		1515
		1516	So when you encounter spurious, unexplained daemon exits, make sure you
		1517	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1518	somewhere, as that would have given you a big clue).
		1519
		1520
		1521	=head3 Watcher-Specific Functions
430		1522
431	=over 4	1523	=over 4
432		1524
433	=item ev_io_init (ev_io *, callback, int fd, int events)	1525	=item ev_io_init (ev_io *, callback, int fd, int events)
434		1526
435	=item ev_io_set (ev_io *, int fd, int events)	1527	=item ev_io_set (ev_io *, int fd, int events)
436		1528
437	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1529	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
438	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1530	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
439	EV_WRITE> to receive the given events.	1531	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1532
		1533	=item int fd [read-only]
		1534
		1535	The file descriptor being watched.
		1536
		1537	=item int events [read-only]
		1538
		1539	The events being watched.
440		1540
441	=back	1541	=back
442		1542
		1543	=head3 Examples
		1544
		1545	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1546	readable, but only once. Since it is likely line-buffered, you could
		1547	attempt to read a whole line in the callback.
		1548
		1549	static void
		1550	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1551	{
		1552	ev_io_stop (loop, w);
		1553	.. read from stdin here (or from w->fd) and handle any I/O errors
		1554	}
		1555
		1556	...
		1557	struct ev_loop *loop = ev_default_init (0);
		1558	ev_io stdin_readable;
		1559	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1560	ev_io_start (loop, &stdin_readable);
		1561	ev_loop (loop, 0);
		1562
		1563
443	=head2 C<ev_timer> - relative and optionally recurring timeouts	1564	=head2 C<ev_timer> - relative and optionally repeating timeouts
444		1565
445	Timer watchers are simple relative timers that generate an event after a	1566	Timer watchers are simple relative timers that generate an event after a
446	given time, and optionally repeating in regular intervals after that.	1567	given time, and optionally repeating in regular intervals after that.
447		1568
448	The timers are based on real time, that is, if you register an event that	1569	The timers are based on real time, that is, if you register an event that
449	times out after an hour and youreset your system clock to last years	1570	times out after an hour and you reset your system clock to January last
450	time, it will still time out after (roughly) and hour. "Roughly" because	1571	year, it will still time out after (roughly) one hour. "Roughly" because
451	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1572	detecting time jumps is hard, and some inaccuracies are unavoidable (the
452	monotonic clock option helps a lot here).	1573	monotonic clock option helps a lot here).
		1574
		1575	The callback is guaranteed to be invoked only I<after> its timeout has
		1576	passed (not I<at>, so on systems with very low-resolution clocks this
		1577	might introduce a small delay). If multiple timers become ready during the
		1578	same loop iteration then the ones with earlier time-out values are invoked
		1579	before ones of the same priority with later time-out values (but this is
		1580	no longer true when a callback calls C<ev_loop> recursively).
		1581
		1582	=head3 Be smart about timeouts
		1583
		1584	Many real-world problems involve some kind of timeout, usually for error
		1585	recovery. A typical example is an HTTP request - if the other side hangs,
		1586	you want to raise some error after a while.
		1587
		1588	What follows are some ways to handle this problem, from obvious and
		1589	inefficient to smart and efficient.
		1590
		1591	In the following, a 60 second activity timeout is assumed - a timeout that
		1592	gets reset to 60 seconds each time there is activity (e.g. each time some
		1593	data or other life sign was received).
		1594
		1595	=over 4
		1596
		1597	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1598
		1599	This is the most obvious, but not the most simple way: In the beginning,
		1600	start the watcher:
		1601
		1602	ev_timer_init (timer, callback, 60., 0.);
		1603	ev_timer_start (loop, timer);
		1604
		1605	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1606	and start it again:
		1607
		1608	ev_timer_stop (loop, timer);
		1609	ev_timer_set (timer, 60., 0.);
		1610	ev_timer_start (loop, timer);
		1611
		1612	This is relatively simple to implement, but means that each time there is
		1613	some activity, libev will first have to remove the timer from its internal
		1614	data structure and then add it again. Libev tries to be fast, but it's
		1615	still not a constant-time operation.
		1616
		1617	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1618
		1619	This is the easiest way, and involves using C<ev_timer_again> instead of
		1620	C<ev_timer_start>.
		1621
		1622	To implement this, configure an C<ev_timer> with a C<repeat> value
		1623	of C<60> and then call C<ev_timer_again> at start and each time you
		1624	successfully read or write some data. If you go into an idle state where
		1625	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1626	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1627
		1628	That means you can ignore both the C<ev_timer_start> function and the
		1629	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1630	member and C<ev_timer_again>.
		1631
		1632	At start:
		1633
		1634	ev_init (timer, callback);
		1635	timer->repeat = 60.;
		1636	ev_timer_again (loop, timer);
		1637
		1638	Each time there is some activity:
		1639
		1640	ev_timer_again (loop, timer);
		1641
		1642	It is even possible to change the time-out on the fly, regardless of
		1643	whether the watcher is active or not:
		1644
		1645	timer->repeat = 30.;
		1646	ev_timer_again (loop, timer);
		1647
		1648	This is slightly more efficient then stopping/starting the timer each time
		1649	you want to modify its timeout value, as libev does not have to completely
		1650	remove and re-insert the timer from/into its internal data structure.
		1651
		1652	It is, however, even simpler than the "obvious" way to do it.
		1653
		1654	=item 3. Let the timer time out, but then re-arm it as required.
		1655
		1656	This method is more tricky, but usually most efficient: Most timeouts are
		1657	relatively long compared to the intervals between other activity - in
		1658	our example, within 60 seconds, there are usually many I/O events with
		1659	associated activity resets.
		1660
		1661	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1662	but remember the time of last activity, and check for a real timeout only
		1663	within the callback:
		1664
		1665	ev_tstamp last_activity; // time of last activity
		1666
		1667	static void
		1668	callback (EV_P_ ev_timer *w, int revents)
		1669	{
		1670	ev_tstamp now = ev_now (EV_A);
		1671	ev_tstamp timeout = last_activity + 60.;
		1672
		1673	// if last_activity + 60. is older than now, we did time out
		1674	if (timeout < now)
		1675	{
		1676	// timeout occured, take action
		1677	}
		1678	else
		1679	{
		1680	// callback was invoked, but there was some activity, re-arm
		1681	// the watcher to fire in last_activity + 60, which is
		1682	// guaranteed to be in the future, so "again" is positive:
		1683	w->repeat = timeout - now;
		1684	ev_timer_again (EV_A_ w);
		1685	}
		1686	}
		1687
		1688	To summarise the callback: first calculate the real timeout (defined
		1689	as "60 seconds after the last activity"), then check if that time has
		1690	been reached, which means something I<did>, in fact, time out. Otherwise
		1691	the callback was invoked too early (C<timeout> is in the future), so
		1692	re-schedule the timer to fire at that future time, to see if maybe we have
		1693	a timeout then.
		1694
		1695	Note how C<ev_timer_again> is used, taking advantage of the
		1696	C<ev_timer_again> optimisation when the timer is already running.
		1697
		1698	This scheme causes more callback invocations (about one every 60 seconds
		1699	minus half the average time between activity), but virtually no calls to
		1700	libev to change the timeout.
		1701
		1702	To start the timer, simply initialise the watcher and set C<last_activity>
		1703	to the current time (meaning we just have some activity :), then call the
		1704	callback, which will "do the right thing" and start the timer:
		1705
		1706	ev_init (timer, callback);
		1707	last_activity = ev_now (loop);
		1708	callback (loop, timer, EV_TIMEOUT);
		1709
		1710	And when there is some activity, simply store the current time in
		1711	C<last_activity>, no libev calls at all:
		1712
		1713	last_actiivty = ev_now (loop);
		1714
		1715	This technique is slightly more complex, but in most cases where the
		1716	time-out is unlikely to be triggered, much more efficient.
		1717
		1718	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1719	callback :) - just change the timeout and invoke the callback, which will
		1720	fix things for you.
		1721
		1722	=item 4. Wee, just use a double-linked list for your timeouts.
		1723
		1724	If there is not one request, but many thousands (millions...), all
		1725	employing some kind of timeout with the same timeout value, then one can
		1726	do even better:
		1727
		1728	When starting the timeout, calculate the timeout value and put the timeout
		1729	at the I<end> of the list.
		1730
		1731	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1732	the list is expected to fire (for example, using the technique #3).
		1733
		1734	When there is some activity, remove the timer from the list, recalculate
		1735	the timeout, append it to the end of the list again, and make sure to
		1736	update the C<ev_timer> if it was taken from the beginning of the list.
		1737
		1738	This way, one can manage an unlimited number of timeouts in O(1) time for
		1739	starting, stopping and updating the timers, at the expense of a major
		1740	complication, and having to use a constant timeout. The constant timeout
		1741	ensures that the list stays sorted.
		1742
		1743	=back
		1744
		1745	So which method the best?
		1746
		1747	Method #2 is a simple no-brain-required solution that is adequate in most
		1748	situations. Method #3 requires a bit more thinking, but handles many cases
		1749	better, and isn't very complicated either. In most case, choosing either
		1750	one is fine, with #3 being better in typical situations.
		1751
		1752	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1753	rather complicated, but extremely efficient, something that really pays
		1754	off after the first million or so of active timers, i.e. it's usually
		1755	overkill :)
		1756
		1757	=head3 The special problem of time updates
		1758
		1759	Establishing the current time is a costly operation (it usually takes at
		1760	least two system calls): EV therefore updates its idea of the current
		1761	time only before and after C<ev_loop> collects new events, which causes a
		1762	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1763	lots of events in one iteration.
453		1764
454	The relative timeouts are calculated relative to the C<ev_now ()>	1765	The relative timeouts are calculated relative to the C<ev_now ()>
455	time. This is usually the right thing as this timestamp refers to the time	1766	time. This is usually the right thing as this timestamp refers to the time
456	of the event triggering whatever timeout you are modifying/starting. If	1767	of the event triggering whatever timeout you are modifying/starting. If
457	you suspect event processing to be delayed and you *need* to base the timeout	1768	you suspect event processing to be delayed and you I<need> to base the
458	ion the current time, use something like this to adjust for this:	1769	timeout on the current time, use something like this to adjust for this:
459		1770
460	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1771	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
461		1772
		1773	If the event loop is suspended for a long time, you can also force an
		1774	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1775	()>.
		1776
		1777	=head3 The special problems of suspended animation
		1778
		1779	When you leave the server world it is quite customary to hit machines that
		1780	can suspend/hibernate - what happens to the clocks during such a suspend?
		1781
		1782	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1783	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1784	to run until the system is suspended, but they will not advance while the
		1785	system is suspended. That means, on resume, it will be as if the program
		1786	was frozen for a few seconds, but the suspend time will not be counted
		1787	towards C<ev_timer> when a monotonic clock source is used. The real time
		1788	clock advanced as expected, but if it is used as sole clocksource, then a
		1789	long suspend would be detected as a time jump by libev, and timers would
		1790	be adjusted accordingly.
		1791
		1792	I would not be surprised to see different behaviour in different between
		1793	operating systems, OS versions or even different hardware.
		1794
		1795	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1796	time jump in the monotonic clocks and the realtime clock. If the program
		1797	is suspended for a very long time, and monotonic clock sources are in use,
		1798	then you can expect C<ev_timer>s to expire as the full suspension time
		1799	will be counted towards the timers. When no monotonic clock source is in
		1800	use, then libev will again assume a timejump and adjust accordingly.
		1801
		1802	It might be beneficial for this latter case to call C<ev_suspend>
		1803	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1804	deterministic behaviour in this case (you can do nothing against
		1805	C<SIGSTOP>).
		1806
		1807	=head3 Watcher-Specific Functions and Data Members
		1808
462	=over 4	1809	=over 4
463		1810
464	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1811	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
465		1812
466	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1813	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
467		1814
468	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1815	Configure the timer to trigger after C<after> seconds. If C<repeat>
469	C<0.>, then it will automatically be stopped. If it is positive, then the	1816	is C<0.>, then it will automatically be stopped once the timeout is
470	timer will automatically be configured to trigger again C<repeat> seconds	1817	reached. If it is positive, then the timer will automatically be
471	later, again, and again, until stopped manually.	1818	configured to trigger again C<repeat> seconds later, again, and again,
		1819	until stopped manually.
472		1820
473	The timer itself will do a best-effort at avoiding drift, that is, if you	1821	The timer itself will do a best-effort at avoiding drift, that is, if
474	configure a timer to trigger every 10 seconds, then it will trigger at	1822	you configure a timer to trigger every 10 seconds, then it will normally
475	exactly 10 second intervals. If, however, your program cannot keep up with	1823	trigger at exactly 10 second intervals. If, however, your program cannot
476	the timer (ecause it takes longer than those 10 seconds to do stuff) the	1824	keep up with the timer (because it takes longer than those 10 seconds to
477	timer will not fire more than once per event loop iteration.	1825	do stuff) the timer will not fire more than once per event loop iteration.
478		1826
479	=item ev_timer_again (loop)	1827	=item ev_timer_again (loop, ev_timer *)
480		1828
481	This will act as if the timer timed out and restart it again if it is	1829	This will act as if the timer timed out and restart it again if it is
482	repeating. The exact semantics are:	1830	repeating. The exact semantics are:
483		1831
		1832	If the timer is pending, its pending status is cleared.
		1833
484	If the timer is started but nonrepeating, stop it.	1834	If the timer is started but non-repeating, stop it (as if it timed out).
485		1835
486	If the timer is repeating, either start it if necessary (with the repeat	1836	If the timer is repeating, either start it if necessary (with the
487	value), or reset the running timer to the repeat value.	1837	C<repeat> value), or reset the running timer to the C<repeat> value.
488		1838
489	This sounds a bit complicated, but here is a useful and typical	1839	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
490	example: Imagine you have a tcp connection and you want a so-called idle	1840	usage example.
491	timeout, that is, you want to be called when there have been, say, 60	1841
492	seconds of inactivity on the socket. The easiest way to do this is to	1842	=item ev_timer_remaining (loop, ev_timer *)
493	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1843
494	time you successfully read or write some data. If you go into an idle	1844	Returns the remaining time until a timer fires. If the timer is active,
495	state where you do not expect data to travel on the socket, you can stop	1845	then this time is relative to the current event loop time, otherwise it's
496	the timer, and again will automatically restart it if need be.	1846	the timeout value currently configured.
