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 1.290 by root, Tue Mar 16 18:03:01 2010 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
9	=head1 DESCRIPTION	9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
		18	ev_io stdin_watcher;
		19	ev_timer timeout_watcher;
		20
		21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
		23	static void
		24	stdin_cb (EV_P_ ev_io *w, int revents)
		25	{
		26	puts ("stdin ready");
		27	// for one-shot events, one must manually stop the watcher
		28	// with its corresponding stop function.
		29	ev_io_stop (EV_A_ w);
		30
		31	// this causes all nested ev_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<struct 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_SIGNALFD>
		376
		377	When this flag is specified, then libev will attempt to use the
		378	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This API
		379	delivers signals synchronously, which makes it both faster and might make
		380	it possible to get the queued signal data. It can also simplify signal
		381	handling with threads, as long as you properly block signals in your
		382	threads that are not interested in handling them.
		383
		384	Signalfd will not be used by default as this changes your signal mask, and
		385	there are a lot of shoddy libraries and programs (glib's threadpool for
		386	example) that can't properly initialise their signal masks.
		387
143	=item C<EVMETHOD_SELECT> (portable select backend)	388	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		389
		390	This is your standard select(2) backend. Not I<completely> standard, as
		391	libev tries to roll its own fd_set with no limits on the number of fds,
		392	but if that fails, expect a fairly low limit on the number of fds when
		393	using this backend. It doesn't scale too well (O(highest_fd)), but its
		394	usually the fastest backend for a low number of (low-numbered :) fds.
		395
		396	To get good performance out of this backend you need a high amount of
		397	parallelism (most of the file descriptors should be busy). If you are
		398	writing a server, you should C<accept ()> in a loop to accept as many
		399	connections as possible during one iteration. You might also want to have
		400	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		401	readiness notifications you get per iteration.
		402
		403	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		404	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		405	C<exceptfds> set on that platform).
		406
145	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	407	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		408
147	=item C<EVMETHOD_EPOLL> (linux only)	409	And this is your standard poll(2) backend. It's more complicated
		410	than select, but handles sparse fds better and has no artificial
		411	limit on the number of fds you can use (except it will slow down
		412	considerably with a lot of inactive fds). It scales similarly to select,
		413	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		414	performance tips.
148		415
149	=item C<EVMETHOD_KQUEUE> (some bsds only)	416	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		417	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
150		418
151	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	419	=item C<EVBACKEND_EPOLL> (value 4, Linux)
152		420
153	=item C<EVMETHOD_PORT> (solaris 10 only)	421	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		422	kernels).
154		423
155	If one or more of these are ored into the flags value, then only these	424	For few fds, this backend is a bit little slower than poll and select,
156	backends will be tried (in the reverse order as given here). If one are	425	but it scales phenomenally better. While poll and select usually scale
157	specified, any backend will do.	426	like O(total_fds) where n is the total number of fds (or the highest fd),
		427	epoll scales either O(1) or O(active_fds).
		428
		429	The epoll mechanism deserves honorable mention as the most misdesigned
		430	of the more advanced event mechanisms: mere annoyances include silently
		431	dropping file descriptors, requiring a system call per change per file
		432	descriptor (and unnecessary guessing of parameters), problems with dup and
		433	so on. The biggest issue is fork races, however - if a program forks then
		434	I<both> parent and child process have to recreate the epoll set, which can
		435	take considerable time (one syscall per file descriptor) and is of course
		436	hard to detect.
		437
		438	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		439	of course I<doesn't>, and epoll just loves to report events for totally
		440	I<different> file descriptors (even already closed ones, so one cannot
		441	even remove them from the set) than registered in the set (especially
		442	on SMP systems). Libev tries to counter these spurious notifications by
		443	employing an additional generation counter and comparing that against the
		444	events to filter out spurious ones, recreating the set when required.
		445
		446	While stopping, setting and starting an I/O watcher in the same iteration
		447	will result in some caching, there is still a system call per such
		448	incident (because the same I<file descriptor> could point to a different
		449	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		450	file descriptors might not work very well if you register events for both
		451	file descriptors.