		1847
		1848	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1849	C<5>. When the timer is started and one second passes, C<ev_timer_remain>
		1850	will return C<4>. When the timer expires and is restarted, it will return
		1851	roughly C<7> (likely slightly less as callback invocation takes some time,
		1852	too), and so on.
		1853
		1854	=item ev_tstamp repeat [read-write]
		1855
		1856	The current C<repeat> value. Will be used each time the watcher times out
		1857	or C<ev_timer_again> is called, and determines the next timeout (if any),
		1858	which is also when any modifications are taken into account.
497		1859
498	=back	1860	=back
499		1861
		1862	=head3 Examples
		1863
		1864	Example: Create a timer that fires after 60 seconds.
		1865
		1866	static void
		1867	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1868	{
		1869	.. one minute over, w is actually stopped right here
		1870	}
		1871
		1872	ev_timer mytimer;
		1873	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1874	ev_timer_start (loop, &mytimer);
		1875
		1876	Example: Create a timeout timer that times out after 10 seconds of
		1877	inactivity.
		1878
		1879	static void
		1880	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1881	{
		1882	.. ten seconds without any activity
		1883	}
		1884
		1885	ev_timer mytimer;
		1886	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1887	ev_timer_again (&mytimer); /* start timer */
		1888	ev_loop (loop, 0);
		1889
		1890	// and in some piece of code that gets executed on any "activity":
		1891	// reset the timeout to start ticking again at 10 seconds
		1892	ev_timer_again (&mytimer);
		1893
		1894
500	=head2 C<ev_periodic> - to cron or not to cron	1895	=head2 C<ev_periodic> - to cron or not to cron?
501		1896
502	Periodic watchers are also timers of a kind, but they are very versatile	1897	Periodic watchers are also timers of a kind, but they are very versatile
503	(and unfortunately a bit complex).	1898	(and unfortunately a bit complex).
504		1899
505	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1900	Unlike C<ev_timer>, periodic watchers are not based on real time (or
506	but on wallclock time (absolute time). You can tell a periodic watcher	1901	relative time, the physical time that passes) but on wall clock time
507	to trigger "at" some specific point in time. For example, if you tell a	1902	(absolute time, the thing you can read on your calender or clock). The
508	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1903	difference is that wall clock time can run faster or slower than real
509	+ 10.>) and then reset your system clock to the last year, then it will	1904	time, and time jumps are not uncommon (e.g. when you adjust your
510	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1905	wrist-watch).
511	roughly 10 seconds later and of course not if you reset your system time
512	again).
513		1906
514	They can also be used to implement vastly more complex timers, such as	1907	You can tell a periodic watcher to trigger after some specific point
		1908	in time: for example, if you tell a periodic watcher to trigger "in 10
		1909	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1910	not a delay) and then reset your system clock to January of the previous
		1911	year, then it will take a year or more to trigger the event (unlike an
		1912	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1913	it, as it uses a relative timeout).
		1914
		1915	C<ev_periodic> watchers can also be used to implement vastly more complex
515	triggering an event on eahc midnight, local time.	1916	timers, such as triggering an event on each "midnight, local time", or
		1917	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1918	those cannot react to time jumps.
		1919
		1920	As with timers, the callback is guaranteed to be invoked only when the
		1921	point in time where it is supposed to trigger has passed. If multiple
		1922	timers become ready during the same loop iteration then the ones with
		1923	earlier time-out values are invoked before ones with later time-out values
		1924	(but this is no longer true when a callback calls C<ev_loop> recursively).
		1925
		1926	=head3 Watcher-Specific Functions and Data Members
516		1927
517	=over 4	1928	=over 4
518		1929
519	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1930	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
520		1931
521	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1932	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
522		1933
523	Lots of arguments, lets sort it out... There are basically three modes of	1934	Lots of arguments, let's sort it out... There are basically three modes of
524	operation, and we will explain them from simplest to complex:	1935	operation, and we will explain them from simplest to most complex:
525
526		1936
527	=over 4	1937	=over 4
528		1938
529	=item * absolute timer (interval = reschedule_cb = 0)	1939	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
530		1940
531	In this configuration the watcher triggers an event at the wallclock time	1941	In this configuration the watcher triggers an event after the wall clock
532	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1942	time C<offset> has passed. It will not repeat and will not adjust when a
533	that is, if it is to be run at January 1st 2011 then it will run when the	1943	time jump occurs, that is, if it is to be run at January 1st 2011 then it
534	system time reaches or surpasses this time.	1944	will be stopped and invoked when the system clock reaches or surpasses
		1945	this point in time.
535		1946
536	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1947	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
537		1948
538	In this mode the watcher will always be scheduled to time out at the next	1949	In this mode the watcher will always be scheduled to time out at the next
539	C<at + N * interval> time (for some integer N) and then repeat, regardless	1950	C<offset + N * interval> time (for some integer N, which can also be
540	of any time jumps.	1951	negative) and then repeat, regardless of any time jumps. The C<offset>
		1952	argument is merely an offset into the C<interval> periods.
541		1953
542	This can be used to create timers that do not drift with respect to system	1954	This can be used to create timers that do not drift with respect to the
543	time:	1955	system clock, for example, here is an C<ev_periodic> that triggers each
		1956	hour, on the hour (with respect to UTC):
544		1957
545	ev_periodic_set (&periodic, 0., 3600., 0);	1958	ev_periodic_set (&periodic, 0., 3600., 0);
546		1959
547	This doesn't mean there will always be 3600 seconds in between triggers,	1960	This doesn't mean there will always be 3600 seconds in between triggers,
548	but only that the the callback will be called when the system time shows a	1961	but only that the callback will be called when the system time shows a
549	full hour (UTC), or more correctly, when the system time is evenly divisible	1962	full hour (UTC), or more correctly, when the system time is evenly divisible
550	by 3600.	1963	by 3600.
551		1964
552	Another way to think about it (for the mathematically inclined) is that	1965	Another way to think about it (for the mathematically inclined) is that
553	C<ev_periodic> will try to run the callback in this mode at the next possible	1966	C<ev_periodic> will try to run the callback in this mode at the next possible
554	time where C<time = at (mod interval)>, regardless of any time jumps.	1967	time where C<time = offset (mod interval)>, regardless of any time jumps.
555		1968
		1969	For numerical stability it is preferable that the C<offset> value is near
		1970	C<ev_now ()> (the current time), but there is no range requirement for
		1971	this value, and in fact is often specified as zero.
		1972
		1973	Note also that there is an upper limit to how often a timer can fire (CPU
		1974	speed for example), so if C<interval> is very small then timing stability
		1975	will of course deteriorate. Libev itself tries to be exact to be about one
		1976	millisecond (if the OS supports it and the machine is fast enough).
		1977
556	=item * manual reschedule mode (reschedule_cb = callback)	1978	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
557		1979
558	In this mode the values for C<interval> and C<at> are both being	1980	In this mode the values for C<interval> and C<offset> are both being
559	ignored. Instead, each time the periodic watcher gets scheduled, the	1981	ignored. Instead, each time the periodic watcher gets scheduled, the
560	reschedule callback will be called with the watcher as first, and the	1982	reschedule callback will be called with the watcher as first, and the
561	current time as second argument.	1983	current time as second argument.
562		1984
563	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	1985	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
564	periodic watcher, ever, or make any event loop modifications>. If you need	1986	or make ANY other event loop modifications whatsoever, unless explicitly
		1987	allowed by documentation here>.
		1988
565	to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards.	1989	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1990	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1991	only event loop modification you are allowed to do).
566		1992
567	Also, I<< this callback must always return a time that is later than the
568	passed C<now> value >>. Not even C<now> itself will be ok.
569
570	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1993	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
571	ev_tstamp now)>, e.g.:	1994	*w, ev_tstamp now)>, e.g.:
572		1995
		1996	static ev_tstamp
573	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1997	my_rescheduler (ev_periodic *w, ev_tstamp now)
574	{	1998	{
575	return now + 60.;	1999	return now + 60.;
576	}	2000	}
577		2001
578	It must return the next time to trigger, based on the passed time value	2002	It must return the next time to trigger, based on the passed time value
579	(that is, the lowest time value larger than to the second argument). It	2003	(that is, the lowest time value larger than to the second argument). It
580	will usually be called just before the callback will be triggered, but	2004	will usually be called just before the callback will be triggered, but
581	might be called at other times, too.	2005	might be called at other times, too.
582		2006
		2007	NOTE: I<< This callback must always return a time that is higher than or
		2008	equal to the passed C<now> value >>.
		2009
583	This can be used to create very complex timers, such as a timer that	2010	This can be used to create very complex timers, such as a timer that
584	triggers on each midnight, local time. To do this, you would calculate the	2011	triggers on "next midnight, local time". To do this, you would calculate the
585	next midnight after C<now> and return the timestamp value for this. How you do this	2012	next midnight after C<now> and return the timestamp value for this. How
586	is, again, up to you (but it is not trivial).	2013	you do this is, again, up to you (but it is not trivial, which is the main
		2014	reason I omitted it as an example).
587		2015
588	=back	2016	=back
589		2017
590	=item ev_periodic_again (loop, ev_periodic *)	2018	=item ev_periodic_again (loop, ev_periodic *)
591		2019
592	Simply stops and restarts the periodic watcher again. This is only useful	2020	Simply stops and restarts the periodic watcher again. This is only useful
593	when you changed some parameters or the reschedule callback would return	2021	when you changed some parameters or the reschedule callback would return
594	a different time than the last time it was called (e.g. in a crond like	2022	a different time than the last time it was called (e.g. in a crond like
595	program when the crontabs have changed).	2023	program when the crontabs have changed).
596		2024
		2025	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2026
		2027	When active, returns the absolute time that the watcher is supposed
		2028	to trigger next. This is not the same as the C<offset> argument to
		2029	C<ev_periodic_set>, but indeed works even in interval and manual
		2030	rescheduling modes.
		2031
		2032	=item ev_tstamp offset [read-write]
		2033
		2034	When repeating, this contains the offset value, otherwise this is the
		2035	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2036	although libev might modify this value for better numerical stability).
		2037
		2038	Can be modified any time, but changes only take effect when the periodic
		2039	timer fires or C<ev_periodic_again> is being called.
		2040
		2041	=item ev_tstamp interval [read-write]
		2042
		2043	The current interval value. Can be modified any time, but changes only
		2044	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2045	called.
		2046
		2047	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2048
		2049	The current reschedule callback, or C<0>, if this functionality is
		2050	switched off. Can be changed any time, but changes only take effect when
		2051	the periodic timer fires or C<ev_periodic_again> is being called.
		2052
597	=back	2053	=back
598		2054
		2055	=head3 Examples
		2056
		2057	Example: Call a callback every hour, or, more precisely, whenever the
		2058	system time is divisible by 3600. The callback invocation times have
		2059	potentially a lot of jitter, but good long-term stability.
		2060
		2061	static void
		2062	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		2063	{
		2064	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2065	}
		2066
		2067	ev_periodic hourly_tick;
		2068	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2069	ev_periodic_start (loop, &hourly_tick);
		2070
		2071	Example: The same as above, but use a reschedule callback to do it:
		2072
		2073	#include <math.h>
		2074
		2075	static ev_tstamp
		2076	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2077	{
		2078	return now + (3600. - fmod (now, 3600.));
		2079	}
		2080
		2081	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2082
		2083	Example: Call a callback every hour, starting now:
		2084
		2085	ev_periodic hourly_tick;
		2086	ev_periodic_init (&hourly_tick, clock_cb,
		2087	fmod (ev_now (loop), 3600.), 3600., 0);
		2088	ev_periodic_start (loop, &hourly_tick);
		2089
		2090
599	=head2 C<ev_signal> - signal me when a signal gets signalled	2091	=head2 C<ev_signal> - signal me when a signal gets signalled!
600		2092
601	Signal watchers will trigger an event when the process receives a specific	2093	Signal watchers will trigger an event when the process receives a specific
602	signal one or more times. Even though signals are very asynchronous, libev	2094	signal one or more times. Even though signals are very asynchronous, libev
603	will try it's best to deliver signals synchronously, i.e. as part of the	2095	will try it's best to deliver signals synchronously, i.e. as part of the
604	normal event processing, like any other event.	2096	normal event processing, like any other event.
605		2097
		2098	If you want signals to be delivered truly asynchronously, just use
		2099	C<sigaction> as you would do without libev and forget about sharing
		2100	the signal. You can even use C<ev_async> from a signal handler to
		2101	synchronously wake up an event loop.
		2102
606	You can configure as many watchers as you like per signal. Only when the	2103	You can configure as many watchers as you like for the same signal, but
		2104	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2105	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2106	C<SIGINT> in both the default loop and another loop at the same time. At
		2107	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2108
607	first watcher gets started will libev actually register a signal watcher	2109	When the first watcher gets started will libev actually register something
608	with the kernel (thus it coexists with your own signal handlers as long	2110	with the kernel (thus it coexists with your own signal handlers as long as
609	as you don't register any with libev). Similarly, when the last signal	2111	you don't register any with libev for the same signal).
610	watcher for a signal is stopped libev will reset the signal handler to	2112
611	SIG_DFL (regardless of what it was set to before).	2113	Both the signal mask state (C<sigprocmask>) and the signal handler state
		2114	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2115	sotpping it again), that is, libev might or might not block the signal,
		2116	and might or might not set or restore the installed signal handler.
		2117
		2118	If possible and supported, libev will install its handlers with
		2119	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2120	not be unduly interrupted. If you have a problem with system calls getting
		2121	interrupted by signals you can block all signals in an C<ev_check> watcher
		2122	and unblock them in an C<ev_prepare> watcher.
		2123
		2124	=head3 Watcher-Specific Functions and Data Members
612		2125
613	=over 4	2126	=over 4
614		2127
615	=item ev_signal_init (ev_signal *, callback, int signum)	2128	=item ev_signal_init (ev_signal *, callback, int signum)
616		2129
617	=item ev_signal_set (ev_signal *, int signum)	2130	=item ev_signal_set (ev_signal *, int signum)
618		2131
619	Configures the watcher to trigger on the given signal number (usually one	2132	Configures the watcher to trigger on the given signal number (usually one
620	of the C<SIGxxx> constants).	2133	of the C<SIGxxx> constants).
621		2134
		2135	=item int signum [read-only]
		2136
		2137	The signal the watcher watches out for.
		2138
622	=back	2139	=back
623		2140
		2141	=head3 Examples
		2142
		2143	Example: Try to exit cleanly on SIGINT.
		2144
		2145	static void
		2146	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2147	{
		2148	ev_unloop (loop, EVUNLOOP_ALL);
		2149	}
		2150
		2151	ev_signal signal_watcher;
		2152	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2153	ev_signal_start (loop, &signal_watcher);
		2154
		2155
624	=head2 C<ev_child> - wait for pid status changes	2156	=head2 C<ev_child> - watch out for process status changes
625		2157
626	Child watchers trigger when your process receives a SIGCHLD in response to	2158	Child watchers trigger when your process receives a SIGCHLD in response to
627	some child status changes (most typically when a child of yours dies).	2159	some child status changes (most typically when a child of yours dies or
		2160	exits). It is permissible to install a child watcher I<after> the child
		2161	has been forked (which implies it might have already exited), as long
		2162	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2163	forking and then immediately registering a watcher for the child is fine,
		2164	but forking and registering a watcher a few event loop iterations later or
		2165	in the next callback invocation is not.
		2166
		2167	Only the default event loop is capable of handling signals, and therefore
		2168	you can only register child watchers in the default event loop.
		2169
		2170	Due to some design glitches inside libev, child watchers will always be
		2171	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2172	libev)
		2173
		2174	=head3 Process Interaction
		2175
		2176	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2177	initialised. This is necessary to guarantee proper behaviour even if the
		2178	first child watcher is started after the child exits. The occurrence
		2179	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2180	synchronously as part of the event loop processing. Libev always reaps all
		2181	children, even ones not watched.
		2182
		2183	=head3 Overriding the Built-In Processing
		2184
		2185	Libev offers no special support for overriding the built-in child
		2186	processing, but if your application collides with libev's default child
		2187	handler, you can override it easily by installing your own handler for
		2188	C<SIGCHLD> after initialising the default loop, and making sure the
		2189	default loop never gets destroyed. You are encouraged, however, to use an
		2190	event-based approach to child reaping and thus use libev's support for
		2191	that, so other libev users can use C<ev_child> watchers freely.
		2192
		2193	=head3 Stopping the Child Watcher
		2194
		2195	Currently, the child watcher never gets stopped, even when the
		2196	child terminates, so normally one needs to stop the watcher in the
		2197	callback. Future versions of libev might stop the watcher automatically
		2198	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2199	problem).
		2200
		2201	=head3 Watcher-Specific Functions and Data Members
628		2202
629	=over 4	2203	=over 4
630		2204
631	=item ev_child_init (ev_child *, callback, int pid)	2205	=item ev_child_init (ev_child *, callback, int pid, int trace)
632		2206
633	=item ev_child_set (ev_child *, int pid)	2207	=item ev_child_set (ev_child *, int pid, int trace)
634		2208
635	Configures the watcher to wait for status changes of process C<pid> (or	2209	Configures the watcher to wait for status changes of process C<pid> (or
636	I<any> process if C<pid> is specified as C<0>). The callback can look	2210	I<any> process if C<pid> is specified as C<0>). The callback can look
637	at the C<rstatus> member of the C<ev_child> watcher structure to see	2211	at the C<rstatus> member of the C<ev_child> watcher structure to see
638	the status word (use the macros from C<sys/wait.h> and see your systems	2212	the status word (use the macros from C<sys/wait.h> and see your systems
639	C<waitpid> documentation). The C<rpid> member contains the pid of the	2213	C<waitpid> documentation). The C<rpid> member contains the pid of the
640	process causing the status change.	2214	process causing the status change. C<trace> must be either C<0> (only
		2215	activate the watcher when the process terminates) or C<1> (additionally
		2216	activate the watcher when the process is stopped or continued).
		2217
		2218	=item int pid [read-only]
		2219
		2220	The process id this watcher watches out for, or C<0>, meaning any process id.
		2221
		2222	=item int rpid [read-write]
		2223
		2224	The process id that detected a status change.
		2225
		2226	=item int rstatus [read-write]
		2227
		2228	The process exit/trace status caused by C<rpid> (see your systems
		2229	C<waitpid> and C<sys/wait.h> documentation for details).
641		2230
642	=back	2231	=back
643		2232
		2233	=head3 Examples
		2234
		2235	Example: C<fork()> a new process and install a child handler to wait for
		2236	its completion.
		2237
		2238	ev_child cw;
		2239
		2240	static void
		2241	child_cb (EV_P_ ev_child *w, int revents)
		2242	{
		2243	ev_child_stop (EV_A_ w);
		2244	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2245	}
		2246
		2247	pid_t pid = fork ();
		2248
		2249	if (pid < 0)
		2250	// error
		2251	else if (pid == 0)
		2252	{
		2253	// the forked child executes here
		2254	exit (1);
		2255	}
		2256	else
		2257	{
		2258	ev_child_init (&cw, child_cb, pid, 0);
		2259	ev_child_start (EV_DEFAULT_ &cw);
		2260	}
		2261
		2262
		2263	=head2 C<ev_stat> - did the file attributes just change?
		2264
		2265	This watches a file system path for attribute changes. That is, it calls
		2266	C<stat> on that path in regular intervals (or when the OS says it changed)
		2267	and sees if it changed compared to the last time, invoking the callback if
		2268	it did.
		2269
		2270	The path does not need to exist: changing from "path exists" to "path does
		2271	not exist" is a status change like any other. The condition "path does not
		2272	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2273	C<st_nlink> field being zero (which is otherwise always forced to be at
		2274	least one) and all the other fields of the stat buffer having unspecified
		2275	contents.
		2276
		2277	The path I<must not> end in a slash or contain special components such as
		2278	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2279	your working directory changes, then the behaviour is undefined.
		2280
		2281	Since there is no portable change notification interface available, the
		2282	portable implementation simply calls C<stat(2)> regularly on the path
		2283	to see if it changed somehow. You can specify a recommended polling
		2284	interval for this case. If you specify a polling interval of C<0> (highly
		2285	recommended!) then a I<suitable, unspecified default> value will be used
		2286	(which you can expect to be around five seconds, although this might
		2287	change dynamically). Libev will also impose a minimum interval which is
		2288	currently around C<0.1>, but that's usually overkill.
		2289
		2290	This watcher type is not meant for massive numbers of stat watchers,
		2291	as even with OS-supported change notifications, this can be
		2292	resource-intensive.
		2293
		2294	At the time of this writing, the only OS-specific interface implemented
		2295	is the Linux inotify interface (implementing kqueue support is left as an
		2296	exercise for the reader. Note, however, that the author sees no way of
		2297	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2298
		2299	=head3 ABI Issues (Largefile Support)
		2300
		2301	Libev by default (unless the user overrides this) uses the default
		2302	compilation environment, which means that on systems with large file
		2303	support disabled by default, you get the 32 bit version of the stat
		2304	structure. When using the library from programs that change the ABI to
		2305	use 64 bit file offsets the programs will fail. In that case you have to
		2306	compile libev with the same flags to get binary compatibility. This is
		2307	obviously the case with any flags that change the ABI, but the problem is
		2308	most noticeably displayed with ev_stat and large file support.
		2309
		2310	The solution for this is to lobby your distribution maker to make large
		2311	file interfaces available by default (as e.g. FreeBSD does) and not
		2312	optional. Libev cannot simply switch on large file support because it has
		2313	to exchange stat structures with application programs compiled using the
		2314	default compilation environment.
		2315
		2316	=head3 Inotify and Kqueue
		2317
		2318	When C<inotify (7)> support has been compiled into libev and present at
		2319	runtime, it will be used to speed up change detection where possible. The
		2320	inotify descriptor will be created lazily when the first C<ev_stat>
		2321	watcher is being started.
		2322
		2323	Inotify presence does not change the semantics of C<ev_stat> watchers
		2324	except that changes might be detected earlier, and in some cases, to avoid
		2325	making regular C<stat> calls. Even in the presence of inotify support
		2326	there are many cases where libev has to resort to regular C<stat> polling,
		2327	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2328	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2329	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2330	xfs are fully working) libev usually gets away without polling.
		2331
		2332	There is no support for kqueue, as apparently it cannot be used to
		2333	implement this functionality, due to the requirement of having a file
		2334	descriptor open on the object at all times, and detecting renames, unlinks
		2335	etc. is difficult.
		2336
		2337	=head3 C<stat ()> is a synchronous operation
		2338
		2339	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2340	the process. The exception are C<ev_stat> watchers - those call C<stat
		2341	()>, which is a synchronous operation.
		2342
		2343	For local paths, this usually doesn't matter: unless the system is very
		2344	busy or the intervals between stat's are large, a stat call will be fast,
		2345	as the path data is usually in memory already (except when starting the
		2346	watcher).
		2347
		2348	For networked file systems, calling C<stat ()> can block an indefinite
		2349	time due to network issues, and even under good conditions, a stat call
		2350	often takes multiple milliseconds.
		2351
		2352	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2353	paths, although this is fully supported by libev.
		2354
		2355	=head3 The special problem of stat time resolution
		2356
		2357	The C<stat ()> system call only supports full-second resolution portably,
		2358	and even on systems where the resolution is higher, most file systems
		2359	still only support whole seconds.
		2360
		2361	That means that, if the time is the only thing that changes, you can
		2362	easily miss updates: on the first update, C<ev_stat> detects a change and
		2363	calls your callback, which does something. When there is another update
		2364	within the same second, C<ev_stat> will be unable to detect unless the
		2365	stat data does change in other ways (e.g. file size).
		2366
		2367	The solution to this is to delay acting on a change for slightly more
		2368	than a second (or till slightly after the next full second boundary), using
		2369	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2370	ev_timer_again (loop, w)>).
		2371
		2372	The C<.02> offset is added to work around small timing inconsistencies
		2373	of some operating systems (where the second counter of the current time
		2374	might be be delayed. One such system is the Linux kernel, where a call to
		2375	C<gettimeofday> might return a timestamp with a full second later than
		2376	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2377	update file times then there will be a small window where the kernel uses
		2378	the previous second to update file times but libev might already execute
		2379	the timer callback).
		2380
		2381	=head3 Watcher-Specific Functions and Data Members
		2382
		2383	=over 4
		2384
		2385	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2386
		2387	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2388
		2389	Configures the watcher to wait for status changes of the given
		2390	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2391	be detected and should normally be specified as C<0> to let libev choose
		2392	a suitable value. The memory pointed to by C<path> must point to the same
		2393	path for as long as the watcher is active.
		2394
		2395	The callback will receive an C<EV_STAT> event when a change was detected,
		2396	relative to the attributes at the time the watcher was started (or the
		2397	last change was detected).
		2398
		2399	=item ev_stat_stat (loop, ev_stat *)
		2400
		2401	Updates the stat buffer immediately with new values. If you change the
		2402	watched path in your callback, you could call this function to avoid
		2403	detecting this change (while introducing a race condition if you are not
		2404	the only one changing the path). Can also be useful simply to find out the
		2405	new values.
		2406
		2407	=item ev_statdata attr [read-only]
		2408
		2409	The most-recently detected attributes of the file. Although the type is
		2410	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2411	suitable for your system, but you can only rely on the POSIX-standardised
		2412	members to be present. If the C<st_nlink> member is C<0>, then there was
		2413	some error while C<stat>ing the file.
		2414
		2415	=item ev_statdata prev [read-only]
		2416
		2417	The previous attributes of the file. The callback gets invoked whenever
		2418	C<prev> != C<attr>, or, more precisely, one or more of these members
		2419	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2420	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2421
		2422	=item ev_tstamp interval [read-only]
		2423
		2424	The specified interval.
		2425
		2426	=item const char *path [read-only]
		2427
		2428	The file system path that is being watched.
		2429
		2430	=back
		2431
		2432	=head3 Examples
		2433
		2434	Example: Watch C</etc/passwd> for attribute changes.
		2435
		2436	static void
		2437	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2438	{
		2439	/* /etc/passwd changed in some way */
		2440	if (w->attr.st_nlink)
		2441	{
		2442	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2443	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2444	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2445	}
		2446	else
		2447	/* you shalt not abuse printf for puts */
		2448	puts ("wow, /etc/passwd is not there, expect problems. "
		2449	"if this is windows, they already arrived\n");
		2450	}
		2451
		2452	...
		2453	ev_stat passwd;
		2454
		2455	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2456	ev_stat_start (loop, &passwd);
		2457
		2458	Example: Like above, but additionally use a one-second delay so we do not
		2459	miss updates (however, frequent updates will delay processing, too, so
		2460	one might do the work both on C<ev_stat> callback invocation I<and> on
		2461	C<ev_timer> callback invocation).
		2462
		2463	static ev_stat passwd;
		2464	static ev_timer timer;
		2465
		2466	static void
		2467	timer_cb (EV_P_ ev_timer *w, int revents)
		2468	{
		2469	ev_timer_stop (EV_A_ w);
		2470
		2471	/* now it's one second after the most recent passwd change */
		2472	}
		2473
		2474	static void
		2475	stat_cb (EV_P_ ev_stat *w, int revents)
		2476	{
		2477	/* reset the one-second timer */
		2478	ev_timer_again (EV_A_ &timer);
		2479	}
		2480
		2481	...
		2482	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2483	ev_stat_start (loop, &passwd);
		2484	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2485
		2486
644	=head2 C<ev_idle> - when you've got nothing better to do	2487	=head2 C<ev_idle> - when you've got nothing better to do...
645		2488
646	Idle watchers trigger events when there are no other events are pending	2489	Idle watchers trigger events when no other events of the same or higher
647	(prepare, check and other idle watchers do not count). That is, as long	2490	priority are pending (prepare, check and other idle watchers do not count
648	as your process is busy handling sockets or timeouts (or even signals,	2491	as receiving "events").
649	imagine) it will not be triggered. But when your process is idle all idle	2492
650	watchers are being called again and again, once per event loop iteration -	2493	That is, as long as your process is busy handling sockets or timeouts
		2494	(or even signals, imagine) of the same or higher priority it will not be
		2495	triggered. But when your process is idle (or only lower-priority watchers
		2496	are pending), the idle watchers are being called once per event loop
651	until stopped, that is, or your process receives more events and becomes	2497	iteration - until stopped, that is, or your process receives more events
652	busy.	2498	and becomes busy again with higher priority stuff.
653		2499
654	The most noteworthy effect is that as long as any idle watchers are	2500	The most noteworthy effect is that as long as any idle watchers are
655	active, the process will not block when waiting for new events.	2501	active, the process will not block when waiting for new events.
656		2502
657	Apart from keeping your process non-blocking (which is a useful	2503	Apart from keeping your process non-blocking (which is a useful
658	effect on its own sometimes), idle watchers are a good place to do	2504	effect on its own sometimes), idle watchers are a good place to do
659	"pseudo-background processing", or delay processing stuff to after the	2505	"pseudo-background processing", or delay processing stuff to after the
660	event loop has handled all outstanding events.	2506	event loop has handled all outstanding events.
661		2507
		2508	=head3 Watcher-Specific Functions and Data Members
		2509
662	=over 4	2510	=over 4
663		2511
664	=item ev_idle_init (ev_signal *, callback)	2512	=item ev_idle_init (ev_idle *, callback)
665		2513
666	Initialises and configures the idle watcher - it has no parameters of any	2514	Initialises and configures the idle watcher - it has no parameters of any
667	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2515	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
668	believe me.	2516	believe me.
669		2517
670	=back	2518	=back
671		2519
		2520	=head3 Examples
		2521
		2522	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2523	callback, free it. Also, use no error checking, as usual.
		2524
		2525	static void
		2526	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2527	{
		2528	free (w);
		2529	// now do something you wanted to do when the program has
		2530	// no longer anything immediate to do.
		2531	}
		2532
		2533	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2534	ev_idle_init (idle_watcher, idle_cb);
		2535	ev_idle_start (loop, idle_watcher);
		2536
		2537
672	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2538	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
673		2539
674	Prepare and check watchers are usually (but not always) used in tandem:	2540	Prepare and check watchers are usually (but not always) used in pairs:
675	Prepare watchers get invoked before the process blocks and check watchers	2541	prepare watchers get invoked before the process blocks and check watchers
676	afterwards.	2542	afterwards.
677		2543
		2544	You I<must not> call C<ev_loop> or similar functions that enter
		2545	the current event loop from either C<ev_prepare> or C<ev_check>
		2546	watchers. Other loops than the current one are fine, however. The
		2547	rationale behind this is that you do not need to check for recursion in
		2548	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2549	C<ev_check> so if you have one watcher of each kind they will always be
		2550	called in pairs bracketing the blocking call.
		2551
678	Their main purpose is to integrate other event mechanisms into libev. This	2552	Their main purpose is to integrate other event mechanisms into libev and
679	could be used, for example, to track variable changes, implement your own	2553	their use is somewhat advanced. They could be used, for example, to track
680	watchers, integrate net-snmp or a coroutine library and lots more.	2554	variable changes, implement your own watchers, integrate net-snmp or a
		2555	coroutine library and lots more. They are also occasionally useful if
		2556	you cache some data and want to flush it before blocking (for example,
		2557	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2558	watcher).
681		2559
682	This is done by examining in each prepare call which file descriptors need	2560	This is done by examining in each prepare call which file descriptors
683	to be watched by the other library, registering C<ev_io> watchers for	2561	need to be watched by the other library, registering C<ev_io> watchers
684	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2562	for them and starting an C<ev_timer> watcher for any timeouts (many
685	provide just this functionality). Then, in the check watcher you check for	2563	libraries provide exactly this functionality). Then, in the check watcher,
686	any events that occured (by checking the pending status of all watchers	2564	you check for any events that occurred (by checking the pending status
687	and stopping them) and call back into the library. The I/O and timer	2565	of all watchers and stopping them) and call back into the library. The
688	callbacks will never actually be called (but must be valid neverthelles,	2566	I/O and timer callbacks will never actually be called (but must be valid
689	because you never know, you know?).	2567	nevertheless, because you never know, you know?).
690		2568
691	As another example, the Perl Coro module uses these hooks to integrate	2569	As another example, the Perl Coro module uses these hooks to integrate
692	coroutines into libev programs, by yielding to other active coroutines	2570	coroutines into libev programs, by yielding to other active coroutines
693	during each prepare and only letting the process block if no coroutines	2571	during each prepare and only letting the process block if no coroutines
694	are ready to run (its actually more complicated, it only runs coroutines	2572	are ready to run (it's actually more complicated: it only runs coroutines
695	with priority higher than the event loop and one lower priority once,	2573	with priority higher than or equal to the event loop and one coroutine
696	using idle watchers to keep the event loop from blocking if lower-priority	2574	of lower priority, but only once, using idle watchers to keep the event
697	coroutines exist, thus mapping low-priority coroutines to idle/background	2575	loop from blocking if lower-priority coroutines are active, thus mapping
698	tasks).	2576	low-priority coroutines to idle/background tasks).
		2577
		2578	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2579	priority, to ensure that they are being run before any other watchers
		2580	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2581
		2582	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2583	activate ("feed") events into libev. While libev fully supports this, they
		2584	might get executed before other C<ev_check> watchers did their job. As
		2585	C<ev_check> watchers are often used to embed other (non-libev) event
		2586	loops those other event loops might be in an unusable state until their
		2587	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2588	others).
		2589
		2590	=head3 Watcher-Specific Functions and Data Members
699		2591
700	=over 4	2592	=over 4
701		2593
702	=item ev_prepare_init (ev_prepare *, callback)	2594	=item ev_prepare_init (ev_prepare *, callback)
703		2595
704	=item ev_check_init (ev_check *, callback)	2596	=item ev_check_init (ev_check *, callback)
705		2597
706	Initialises and configures the prepare or check watcher - they have no	2598	Initialises and configures the prepare or check watcher - they have no
707	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2599	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
708	macros, but using them is utterly, utterly and completely pointless.	2600	macros, but using them is utterly, utterly, utterly and completely
		2601	pointless.
709		2602
710	=back	2603	=back
711		2604
		2605	=head3 Examples
		2606
		2607	There are a number of principal ways to embed other event loops or modules
		2608	into libev. Here are some ideas on how to include libadns into libev
		2609	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2610	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2611	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2612	Glib event loop).
		2613
		2614	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2615	and in a check watcher, destroy them and call into libadns. What follows
		2616	is pseudo-code only of course. This requires you to either use a low
		2617	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2618	the callbacks for the IO/timeout watchers might not have been called yet.
		2619
		2620	static ev_io iow [nfd];
		2621	static ev_timer tw;
		2622
		2623	static void
		2624	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2625	{
		2626	}
		2627
		2628	// create io watchers for each fd and a timer before blocking
		2629	static void
		2630	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2631	{
		2632	int timeout = 3600000;
		2633	struct pollfd fds [nfd];
		2634	// actual code will need to loop here and realloc etc.
		2635	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2636
		2637	/* the callback is illegal, but won't be called as we stop during check */
		2638	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2639	ev_timer_start (loop, &tw);
		2640
		2641	// create one ev_io per pollfd
		2642	for (int i = 0; i < nfd; ++i)
		2643	{
		2644	ev_io_init (iow + i, io_cb, fds [i].fd,
		2645	((fds [i].events & POLLIN ? EV_READ : 0)
		2646	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2647
		2648	fds [i].revents = 0;
		2649	ev_io_start (loop, iow + i);
		2650	}
		2651	}
		2652
		2653	// stop all watchers after blocking
		2654	static void
		2655	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2656	{
		2657	ev_timer_stop (loop, &tw);
		2658
		2659	for (int i = 0; i < nfd; ++i)
		2660	{
		2661	// set the relevant poll flags
		2662	// could also call adns_processreadable etc. here
		2663	struct pollfd *fd = fds + i;
		2664	int revents = ev_clear_pending (iow + i);
		2665	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2666	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2667
		2668	// now stop the watcher
		2669	ev_io_stop (loop, iow + i);
		2670	}
		2671
		2672	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2673	}
		2674
		2675	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2676	in the prepare watcher and would dispose of the check watcher.
		2677
		2678	Method 3: If the module to be embedded supports explicit event
		2679	notification (libadns does), you can also make use of the actual watcher
		2680	callbacks, and only destroy/create the watchers in the prepare watcher.
		2681
		2682	static void
		2683	timer_cb (EV_P_ ev_timer *w, int revents)
		2684	{
		2685	adns_state ads = (adns_state)w->data;
		2686	update_now (EV_A);
		2687
		2688	adns_processtimeouts (ads, &tv_now);
		2689	}
		2690
		2691	static void
		2692	io_cb (EV_P_ ev_io *w, int revents)
		2693	{
		2694	adns_state ads = (adns_state)w->data;
		2695	update_now (EV_A);
		2696
		2697	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2698	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2699	}
		2700
		2701	// do not ever call adns_afterpoll
		2702
		2703	Method 4: Do not use a prepare or check watcher because the module you
		2704	want to embed is not flexible enough to support it. Instead, you can
		2705	override their poll function. The drawback with this solution is that the
		2706	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2707	this approach, effectively embedding EV as a client into the horrible
		2708	libglib event loop.
		2709
		2710	static gint
		2711	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2712	{
		2713	int got_events = 0;
		2714
		2715	for (n = 0; n < nfds; ++n)
		2716	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2717
		2718	if (timeout >= 0)
		2719	// create/start timer
		2720
		2721	// poll
		2722	ev_loop (EV_A_ 0);
		2723
		2724	// stop timer again
		2725	if (timeout >= 0)
		2726	ev_timer_stop (EV_A_ &to);
		2727
		2728	// stop io watchers again - their callbacks should have set
		2729	for (n = 0; n < nfds; ++n)
		2730	ev_io_stop (EV_A_ iow [n]);
		2731
		2732	return got_events;
		2733	}
		2734
		2735
		2736	=head2 C<ev_embed> - when one backend isn't enough...
		2737
		2738	This is a rather advanced watcher type that lets you embed one event loop
		2739	into another (currently only C<ev_io> events are supported in the embedded
		2740	loop, other types of watchers might be handled in a delayed or incorrect
		2741	fashion and must not be used).
		2742
		2743	There are primarily two reasons you would want that: work around bugs and
		2744	prioritise I/O.
		2745
		2746	As an example for a bug workaround, the kqueue backend might only support
		2747	sockets on some platform, so it is unusable as generic backend, but you
		2748	still want to make use of it because you have many sockets and it scales
		2749	so nicely. In this case, you would create a kqueue-based loop and embed
		2750	it into your default loop (which might use e.g. poll). Overall operation
		2751	will be a bit slower because first libev has to call C<poll> and then
		2752	C<kevent>, but at least you can use both mechanisms for what they are
		2753	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2754
		2755	As for prioritising I/O: under rare circumstances you have the case where
		2756	some fds have to be watched and handled very quickly (with low latency),
		2757	and even priorities and idle watchers might have too much overhead. In
		2758	this case you would put all the high priority stuff in one loop and all
		2759	the rest in a second one, and embed the second one in the first.
		2760
		2761	As long as the watcher is active, the callback will be invoked every
		2762	time there might be events pending in the embedded loop. The callback
		2763	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2764	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2765	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2766	to give the embedded loop strictly lower priority for example).
		2767
		2768	You can also set the callback to C<0>, in which case the embed watcher
		2769	will automatically execute the embedded loop sweep whenever necessary.
		2770
		2771	Fork detection will be handled transparently while the C<ev_embed> watcher
		2772	is active, i.e., the embedded loop will automatically be forked when the
		2773	embedding loop forks. In other cases, the user is responsible for calling
		2774	C<ev_loop_fork> on the embedded loop.
		2775
		2776	Unfortunately, not all backends are embeddable: only the ones returned by
		2777	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2778	portable one.
		2779
		2780	So when you want to use this feature you will always have to be prepared
		2781	that you cannot get an embeddable loop. The recommended way to get around
		2782	this is to have a separate variables for your embeddable loop, try to
		2783	create it, and if that fails, use the normal loop for everything.
		2784
		2785	=head3 C<ev_embed> and fork
		2786
		2787	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2788	automatically be applied to the embedded loop as well, so no special
		2789	fork handling is required in that case. When the watcher is not running,
		2790	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2791	as applicable.
		2792
		2793	=head3 Watcher-Specific Functions and Data Members
		2794
		2795	=over 4
		2796
		2797	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2798
		2799	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2800
		2801	Configures the watcher to embed the given loop, which must be
		2802	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2803	invoked automatically, otherwise it is the responsibility of the callback
		2804	to invoke it (it will continue to be called until the sweep has been done,
		2805	if you do not want that, you need to temporarily stop the embed watcher).
		2806
		2807	=item ev_embed_sweep (loop, ev_embed *)
		2808
		2809	Make a single, non-blocking sweep over the embedded loop. This works
		2810	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2811	appropriate way for embedded loops.
		2812
		2813	=item struct ev_loop *other [read-only]
		2814
		2815	The embedded event loop.
		2816
		2817	=back
		2818
		2819	=head3 Examples
		2820
		2821	Example: Try to get an embeddable event loop and embed it into the default
		2822	event loop. If that is not possible, use the default loop. The default
		2823	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2824	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2825	used).
		2826
		2827	struct ev_loop *loop_hi = ev_default_init (0);
		2828	struct ev_loop *loop_lo = 0;
		2829	ev_embed embed;
		2830
		2831	// see if there is a chance of getting one that works
		2832	// (remember that a flags value of 0 means autodetection)
		2833	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2834	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2835	: 0;
		2836
		2837	// if we got one, then embed it, otherwise default to loop_hi
		2838	if (loop_lo)
		2839	{
		2840	ev_embed_init (&embed, 0, loop_lo);
		2841	ev_embed_start (loop_hi, &embed);
		2842	}
		2843	else
		2844	loop_lo = loop_hi;
		2845
		2846	Example: Check if kqueue is available but not recommended and create
		2847	a kqueue backend for use with sockets (which usually work with any
		2848	kqueue implementation). Store the kqueue/socket-only event loop in
		2849	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2850
		2851	struct ev_loop *loop = ev_default_init (0);
		2852	struct ev_loop *loop_socket = 0;
		2853	ev_embed embed;
		2854
		2855	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2856	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2857	{
		2858	ev_embed_init (&embed, 0, loop_socket);
		2859	ev_embed_start (loop, &embed);
		2860	}
		2861
		2862	if (!loop_socket)
		2863	loop_socket = loop;
		2864
		2865	// now use loop_socket for all sockets, and loop for everything else
		2866
		2867
		2868	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2869
		2870	Fork watchers are called when a C<fork ()> was detected (usually because
		2871	whoever is a good citizen cared to tell libev about it by calling
		2872	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2873	event loop blocks next and before C<ev_check> watchers are being called,
		2874	and only in the child after the fork. If whoever good citizen calling
		2875	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2876	handlers will be invoked, too, of course.
		2877
		2878	=head3 The special problem of life after fork - how is it possible?
		2879
		2880	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2881	up/change the process environment, followed by a call to C<exec()>. This
		2882	sequence should be handled by libev without any problems.
		2883
		2884	This changes when the application actually wants to do event handling
		2885	in the child, or both parent in child, in effect "continuing" after the
		2886	fork.
		2887
		2888	The default mode of operation (for libev, with application help to detect
		2889	forks) is to duplicate all the state in the child, as would be expected
		2890	when I<either> the parent I<or> the child process continues.
		2891
		2892	When both processes want to continue using libev, then this is usually the
		2893	wrong result. In that case, usually one process (typically the parent) is
		2894	supposed to continue with all watchers in place as before, while the other
		2895	process typically wants to start fresh, i.e. without any active watchers.
		2896
		2897	The cleanest and most efficient way to achieve that with libev is to
		2898	simply create a new event loop, which of course will be "empty", and
		2899	use that for new watchers. This has the advantage of not touching more
		2900	memory than necessary, and thus avoiding the copy-on-write, and the
		2901	disadvantage of having to use multiple event loops (which do not support
		2902	signal watchers).
		2903
		2904	When this is not possible, or you want to use the default loop for
		2905	other reasons, then in the process that wants to start "fresh", call
		2906	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2907	the default loop will "orphan" (not stop) all registered watchers, so you
		2908	have to be careful not to execute code that modifies those watchers. Note
		2909	also that in that case, you have to re-register any signal watchers.
		2910
		2911	=head3 Watcher-Specific Functions and Data Members
		2912
		2913	=over 4
		2914
		2915	=item ev_fork_init (ev_signal *, callback)
		2916
		2917	Initialises and configures the fork watcher - it has no parameters of any
		2918	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2919	believe me.
		2920
		2921	=back
		2922
		2923
		2924	=head2 C<ev_async> - how to wake up another event loop
		2925
		2926	In general, you cannot use an C<ev_loop> from multiple threads or other
		2927	asynchronous sources such as signal handlers (as opposed to multiple event
		2928	loops - those are of course safe to use in different threads).
		2929
		2930	Sometimes, however, you need to wake up another event loop you do not
		2931	control, for example because it belongs to another thread. This is what
		2932	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2933	can signal it by calling C<ev_async_send>, which is thread- and signal
		2934	safe.
		2935
		2936	This functionality is very similar to C<ev_signal> watchers, as signals,
		2937	too, are asynchronous in nature, and signals, too, will be compressed
		2938	(i.e. the number of callback invocations may be less than the number of
		2939	C<ev_async_sent> calls).
		2940
		2941	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2942	just the default loop.
		2943
		2944	=head3 Queueing
		2945
		2946	C<ev_async> does not support queueing of data in any way. The reason
		2947	is that the author does not know of a simple (or any) algorithm for a
		2948	multiple-writer-single-reader queue that works in all cases and doesn't
		2949	need elaborate support such as pthreads.
		2950
		2951	That means that if you want to queue data, you have to provide your own
		2952	queue. But at least I can tell you how to implement locking around your
		2953	queue:
		2954
		2955	=over 4
		2956
		2957	=item queueing from a signal handler context
		2958
		2959	To implement race-free queueing, you simply add to the queue in the signal
		2960	handler but you block the signal handler in the watcher callback. Here is
		2961	an example that does that for some fictitious SIGUSR1 handler:
		2962
		2963	static ev_async mysig;
		2964
		2965	static void
		2966	sigusr1_handler (void)
		2967	{
		2968	sometype data;
		2969
		2970	// no locking etc.
		2971	queue_put (data);
		2972	ev_async_send (EV_DEFAULT_ &mysig);
		2973	}
		2974
		2975	static void
		2976	mysig_cb (EV_P_ ev_async *w, int revents)
		2977	{
		2978	sometype data;
		2979	sigset_t block, prev;
		2980
		2981	sigemptyset (&block);
		2982	sigaddset (&block, SIGUSR1);
		2983	sigprocmask (SIG_BLOCK, &block, &prev);
		2984
		2985	while (queue_get (&data))
		2986	process (data);
		2987
		2988	if (sigismember (&prev, SIGUSR1)
		2989	sigprocmask (SIG_UNBLOCK, &block, 0);
		2990	}
		2991
		2992	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2993	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2994	either...).
		2995
		2996	=item queueing from a thread context
		2997
		2998	The strategy for threads is different, as you cannot (easily) block
		2999	threads but you can easily preempt them, so to queue safely you need to
		3000	employ a traditional mutex lock, such as in this pthread example:
		3001
		3002	static ev_async mysig;
		3003	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3004
		3005	static void
		3006	otherthread (void)
		3007	{
		3008	// only need to lock the actual queueing operation
		3009	pthread_mutex_lock (&mymutex);
		3010	queue_put (data);
		3011	pthread_mutex_unlock (&mymutex);
		3012
		3013	ev_async_send (EV_DEFAULT_ &mysig);
		3014	}
		3015
		3016	static void
		3017	mysig_cb (EV_P_ ev_async *w, int revents)
		3018	{
		3019	pthread_mutex_lock (&mymutex);
		3020
		3021	while (queue_get (&data))
		3022	process (data);
		3023
		3024	pthread_mutex_unlock (&mymutex);
		3025	}
		3026
		3027	=back
		3028
		3029
		3030	=head3 Watcher-Specific Functions and Data Members
		3031
		3032	=over 4
		3033
		3034	=item ev_async_init (ev_async *, callback)
		3035
		3036	Initialises and configures the async watcher - it has no parameters of any
		3037	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3038	trust me.
		3039
		3040	=item ev_async_send (loop, ev_async *)
		3041
		3042	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3043	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3044	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3045	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3046	section below on what exactly this means).
		3047
		3048	Note that, as with other watchers in libev, multiple events might get
		3049	compressed into a single callback invocation (another way to look at this
		3050	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3051	reset when the event loop detects that).
		3052
		3053	This call incurs the overhead of a system call only once per event loop
		3054	iteration, so while the overhead might be noticeable, it doesn't apply to
		3055	repeated calls to C<ev_async_send> for the same event loop.
		3056
		3057	=item bool = ev_async_pending (ev_async *)
		3058
		3059	Returns a non-zero value when C<ev_async_send> has been called on the
		3060	watcher but the event has not yet been processed (or even noted) by the
		3061	event loop.
		3062
		3063	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3064	the loop iterates next and checks for the watcher to have become active,
		3065	it will reset the flag again. C<ev_async_pending> can be used to very
		3066	quickly check whether invoking the loop might be a good idea.
		3067
		3068	Not that this does I<not> check whether the watcher itself is pending,
		3069	only whether it has been requested to make this watcher pending: there
		3070	is a time window between the event loop checking and resetting the async
		3071	notification, and the callback being invoked.
		3072
		3073	=back
		3074
		3075
712	=head1 OTHER FUNCTIONS	3076	=head1 OTHER FUNCTIONS
713		3077
714	There are some other functions of possible interest. Described. Here. Now.	3078	There are some other functions of possible interest. Described. Here. Now.
715		3079
716	=over 4	3080	=over 4
717		3081
718	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3082	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
719		3083
720	This function combines a simple timer and an I/O watcher, calls your	3084	This function combines a simple timer and an I/O watcher, calls your
721	callback on whichever event happens first and automatically stop both	3085	callback on whichever event happens first and automatically stops both
722	watchers. This is useful if you want to wait for a single event on an fd	3086	watchers. This is useful if you want to wait for a single event on an fd
723	or timeout without havign to allocate/configure/start/stop/free one or	3087	or timeout without having to allocate/configure/start/stop/free one or
724	more watchers yourself.	3088	more watchers yourself.
725		3089
726	If C<fd> is less than 0, then no I/O watcher will be started and events	3090	If C<fd> is less than 0, then no I/O watcher will be started and the
727	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3091	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
728	C<events> set will be craeted and started.	3092	the given C<fd> and C<events> set will be created and started.
729		3093
730	If C<timeout> is less than 0, then no timeout watcher will be	3094	If C<timeout> is less than 0, then no timeout watcher will be
731	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3095	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
732	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3096	repeat = 0) will be started. C<0> is a valid timeout.
733	dubious value.
734		3097
735	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3098	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
736	passed an events set like normal event callbacks (with a combination of	3099	passed an C<revents> set like normal event callbacks (a combination of
737	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3100	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
738	value passed to C<ev_once>:	3101	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3102	a timeout and an io event at the same time - you probably should give io
		3103	events precedence.
739		3104
		3105	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3106
740	static void stdin_ready (int revents, void *arg)	3107	static void stdin_ready (int revents, void *arg)
741	{	3108	{
742	if (revents & EV_TIMEOUT)
743	/* doh, nothing entered */;
744	else if (revents & EV_READ)	3109	if (revents & EV_READ)
745	/* stdin might have data for us, joy! */;	3110	/* stdin might have data for us, joy! */;
		3111	else if (revents & EV_TIMEOUT)
		3112	/* doh, nothing entered */;
746	}	3113	}
747		3114
748	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3115	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
749		3116
750	=item ev_feed_event (loop, watcher, int events)	3117	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
751		3118
752	Feeds the given event set into the event loop, as if the specified event	3119	Feeds the given event set into the event loop, as if the specified event
753	had happened for the specified watcher (which must be a pointer to an	3120	had happened for the specified watcher (which must be a pointer to an
754	initialised but not necessarily started event watcher).	3121	initialised but not necessarily started event watcher).
755		3122
756	=item ev_feed_fd_event (loop, int fd, int revents)	3123	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
757		3124
758	Feed an event on the given fd, as if a file descriptor backend detected	3125	Feed an event on the given fd, as if a file descriptor backend detected
759	the given events it.	3126	the given events it.
760		3127
761	=item ev_feed_signal_event (loop, int signum)	3128	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
762		3129
763	Feed an event as if the given signal occured (loop must be the default loop!).	3130	Feed an event as if the given signal occurred (C<loop> must be the default
		3131	loop!).
764		3132
765	=back	3133	=back
766		3134
		3135
		3136	=head1 LIBEVENT EMULATION
		3137
		3138	Libev offers a compatibility emulation layer for libevent. It cannot
		3139	emulate the internals of libevent, so here are some usage hints:
		3140
		3141	=over 4
		3142
		3143	=item * Use it by including <event.h>, as usual.
		3144
		3145	=item * The following members are fully supported: ev_base, ev_callback,
		3146	ev_arg, ev_fd, ev_res, ev_events.
		3147
		3148	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		3149	maintained by libev, it does not work exactly the same way as in libevent (consider
		3150	it a private API).
		3151
		3152	=item * Priorities are not currently supported. Initialising priorities
		3153	will fail and all watchers will have the same priority, even though there
		3154	is an ev_pri field.
		3155
		3156	=item * In libevent, the last base created gets the signals, in libev, the
		3157	first base created (== the default loop) gets the signals.
		3158
		3159	=item * Other members are not supported.
		3160
		3161	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		3162	to use the libev header file and library.
		3163
		3164	=back
		3165
		3166	=head1 C++ SUPPORT
		3167
		3168	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3169	you to use some convenience methods to start/stop watchers and also change
		3170	the callback model to a model using method callbacks on objects.
		3171
		3172	To use it,
		3173
		3174	#include <ev++.h>
		3175
		3176	This automatically includes F<ev.h> and puts all of its definitions (many
		3177	of them macros) into the global namespace. All C++ specific things are
		3178	put into the C<ev> namespace. It should support all the same embedding
		3179	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3180
		3181	Care has been taken to keep the overhead low. The only data member the C++
		3182	classes add (compared to plain C-style watchers) is the event loop pointer
		3183	that the watcher is associated with (or no additional members at all if
		3184	you disable C<EV_MULTIPLICITY> when embedding libev).
		3185
		3186	Currently, functions, and static and non-static member functions can be
		3187	used as callbacks. Other types should be easy to add as long as they only
		3188	need one additional pointer for context. If you need support for other
		3189	types of functors please contact the author (preferably after implementing
		3190	it).
		3191
		3192	Here is a list of things available in the C<ev> namespace:
		3193
		3194	=over 4
		3195
		3196	=item C<ev::READ>, C<ev::WRITE> etc.
		3197
		3198	These are just enum values with the same values as the C<EV_READ> etc.
		3199	macros from F<ev.h>.
		3200
		3201	=item C<ev::tstamp>, C<ev::now>
		3202
		3203	Aliases to the same types/functions as with the C<ev_> prefix.
		3204
		3205	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3206
		3207	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3208	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3209	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3210	defines by many implementations.
		3211
		3212	All of those classes have these methods:
		3213
		3214	=over 4
		3215
		3216	=item ev::TYPE::TYPE ()
		3217
		3218	=item ev::TYPE::TYPE (struct ev_loop *)
		3219
		3220	=item ev::TYPE::~TYPE
		3221
		3222	The constructor (optionally) takes an event loop to associate the watcher
		3223	with. If it is omitted, it will use C<EV_DEFAULT>.
		3224
		3225	The constructor calls C<ev_init> for you, which means you have to call the
		3226	C<set> method before starting it.
		3227
		3228	It will not set a callback, however: You have to call the templated C<set>
		3229	method to set a callback before you can start the watcher.
		3230
		3231	(The reason why you have to use a method is a limitation in C++ which does
		3232	not allow explicit template arguments for constructors).
		3233
		3234	The destructor automatically stops the watcher if it is active.
		3235
		3236	=item w->set<class, &class::method> (object *)
		3237
		3238	This method sets the callback method to call. The method has to have a
		3239	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3240	first argument and the C<revents> as second. The object must be given as
		3241	parameter and is stored in the C<data> member of the watcher.
		3242
		3243	This method synthesizes efficient thunking code to call your method from
		3244	the C callback that libev requires. If your compiler can inline your
		3245	callback (i.e. it is visible to it at the place of the C<set> call and
		3246	your compiler is good :), then the method will be fully inlined into the
		3247	thunking function, making it as fast as a direct C callback.
		3248
		3249	Example: simple class declaration and watcher initialisation
		3250
		3251	struct myclass
		3252	{
		3253	void io_cb (ev::io &w, int revents) { }
		3254	}
		3255
		3256	myclass obj;
		3257	ev::io iow;
		3258	iow.set <myclass, &myclass::io_cb> (&obj);
		3259
		3260	=item w->set (object *)
		3261
		3262	This is an B<experimental> feature that might go away in a future version.
		3263
		3264	This is a variation of a method callback - leaving out the method to call
		3265	will default the method to C<operator ()>, which makes it possible to use
		3266	functor objects without having to manually specify the C<operator ()> all
		3267	the time. Incidentally, you can then also leave out the template argument
		3268	list.
		3269
		3270	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3271	int revents)>.
		3272
		3273	See the method-C<set> above for more details.
		3274
		3275	Example: use a functor object as callback.
		3276
		3277	struct myfunctor
		3278	{
		3279	void operator() (ev::io &w, int revents)
		3280	{
		3281	...
		3282	}
		3283	}
		3284
		3285	myfunctor f;
		3286
		3287	ev::io w;
		3288	w.set (&f);
		3289
		3290	=item w->set<function> (void *data = 0)
		3291
		3292	Also sets a callback, but uses a static method or plain function as
		3293	callback. The optional C<data> argument will be stored in the watcher's
		3294	C<data> member and is free for you to use.
		3295
		3296	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3297
		3298	See the method-C<set> above for more details.
		3299
		3300	Example: Use a plain function as callback.
		3301
		3302	static void io_cb (ev::io &w, int revents) { }
		3303	iow.set <io_cb> ();
		3304
		3305	=item w->set (struct ev_loop *)
		3306
		3307	Associates a different C<struct ev_loop> with this watcher. You can only
		3308	do this when the watcher is inactive (and not pending either).
		3309
		3310	=item w->set ([arguments])
		3311
		3312	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		3313	called at least once. Unlike the C counterpart, an active watcher gets
		3314	automatically stopped and restarted when reconfiguring it with this
		3315	method.
		3316
		3317	=item w->start ()
		3318
		3319	Starts the watcher. Note that there is no C<loop> argument, as the
		3320	constructor already stores the event loop.
		3321
		3322	=item w->stop ()
		3323
		3324	Stops the watcher if it is active. Again, no C<loop> argument.
		3325
		3326	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3327
		3328	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3329	C<ev_TYPE_again> function.
		3330
		3331	=item w->sweep () (C<ev::embed> only)
		3332
		3333	Invokes C<ev_embed_sweep>.
		3334
		3335	=item w->update () (C<ev::stat> only)
		3336
		3337	Invokes C<ev_stat_stat>.
		3338
		3339	=back
		3340
		3341	=back
		3342
		3343	Example: Define a class with an IO and idle watcher, start one of them in
		3344	the constructor.
		3345
		3346	class myclass
		3347	{
		3348	ev::io io ; void io_cb (ev::io &w, int revents);
		3349	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3350
		3351	myclass (int fd)
		3352	{
		3353	io .set <myclass, &myclass::io_cb > (this);
		3354	idle.set <myclass, &myclass::idle_cb> (this);
		3355
		3356	io.start (fd, ev::READ);
		3357	}
		3358	};
		3359
		3360
		3361	=head1 OTHER LANGUAGE BINDINGS
		3362
		3363	Libev does not offer other language bindings itself, but bindings for a
		3364	number of languages exist in the form of third-party packages. If you know
		3365	any interesting language binding in addition to the ones listed here, drop
		3366	me a note.
		3367
		3368	=over 4
		3369
		3370	=item Perl
		3371
		3372	The EV module implements the full libev API and is actually used to test
		3373	libev. EV is developed together with libev. Apart from the EV core module,
		3374	there are additional modules that implement libev-compatible interfaces
		3375	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3376	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3377	and C<EV::Glib>).
		3378
		3379	It can be found and installed via CPAN, its homepage is at
		3380	L<http://software.schmorp.de/pkg/EV>.
		3381
		3382	=item Python
		3383
		3384	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3385	seems to be quite complete and well-documented.
		3386
		3387	=item Ruby
		3388
		3389	Tony Arcieri has written a ruby extension that offers access to a subset
		3390	of the libev API and adds file handle abstractions, asynchronous DNS and
		3391	more on top of it. It can be found via gem servers. Its homepage is at
		3392	L<http://rev.rubyforge.org/>.
		3393
		3394	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3395	makes rev work even on mingw.
		3396
		3397	=item Haskell
		3398
		3399	A haskell binding to libev is available at
		3400	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3401
		3402	=item D
		3403
		3404	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3405	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3406
		3407	=item Ocaml
		3408
		3409	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3410	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3411
		3412	=back
		3413
		3414
		3415	=head1 MACRO MAGIC
		3416
		3417	Libev can be compiled with a variety of options, the most fundamental
		3418	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3419	functions and callbacks have an initial C<struct ev_loop *> argument.
		3420
		3421	To make it easier to write programs that cope with either variant, the
		3422	following macros are defined:
		3423
		3424	=over 4
		3425
		3426	=item C<EV_A>, C<EV_A_>
		3427
		3428	This provides the loop I<argument> for functions, if one is required ("ev
		3429	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3430	C<EV_A_> is used when other arguments are following. Example:
		3431
		3432	ev_unref (EV_A);
		3433	ev_timer_add (EV_A_ watcher);
		3434	ev_loop (EV_A_ 0);
		3435
		3436	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3437	which is often provided by the following macro.
		3438
		3439	=item C<EV_P>, C<EV_P_>
		3440
		3441	This provides the loop I<parameter> for functions, if one is required ("ev
		3442	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3443	C<EV_P_> is used when other parameters are following. Example:
		3444
		3445	// this is how ev_unref is being declared
		3446	static void ev_unref (EV_P);
		3447
		3448	// this is how you can declare your typical callback
		3449	static void cb (EV_P_ ev_timer *w, int revents)
		3450
		3451	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3452	suitable for use with C<EV_A>.
		3453
		3454	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3455
		3456	Similar to the other two macros, this gives you the value of the default
		3457	loop, if multiple loops are supported ("ev loop default").
		3458
		3459	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3460
		3461	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3462	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3463	is undefined when the default loop has not been initialised by a previous
		3464	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3465
		3466	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3467	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3468
		3469	=back
		3470
		3471	Example: Declare and initialise a check watcher, utilising the above
		3472	macros so it will work regardless of whether multiple loops are supported
		3473	or not.
		3474
		3475	static void
		3476	check_cb (EV_P_ ev_timer *w, int revents)
		3477	{
		3478	ev_check_stop (EV_A_ w);
		3479	}
		3480
		3481	ev_check check;
		3482	ev_check_init (&check, check_cb);
		3483	ev_check_start (EV_DEFAULT_ &check);
		3484	ev_loop (EV_DEFAULT_ 0);
		3485
		3486	=head1 EMBEDDING
		3487
		3488	Libev can (and often is) directly embedded into host
		3489	applications. Examples of applications that embed it include the Deliantra
		3490	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3491	and rxvt-unicode.
		3492
		3493	The goal is to enable you to just copy the necessary files into your
		3494	source directory without having to change even a single line in them, so
		3495	you can easily upgrade by simply copying (or having a checked-out copy of
		3496	libev somewhere in your source tree).
		3497
		3498	=head2 FILESETS
		3499
		3500	Depending on what features you need you need to include one or more sets of files
		3501	in your application.
		3502
		3503	=head3 CORE EVENT LOOP
		3504
		3505	To include only the libev core (all the C<ev_*> functions), with manual
		3506	configuration (no autoconf):
		3507
		3508	#define EV_STANDALONE 1
		3509	#include "ev.c"
		3510
		3511	This will automatically include F<ev.h>, too, and should be done in a
		3512	single C source file only to provide the function implementations. To use
		3513	it, do the same for F<ev.h> in all files wishing to use this API (best
		3514	done by writing a wrapper around F<ev.h> that you can include instead and
		3515	where you can put other configuration options):
		3516
		3517	#define EV_STANDALONE 1
		3518	#include "ev.h"
		3519
		3520	Both header files and implementation files can be compiled with a C++
		3521	compiler (at least, that's a stated goal, and breakage will be treated
		3522	as a bug).
		3523
		3524	You need the following files in your source tree, or in a directory
		3525	in your include path (e.g. in libev/ when using -Ilibev):
		3526
		3527	ev.h
		3528	ev.c
		3529	ev_vars.h
		3530	ev_wrap.h
		3531
		3532	ev_win32.c required on win32 platforms only
		3533
		3534	ev_select.c only when select backend is enabled (which is enabled by default)
		3535	ev_poll.c only when poll backend is enabled (disabled by default)
		3536	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3537	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3538	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3539
		3540	F<ev.c> includes the backend files directly when enabled, so you only need
		3541	to compile this single file.
		3542
		3543	=head3 LIBEVENT COMPATIBILITY API
		3544
		3545	To include the libevent compatibility API, also include:
		3546
		3547	#include "event.c"
		3548
		3549	in the file including F<ev.c>, and:
		3550
		3551	#include "event.h"
		3552
		3553	in the files that want to use the libevent API. This also includes F<ev.h>.
		3554
		3555	You need the following additional files for this:
		3556
		3557	event.h
		3558	event.c
		3559
		3560	=head3 AUTOCONF SUPPORT
		3561
		3562	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3563	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3564	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3565	include F<config.h> and configure itself accordingly.
		3566
		3567	For this of course you need the m4 file:
		3568
		3569	libev.m4
		3570
		3571	=head2 PREPROCESSOR SYMBOLS/MACROS
		3572
		3573	Libev can be configured via a variety of preprocessor symbols you have to
		3574	define before including any of its files. The default in the absence of
		3575	autoconf is documented for every option.
		3576
		3577	=over 4
		3578
		3579	=item EV_STANDALONE
		3580
		3581	Must always be C<1> if you do not use autoconf configuration, which
		3582	keeps libev from including F<config.h>, and it also defines dummy
		3583	implementations for some libevent functions (such as logging, which is not
		3584	supported). It will also not define any of the structs usually found in
		3585	F<event.h> that are not directly supported by the libev core alone.
		3586
		3587	In standalone mode, libev will still try to automatically deduce the
		3588	configuration, but has to be more conservative.
		3589
		3590	=item EV_USE_MONOTONIC
		3591
		3592	If defined to be C<1>, libev will try to detect the availability of the
		3593	monotonic clock option at both compile time and runtime. Otherwise no
		3594	use of the monotonic clock option will be attempted. If you enable this,
		3595	you usually have to link against librt or something similar. Enabling it
		3596	when the functionality isn't available is safe, though, although you have
		3597	to make sure you link against any libraries where the C<clock_gettime>
		3598	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3599
		3600	=item EV_USE_REALTIME
		3601
		3602	If defined to be C<1>, libev will try to detect the availability of the
		3603	real-time clock option at compile time (and assume its availability
		3604	at runtime if successful). Otherwise no use of the real-time clock
		3605	option will be attempted. This effectively replaces C<gettimeofday>
		3606	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3607	correctness. See the note about libraries in the description of
		3608	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3609	C<EV_USE_CLOCK_SYSCALL>.
		3610
		3611	=item EV_USE_CLOCK_SYSCALL
		3612
		3613	If defined to be C<1>, libev will try to use a direct syscall instead
		3614	of calling the system-provided C<clock_gettime> function. This option
		3615	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3616	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3617	programs needlessly. Using a direct syscall is slightly slower (in
		3618	theory), because no optimised vdso implementation can be used, but avoids
		3619	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3620	higher, as it simplifies linking (no need for C<-lrt>).
		3621
		3622	=item EV_USE_NANOSLEEP
		3623
		3624	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3625	and will use it for delays. Otherwise it will use C<select ()>.
		3626
		3627	=item EV_USE_EVENTFD
		3628
		3629	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3630	available and will probe for kernel support at runtime. This will improve
		3631	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3632	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3633	2.7 or newer, otherwise disabled.
		3634
		3635	=item EV_USE_SELECT
		3636
		3637	If undefined or defined to be C<1>, libev will compile in support for the
		3638	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3639	other method takes over, select will be it. Otherwise the select backend
		3640	will not be compiled in.
		3641
		3642	=item EV_SELECT_USE_FD_SET
		3643
		3644	If defined to C<1>, then the select backend will use the system C<fd_set>
		3645	structure. This is useful if libev doesn't compile due to a missing
		3646	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3647	on exotic systems. This usually limits the range of file descriptors to
		3648	some low limit such as 1024 or might have other limitations (winsocket
		3649	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3650	configures the maximum size of the C<fd_set>.
		3651
		3652	=item EV_SELECT_IS_WINSOCKET
		3653
		3654	When defined to C<1>, the select backend will assume that
		3655	select/socket/connect etc. don't understand file descriptors but
		3656	wants osf handles on win32 (this is the case when the select to
		3657	be used is the winsock select). This means that it will call
		3658	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3659	it is assumed that all these functions actually work on fds, even
		3660	on win32. Should not be defined on non-win32 platforms.
		3661
		3662	=item EV_FD_TO_WIN32_HANDLE
		3663
		3664	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3665	file descriptors to socket handles. When not defining this symbol (the
		3666	default), then libev will call C<_get_osfhandle>, which is usually
		3667	correct. In some cases, programs use their own file descriptor management,
		3668	in which case they can provide this function to map fds to socket handles.
		3669
		3670	=item EV_USE_POLL
		3671
		3672	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3673	backend. Otherwise it will be enabled on non-win32 platforms. It
		3674	takes precedence over select.
		3675
		3676	=item EV_USE_EPOLL
		3677
		3678	If defined to be C<1>, libev will compile in support for the Linux
		3679	C<epoll>(7) backend. Its availability will be detected at runtime,
		3680	otherwise another method will be used as fallback. This is the preferred
		3681	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3682	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3683
		3684	=item EV_USE_KQUEUE
		3685
		3686	If defined to be C<1>, libev will compile in support for the BSD style
		3687	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3688	otherwise another method will be used as fallback. This is the preferred
		3689	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3690	supports some types of fds correctly (the only platform we found that
		3691	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3692	not be used unless explicitly requested. The best way to use it is to find
		3693	out whether kqueue supports your type of fd properly and use an embedded
		3694	kqueue loop.
		3695
		3696	=item EV_USE_PORT
		3697
		3698	If defined to be C<1>, libev will compile in support for the Solaris
		3699	10 port style backend. Its availability will be detected at runtime,
		3700	otherwise another method will be used as fallback. This is the preferred
		3701	backend for Solaris 10 systems.
		3702
		3703	=item EV_USE_DEVPOLL
		3704
		3705	Reserved for future expansion, works like the USE symbols above.
		3706
		3707	=item EV_USE_INOTIFY
		3708
		3709	If defined to be C<1>, libev will compile in support for the Linux inotify
		3710	interface to speed up C<ev_stat> watchers. Its actual availability will
		3711	be detected at runtime. If undefined, it will be enabled if the headers
		3712	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3713
		3714	=item EV_ATOMIC_T
		3715
		3716	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3717	access is atomic with respect to other threads or signal contexts. No such
		3718	type is easily found in the C language, so you can provide your own type
		3719	that you know is safe for your purposes. It is used both for signal handler "locking"
		3720	as well as for signal and thread safety in C<ev_async> watchers.
		3721
		3722	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3723	(from F<signal.h>), which is usually good enough on most platforms.
		3724
		3725	=item EV_H
		3726
		3727	The name of the F<ev.h> header file used to include it. The default if
		3728	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3729	used to virtually rename the F<ev.h> header file in case of conflicts.
		3730
		3731	=item EV_CONFIG_H
		3732
		3733	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3734	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3735	C<EV_H>, above.
		3736
		3737	=item EV_EVENT_H
		3738
		3739	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3740	of how the F<event.h> header can be found, the default is C<"event.h">.
		3741
		3742	=item EV_PROTOTYPES
		3743
		3744	If defined to be C<0>, then F<ev.h> will not define any function
		3745	prototypes, but still define all the structs and other symbols. This is
		3746	occasionally useful if you want to provide your own wrapper functions
		3747	around libev functions.
		3748
		3749	=item EV_MULTIPLICITY
		3750
		3751	If undefined or defined to C<1>, then all event-loop-specific functions
		3752	will have the C<struct ev_loop *> as first argument, and you can create
		3753	additional independent event loops. Otherwise there will be no support
		3754	for multiple event loops and there is no first event loop pointer
		3755	argument. Instead, all functions act on the single default loop.
		3756
		3757	=item EV_MINPRI
		3758
		3759	=item EV_MAXPRI
		3760
		3761	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3762	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3763	provide for more priorities by overriding those symbols (usually defined
		3764	to be C<-2> and C<2>, respectively).
		3765
		3766	When doing priority-based operations, libev usually has to linearly search
		3767	all the priorities, so having many of them (hundreds) uses a lot of space
		3768	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3769	fine.
		3770
		3771	If your embedding application does not need any priorities, defining these
		3772	both to C<0> will save some memory and CPU.
		3773
		3774	=item EV_PERIODIC_ENABLE
		3775
		3776	If undefined or defined to be C<1>, then periodic timers are supported. If
		3777	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3778	code.
		3779
		3780	=item EV_IDLE_ENABLE
		3781
		3782	If undefined or defined to be C<1>, then idle watchers are supported. If
		3783	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3784	code.
		3785
		3786	=item EV_EMBED_ENABLE
		3787
		3788	If undefined or defined to be C<1>, then embed watchers are supported. If
		3789	defined to be C<0>, then they are not. Embed watchers rely on most other
		3790	watcher types, which therefore must not be disabled.
		3791
		3792	=item EV_STAT_ENABLE
		3793
		3794	If undefined or defined to be C<1>, then stat watchers are supported. If
		3795	defined to be C<0>, then they are not.
		3796
		3797	=item EV_FORK_ENABLE
		3798
		3799	If undefined or defined to be C<1>, then fork watchers are supported. If
		3800	defined to be C<0>, then they are not.
		3801
		3802	=item EV_ASYNC_ENABLE
		3803
		3804	If undefined or defined to be C<1>, then async watchers are supported. If
		3805	defined to be C<0>, then they are not.
		3806
		3807	=item EV_MINIMAL
		3808
		3809	If you need to shave off some kilobytes of code at the expense of some
		3810	speed (but with the full API), define this symbol to C<1>. Currently this
		3811	is used to override some inlining decisions, saves roughly 30% code size
		3812	on amd64. It also selects a much smaller 2-heap for timer management over
		3813	the default 4-heap.
		3814
		3815	You can save even more by disabling watcher types you do not need
		3816	and setting C<EV_MAXPRI> == C<EV_MINPRI>. Also, disabling C<assert>
		3817	(C<-DNDEBUG>) will usually reduce code size a lot.
		3818
		3819	Defining C<EV_MINIMAL> to C<2> will additionally reduce the core API to
		3820	provide a bare-bones event library. See C<ev.h> for details on what parts
		3821	of the API are still available, and do not complain if this subset changes
		3822	over time.
		3823
		3824	=item EV_NSIG
		3825
		3826	The highest supported signal number, +1 (or, the number of
		3827	signals): Normally, libev tries to deduce the maximum number of signals
		3828	automatically, but sometimes this fails, in which case it can be
		3829	specified. Also, using a lower number than detected (C<32> should be
		3830	good for about any system in existance) can save some memory, as libev
		3831	statically allocates some 12-24 bytes per signal number.
		3832
		3833	=item EV_PID_HASHSIZE
		3834
		3835	C<ev_child> watchers use a small hash table to distribute workload by
		3836	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3837	than enough. If you need to manage thousands of children you might want to
		3838	increase this value (I<must> be a power of two).
		3839
		3840	=item EV_INOTIFY_HASHSIZE
		3841
		3842	C<ev_stat> watchers use a small hash table to distribute workload by
		3843	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3844	usually more than enough. If you need to manage thousands of C<ev_stat>
		3845	watchers you might want to increase this value (I<must> be a power of
		3846	two).
		3847
		3848	=item EV_USE_4HEAP
		3849
		3850	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3851	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3852	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3853	faster performance with many (thousands) of watchers.
		3854
		3855	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3856	(disabled).
		3857
		3858	=item EV_HEAP_CACHE_AT
		3859
		3860	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3861	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3862	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3863	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3864	but avoids random read accesses on heap changes. This improves performance
		3865	noticeably with many (hundreds) of watchers.
		3866
		3867	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3868	(disabled).
		3869
		3870	=item EV_VERIFY
		3871
		3872	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3873	be done: If set to C<0>, no internal verification code will be compiled
		3874	in. If set to C<1>, then verification code will be compiled in, but not
		3875	called. If set to C<2>, then the internal verification code will be
		3876	called once per loop, which can slow down libev. If set to C<3>, then the
		3877	verification code will be called very frequently, which will slow down
		3878	libev considerably.
		3879
		3880	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3881	C<0>.
		3882
		3883	=item EV_COMMON
		3884
		3885	By default, all watchers have a C<void *data> member. By redefining
		3886	this macro to a something else you can include more and other types of
		3887	members. You have to define it each time you include one of the files,
		3888	though, and it must be identical each time.
		3889
		3890	For example, the perl EV module uses something like this:
		3891
		3892	#define EV_COMMON \
		3893	SV *self; /* contains this struct */ \
		3894	SV *cb_sv, *fh /* note no trailing ";" */
		3895
		3896	=item EV_CB_DECLARE (type)
		3897
		3898	=item EV_CB_INVOKE (watcher, revents)
		3899
		3900	=item ev_set_cb (ev, cb)
		3901
		3902	Can be used to change the callback member declaration in each watcher,
		3903	and the way callbacks are invoked and set. Must expand to a struct member
		3904	definition and a statement, respectively. See the F<ev.h> header file for
		3905	their default definitions. One possible use for overriding these is to
		3906	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3907	method calls instead of plain function calls in C++.
		3908
		3909	=back
		3910
		3911	=head2 EXPORTED API SYMBOLS
		3912
		3913	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3914	exported symbols, you can use the provided F<Symbol.*> files which list
		3915	all public symbols, one per line:
		3916
		3917	Symbols.ev for libev proper
		3918	Symbols.event for the libevent emulation
		3919
		3920	This can also be used to rename all public symbols to avoid clashes with
		3921	multiple versions of libev linked together (which is obviously bad in
		3922	itself, but sometimes it is inconvenient to avoid this).
		3923
		3924	A sed command like this will create wrapper C<#define>'s that you need to
		3925	include before including F<ev.h>:
		3926
		3927	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3928
		3929	This would create a file F<wrap.h> which essentially looks like this:
		3930
		3931	#define ev_backend myprefix_ev_backend
		3932	#define ev_check_start myprefix_ev_check_start
		3933	#define ev_check_stop myprefix_ev_check_stop
		3934	...
		3935
		3936	=head2 EXAMPLES
		3937
		3938	For a real-world example of a program the includes libev
		3939	verbatim, you can have a look at the EV perl module
		3940	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3941	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3942	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3943	will be compiled. It is pretty complex because it provides its own header
		3944	file.
		3945
		3946	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3947	that everybody includes and which overrides some configure choices:
		3948
		3949	#define EV_MINIMAL 1
		3950	#define EV_USE_POLL 0
		3951	#define EV_MULTIPLICITY 0
		3952	#define EV_PERIODIC_ENABLE 0
		3953	#define EV_STAT_ENABLE 0
		3954	#define EV_FORK_ENABLE 0
		3955	#define EV_CONFIG_H <config.h>
		3956	#define EV_MINPRI 0
		3957	#define EV_MAXPRI 0
		3958
		3959	#include "ev++.h"
		3960
		3961	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3962
		3963	#include "ev_cpp.h"
		3964	#include "ev.c"
		3965
		3966	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3967
		3968	=head2 THREADS AND COROUTINES
		3969
		3970	=head3 THREADS
		3971
		3972	All libev functions are reentrant and thread-safe unless explicitly
		3973	documented otherwise, but libev implements no locking itself. This means
		3974	that you can use as many loops as you want in parallel, as long as there
		3975	are no concurrent calls into any libev function with the same loop
		3976	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3977	of course): libev guarantees that different event loops share no data
		3978	structures that need any locking.
		3979
		3980	Or to put it differently: calls with different loop parameters can be done
		3981	concurrently from multiple threads, calls with the same loop parameter
		3982	must be done serially (but can be done from different threads, as long as
		3983	only one thread ever is inside a call at any point in time, e.g. by using
		3984	a mutex per loop).
		3985
		3986	Specifically to support threads (and signal handlers), libev implements
		3987	so-called C<ev_async> watchers, which allow some limited form of
		3988	concurrency on the same event loop, namely waking it up "from the
		3989	outside".
		3990
		3991	If you want to know which design (one loop, locking, or multiple loops
		3992	without or something else still) is best for your problem, then I cannot
		3993	help you, but here is some generic advice:
		3994
		3995	=over 4
		3996
		3997	=item * most applications have a main thread: use the default libev loop
		3998	in that thread, or create a separate thread running only the default loop.
		3999
		4000	This helps integrating other libraries or software modules that use libev
		4001	themselves and don't care/know about threading.
		4002
		4003	=item * one loop per thread is usually a good model.
		4004
		4005	Doing this is almost never wrong, sometimes a better-performance model
		4006	exists, but it is always a good start.
		4007
		4008	=item * other models exist, such as the leader/follower pattern, where one
		4009	loop is handed through multiple threads in a kind of round-robin fashion.
		4010
		4011	Choosing a model is hard - look around, learn, know that usually you can do
		4012	better than you currently do :-)
		4013
		4014	=item * often you need to talk to some other thread which blocks in the
		4015	event loop.
		4016
		4017	C<ev_async> watchers can be used to wake them up from other threads safely
		4018	(or from signal contexts...).
		4019
		4020	An example use would be to communicate signals or other events that only
		4021	work in the default loop by registering the signal watcher with the
		4022	default loop and triggering an C<ev_async> watcher from the default loop
		4023	watcher callback into the event loop interested in the signal.
		4024
		4025	=back
		4026
		4027	=head4 THREAD LOCKING EXAMPLE
		4028
		4029	Here is a fictitious example of how to run an event loop in a different
		4030	thread than where callbacks are being invoked and watchers are
		4031	created/added/removed.
		4032
		4033	For a real-world example, see the C<EV::Loop::Async> perl module,
		4034	which uses exactly this technique (which is suited for many high-level
		4035	languages).
		4036
		4037	The example uses a pthread mutex to protect the loop data, a condition
		4038	variable to wait for callback invocations, an async watcher to notify the
		4039	event loop thread and an unspecified mechanism to wake up the main thread.
		4040
		4041	First, you need to associate some data with the event loop:
		4042
		4043	typedef struct {
		4044	mutex_t lock; /* global loop lock */
		4045	ev_async async_w;
		4046	thread_t tid;
		4047	cond_t invoke_cv;
		4048	} userdata;
		4049
		4050	void prepare_loop (EV_P)
		4051	{
		4052	// for simplicity, we use a static userdata struct.
		4053	static userdata u;
		4054
		4055	ev_async_init (&u->async_w, async_cb);
		4056	ev_async_start (EV_A_ &u->async_w);
		4057
		4058	pthread_mutex_init (&u->lock, 0);
		4059	pthread_cond_init (&u->invoke_cv, 0);
		4060
		4061	// now associate this with the loop
		4062	ev_set_userdata (EV_A_ u);
		4063	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4064	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4065
		4066	// then create the thread running ev_loop
		4067	pthread_create (&u->tid, 0, l_run, EV_A);
		4068	}
		4069
		4070	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4071	solely to wake up the event loop so it takes notice of any new watchers
		4072	that might have been added:
		4073
		4074	static void
		4075	async_cb (EV_P_ ev_async *w, int revents)
		4076	{
		4077	// just used for the side effects
		4078	}
		4079
		4080	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4081	protecting the loop data, respectively.
		4082
		4083	static void
		4084	l_release (EV_P)
		4085	{
		4086	userdata *u = ev_userdata (EV_A);
		4087	pthread_mutex_unlock (&u->lock);
		4088	}
		4089
		4090	static void
		4091	l_acquire (EV_P)
		4092	{
		4093	userdata *u = ev_userdata (EV_A);
		4094	pthread_mutex_lock (&u->lock);
		4095	}
		4096
		4097	The event loop thread first acquires the mutex, and then jumps straight
		4098	into C<ev_loop>:
		4099
		4100	void *
		4101	l_run (void *thr_arg)
		4102	{
		4103	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4104
		4105	l_acquire (EV_A);
		4106	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4107	ev_loop (EV_A_ 0);
		4108	l_release (EV_A);
		4109
		4110	return 0;
		4111	}
		4112
		4113	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4114	signal the main thread via some unspecified mechanism (signals? pipe
		4115	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4116	have been called (in a while loop because a) spurious wakeups are possible
		4117	and b) skipping inter-thread-communication when there are no pending
		4118	watchers is very beneficial):
		4119
		4120	static void
		4121	l_invoke (EV_P)
		4122	{
		4123	userdata *u = ev_userdata (EV_A);
		4124
		4125	while (ev_pending_count (EV_A))
		4126	{
		4127	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4128	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4129	}
		4130	}
		4131
		4132	Now, whenever the main thread gets told to invoke pending watchers, it
		4133	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4134	thread to continue:
		4135
		4136	static void
		4137	real_invoke_pending (EV_P)
		4138	{
		4139	userdata *u = ev_userdata (EV_A);
		4140
		4141	pthread_mutex_lock (&u->lock);
		4142	ev_invoke_pending (EV_A);
		4143	pthread_cond_signal (&u->invoke_cv);
		4144	pthread_mutex_unlock (&u->lock);
		4145	}
		4146
		4147	Whenever you want to start/stop a watcher or do other modifications to an
		4148	event loop, you will now have to lock:
		4149
		4150	ev_timer timeout_watcher;
		4151	userdata *u = ev_userdata (EV_A);
		4152
		4153	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4154
		4155	pthread_mutex_lock (&u->lock);
		4156	ev_timer_start (EV_A_ &timeout_watcher);
		4157	ev_async_send (EV_A_ &u->async_w);
		4158	pthread_mutex_unlock (&u->lock);
		4159
		4160	Note that sending the C<ev_async> watcher is required because otherwise
		4161	an event loop currently blocking in the kernel will have no knowledge
		4162	about the newly added timer. By waking up the loop it will pick up any new
		4163	watchers in the next event loop iteration.
		4164
		4165	=head3 COROUTINES
		4166
		4167	Libev is very accommodating to coroutines ("cooperative threads"):
		4168	libev fully supports nesting calls to its functions from different
		4169	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		4170	different coroutines, and switch freely between both coroutines running
		4171	the loop, as long as you don't confuse yourself). The only exception is
		4172	that you must not do this from C<ev_periodic> reschedule callbacks.
		4173
		4174	Care has been taken to ensure that libev does not keep local state inside
		4175	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		4176	they do not call any callbacks.
		4177
		4178	=head2 COMPILER WARNINGS
		4179
		4180	Depending on your compiler and compiler settings, you might get no or a
		4181	lot of warnings when compiling libev code. Some people are apparently
		4182	scared by this.
		4183
		4184	However, these are unavoidable for many reasons. For one, each compiler
		4185	has different warnings, and each user has different tastes regarding
		4186	warning options. "Warn-free" code therefore cannot be a goal except when
		4187	targeting a specific compiler and compiler-version.
		4188
		4189	Another reason is that some compiler warnings require elaborate
		4190	workarounds, or other changes to the code that make it less clear and less
		4191	maintainable.
		4192
		4193	And of course, some compiler warnings are just plain stupid, or simply
		4194	wrong (because they don't actually warn about the condition their message
		4195	seems to warn about). For example, certain older gcc versions had some
		4196	warnings that resulted an extreme number of false positives. These have
		4197	been fixed, but some people still insist on making code warn-free with
		4198	such buggy versions.
		4199
		4200	While libev is written to generate as few warnings as possible,
		4201	"warn-free" code is not a goal, and it is recommended not to build libev
		4202	with any compiler warnings enabled unless you are prepared to cope with
		4203	them (e.g. by ignoring them). Remember that warnings are just that:
		4204	warnings, not errors, or proof of bugs.
		4205
		4206
		4207	=head2 VALGRIND
		4208
		4209	Valgrind has a special section here because it is a popular tool that is
		4210	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4211
		4212	If you think you found a bug (memory leak, uninitialised data access etc.)
		4213	in libev, then check twice: If valgrind reports something like:
		4214
		4215	==2274== definitely lost: 0 bytes in 0 blocks.
		4216	==2274== possibly lost: 0 bytes in 0 blocks.
		4217	==2274== still reachable: 256 bytes in 1 blocks.
		4218
		4219	Then there is no memory leak, just as memory accounted to global variables
		4220	is not a memleak - the memory is still being referenced, and didn't leak.
		4221
		4222	Similarly, under some circumstances, valgrind might report kernel bugs
		4223	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4224	although an acceptable workaround has been found here), or it might be
		4225	confused.
		4226
		4227	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4228	make it into some kind of religion.
		4229
		4230	If you are unsure about something, feel free to contact the mailing list
		4231	with the full valgrind report and an explanation on why you think this
		4232	is a bug in libev (best check the archives, too :). However, don't be
		4233	annoyed when you get a brisk "this is no bug" answer and take the chance
		4234	of learning how to interpret valgrind properly.
		4235
		4236	If you need, for some reason, empty reports from valgrind for your project
		4237	I suggest using suppression lists.
		4238
		4239
		4240	=head1 PORTABILITY NOTES
		4241
		4242	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4243
		4244	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4245	requires, and its I/O model is fundamentally incompatible with the POSIX
		4246	model. Libev still offers limited functionality on this platform in
		4247	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4248	descriptors. This only applies when using Win32 natively, not when using
		4249	e.g. cygwin.
		4250
		4251	Lifting these limitations would basically require the full
		4252	re-implementation of the I/O system. If you are into these kinds of
		4253	things, then note that glib does exactly that for you in a very portable
		4254	way (note also that glib is the slowest event library known to man).
		4255
		4256	There is no supported compilation method available on windows except
		4257	embedding it into other applications.
		4258
		4259	Sensible signal handling is officially unsupported by Microsoft - libev
		4260	tries its best, but under most conditions, signals will simply not work.
		4261
		4262	Not a libev limitation but worth mentioning: windows apparently doesn't
		4263	accept large writes: instead of resulting in a partial write, windows will
		4264	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4265	so make sure you only write small amounts into your sockets (less than a
		4266	megabyte seems safe, but this apparently depends on the amount of memory
		4267	available).
		4268
		4269	Due to the many, low, and arbitrary limits on the win32 platform and
		4270	the abysmal performance of winsockets, using a large number of sockets
		4271	is not recommended (and not reasonable). If your program needs to use
		4272	more than a hundred or so sockets, then likely it needs to use a totally
		4273	different implementation for windows, as libev offers the POSIX readiness
		4274	notification model, which cannot be implemented efficiently on windows
		4275	(due to Microsoft monopoly games).
		4276
		4277	A typical way to use libev under windows is to embed it (see the embedding
		4278	section for details) and use the following F<evwrap.h> header file instead
		4279	of F<ev.h>:
		4280
		4281	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4282	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4283
		4284	#include "ev.h"
		4285
		4286	And compile the following F<evwrap.c> file into your project (make sure
		4287	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4288
		4289	#include "evwrap.h"
		4290	#include "ev.c"
		4291
		4292	=over 4
		4293
		4294	=item The winsocket select function
		4295
		4296	The winsocket C<select> function doesn't follow POSIX in that it
		4297	requires socket I<handles> and not socket I<file descriptors> (it is
		4298	also extremely buggy). This makes select very inefficient, and also
		4299	requires a mapping from file descriptors to socket handles (the Microsoft
		4300	C runtime provides the function C<_open_osfhandle> for this). See the
		4301	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4302	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4303
		4304	The configuration for a "naked" win32 using the Microsoft runtime
		4305	libraries and raw winsocket select is:
		4306
		4307	#define EV_USE_SELECT 1
		4308	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4309
		4310	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4311	complexity in the O(n²) range when using win32.
		4312
		4313	=item Limited number of file descriptors
		4314
		4315	Windows has numerous arbitrary (and low) limits on things.
		4316
		4317	Early versions of winsocket's select only supported waiting for a maximum
		4318	of C<64> handles (probably owning to the fact that all windows kernels
		4319	can only wait for C<64> things at the same time internally; Microsoft
		4320	recommends spawning a chain of threads and wait for 63 handles and the
		4321	previous thread in each. Sounds great!).
		4322
		4323	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4324	to some high number (e.g. C<2048>) before compiling the winsocket select
		4325	call (which might be in libev or elsewhere, for example, perl and many
		4326	other interpreters do their own select emulation on windows).
		4327
		4328	Another limit is the number of file descriptors in the Microsoft runtime
		4329	libraries, which by default is C<64> (there must be a hidden I<64>
		4330	fetish or something like this inside Microsoft). You can increase this
		4331	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4332	(another arbitrary limit), but is broken in many versions of the Microsoft
		4333	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4334	(depending on windows version and/or the phase of the moon). To get more,
		4335	you need to wrap all I/O functions and provide your own fd management, but
		4336	the cost of calling select (O(n²)) will likely make this unworkable.
		4337
		4338	=back
		4339
		4340	=head2 PORTABILITY REQUIREMENTS
		4341
		4342	In addition to a working ISO-C implementation and of course the
		4343	backend-specific APIs, libev relies on a few additional extensions:
		4344
		4345	=over 4
		4346
		4347	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4348	calling conventions regardless of C<ev_watcher_type *>.
		4349
		4350	Libev assumes not only that all watcher pointers have the same internal
		4351	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4352	assumes that the same (machine) code can be used to call any watcher
		4353	callback: The watcher callbacks have different type signatures, but libev
		4354	calls them using an C<ev_watcher *> internally.
		4355
		4356	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4357
		4358	The type C<sig_atomic_t volatile> (or whatever is defined as
		4359	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4360	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4361	believed to be sufficiently portable.
		4362
		4363	=item C<sigprocmask> must work in a threaded environment
		4364
		4365	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4366	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4367	pthread implementations will either allow C<sigprocmask> in the "main
		4368	thread" or will block signals process-wide, both behaviours would
		4369	be compatible with libev. Interaction between C<sigprocmask> and
		4370	C<pthread_sigmask> could complicate things, however.
		4371
		4372	The most portable way to handle signals is to block signals in all threads
		4373	except the initial one, and run the default loop in the initial thread as
		4374	well.
		4375
		4376	=item C<long> must be large enough for common memory allocation sizes
		4377
		4378	To improve portability and simplify its API, libev uses C<long> internally
		4379	instead of C<size_t> when allocating its data structures. On non-POSIX
		4380	systems (Microsoft...) this might be unexpectedly low, but is still at
		4381	least 31 bits everywhere, which is enough for hundreds of millions of
		4382	watchers.
		4383
		4384	=item C<double> must hold a time value in seconds with enough accuracy
		4385
		4386	The type C<double> is used to represent timestamps. It is required to
		4387	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		4388	enough for at least into the year 4000. This requirement is fulfilled by
		4389	implementations implementing IEEE 754, which is basically all existing
		4390	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4391	2200.
		4392
		4393	=back
		4394
		4395	If you know of other additional requirements drop me a note.
		4396
		4397
		4398	=head1 ALGORITHMIC COMPLEXITIES
		4399
		4400	In this section the complexities of (many of) the algorithms used inside
		4401	libev will be documented. For complexity discussions about backends see
		4402	the documentation for C<ev_default_init>.
		4403
		4404	All of the following are about amortised time: If an array needs to be
		4405	extended, libev needs to realloc and move the whole array, but this
		4406	happens asymptotically rarer with higher number of elements, so O(1) might
		4407	mean that libev does a lengthy realloc operation in rare cases, but on
		4408	average it is much faster and asymptotically approaches constant time.
		4409
		4410	=over 4
		4411
		4412	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4413
		4414	This means that, when you have a watcher that triggers in one hour and
		4415	there are 100 watchers that would trigger before that, then inserting will
		4416	have to skip roughly seven (C<ld 100>) of these watchers.
		4417
		4418	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4419
		4420	That means that changing a timer costs less than removing/adding them,
		4421	as only the relative motion in the event queue has to be paid for.
		4422
		4423	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4424
		4425	These just add the watcher into an array or at the head of a list.
		4426
		4427	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4428
		4429	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4430
		4431	These watchers are stored in lists, so they need to be walked to find the
		4432	correct watcher to remove. The lists are usually short (you don't usually
		4433	have many watchers waiting for the same fd or signal: one is typical, two
		4434	is rare).
		4435
		4436	=item Finding the next timer in each loop iteration: O(1)
		4437
		4438	By virtue of using a binary or 4-heap, the next timer is always found at a
		4439	fixed position in the storage array.
		4440
		4441	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4442
		4443	A change means an I/O watcher gets started or stopped, which requires
		4444	libev to recalculate its status (and possibly tell the kernel, depending
		4445	on backend and whether C<ev_io_set> was used).
		4446
		4447	=item Activating one watcher (putting it into the pending state): O(1)
		4448
		4449	=item Priority handling: O(number_of_priorities)
		4450
		4451	Priorities are implemented by allocating some space for each
		4452	priority. When doing priority-based operations, libev usually has to
		4453	linearly search all the priorities, but starting/stopping and activating
		4454	watchers becomes O(1) with respect to priority handling.
		4455
		4456	=item Sending an ev_async: O(1)
		4457
		4458	=item Processing ev_async_send: O(number_of_async_watchers)
		4459
		4460	=item Processing signals: O(max_signal_number)
		4461
		4462	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4463	calls in the current loop iteration. Checking for async and signal events
		4464	involves iterating over all running async watchers or all signal numbers.
		4465
		4466	=back
		4467
		4468
		4469	=head1 GLOSSARY
		4470
		4471	=over 4
		4472
		4473	=item active
		4474
		4475	A watcher is active as long as it has been started (has been attached to
		4476	an event loop) but not yet stopped (disassociated from the event loop).
		4477
		4478	=item application
		4479
		4480	In this document, an application is whatever is using libev.
		4481
		4482	=item callback
		4483
		4484	The address of a function that is called when some event has been
		4485	detected. Callbacks are being passed the event loop, the watcher that
		4486	received the event, and the actual event bitset.
		4487
		4488	=item callback invocation
		4489
		4490	The act of calling the callback associated with a watcher.
		4491
		4492	=item event
		4493
		4494	A change of state of some external event, such as data now being available
		4495	for reading on a file descriptor, time having passed or simply not having
		4496	any other events happening anymore.
		4497
		4498	In libev, events are represented as single bits (such as C<EV_READ> or
		4499	C<EV_TIMEOUT>).
		4500
		4501	=item event library
		4502
		4503	A software package implementing an event model and loop.
		4504
		4505	=item event loop
		4506
		4507	An entity that handles and processes external events and converts them
		4508	into callback invocations.
		4509
		4510	=item event model
		4511
		4512	The model used to describe how an event loop handles and processes
		4513	watchers and events.
		4514
		4515	=item pending
		4516
		4517	A watcher is pending as soon as the corresponding event has been detected,
		4518	and stops being pending as soon as the watcher will be invoked or its
		4519	pending status is explicitly cleared by the application.
		4520
		4521	A watcher can be pending, but not active. Stopping a watcher also clears
		4522	its pending status.
		4523
		4524	=item real time
		4525
		4526	The physical time that is observed. It is apparently strictly monotonic :)
		4527
		4528	=item wall-clock time
		4529
		4530	The time and date as shown on clocks. Unlike real time, it can actually
		4531	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4532	clock.
		4533
		4534	=item watcher
		4535
		4536	A data structure that describes interest in certain events. Watchers need
		4537	to be started (attached to an event loop) before they can receive events.
		4538
		4539	=item watcher invocation
		4540
		4541	The act of calling the callback associated with a watcher.
		4542
		4543	=back
		4544
767	=head1 AUTHOR	4545	=head1 AUTHOR
768		4546
769	Marc Lehmann <libev@schmorp.de>.	4547	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
770		4548

Diff Legend

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