		452
		453	Best performance from this backend is achieved by not unregistering all
		454	watchers for a file descriptor until it has been closed, if possible,
		455	i.e. keep at least one watcher active per fd at all times. Stopping and
		456	starting a watcher (without re-setting it) also usually doesn't cause
		457	extra overhead. A fork can both result in spurious notifications as well
		458	as in libev having to destroy and recreate the epoll object, which can
		459	take considerable time and thus should be avoided.
		460
		461	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		462	faster than epoll for maybe up to a hundred file descriptors, depending on
		463	the usage. So sad.
		464
		465	While nominally embeddable in other event loops, this feature is broken in
		466	all kernel versions tested so far.
		467
		468	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		469	C<EVBACKEND_POLL>.
		470
		471	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		472
		473	Kqueue deserves special mention, as at the time of this writing, it
		474	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		475	with anything but sockets and pipes, except on Darwin, where of course
		476	it's completely useless). Unlike epoll, however, whose brokenness
		477	is by design, these kqueue bugs can (and eventually will) be fixed
		478	without API changes to existing programs. For this reason it's not being
		479	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		480	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		481	system like NetBSD.
		482
		483	You still can embed kqueue into a normal poll or select backend and use it
		484	only for sockets (after having made sure that sockets work with kqueue on
		485	the target platform). See C<ev_embed> watchers for more info.
		486
		487	It scales in the same way as the epoll backend, but the interface to the
		488	kernel is more efficient (which says nothing about its actual speed, of
		489	course). While stopping, setting and starting an I/O watcher does never
		490	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		491	two event changes per incident. Support for C<fork ()> is very bad (but
		492	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		493	cases
		494
		495	This backend usually performs well under most conditions.
		496
		497	While nominally embeddable in other event loops, this doesn't work
		498	everywhere, so you might need to test for this. And since it is broken
		499	almost everywhere, you should only use it when you have a lot of sockets
		500	(for which it usually works), by embedding it into another event loop
		501	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		502	also broken on OS X)) and, did I mention it, using it only for sockets.
		503
		504	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		505	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		506	C<NOTE_EOF>.
		507
		508	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		509
		510	This is not implemented yet (and might never be, unless you send me an
		511	implementation). According to reports, C</dev/poll> only supports sockets
		512	and is not embeddable, which would limit the usefulness of this backend
		513	immensely.
		514
		515	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		516
		517	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		518	it's really slow, but it still scales very well (O(active_fds)).
		519
		520	Please note that Solaris event ports can deliver a lot of spurious
		521	notifications, so you need to use non-blocking I/O or other means to avoid
		522	blocking when no data (or space) is available.
		523
		524	While this backend scales well, it requires one system call per active
		525	file descriptor per loop iteration. For small and medium numbers of file
		526	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		527	might perform better.
		528
		529	On the positive side, with the exception of the spurious readiness
		530	notifications, this backend actually performed fully to specification
		531	in all tests and is fully embeddable, which is a rare feat among the
		532	OS-specific backends (I vastly prefer correctness over speed hacks).
		533
		534	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		535	C<EVBACKEND_POLL>.
		536
		537	=item C<EVBACKEND_ALL>
		538
		539	Try all backends (even potentially broken ones that wouldn't be tried
		540	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		541	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		542
		543	It is definitely not recommended to use this flag.
158		544
159	=back	545	=back
160		546
		547	If one or more of the backend flags are or'ed into the flags value,
		548	then only these backends will be tried (in the reverse order as listed
		549	here). If none are specified, all backends in C<ev_recommended_backends
		550	()> will be tried.
		551
		552	Example: This is the most typical usage.
		553
		554	if (!ev_default_loop (0))
		555	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		556
		557	Example: Restrict libev to the select and poll backends, and do not allow
		558	environment settings to be taken into account:
		559
		560	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		561
		562	Example: Use whatever libev has to offer, but make sure that kqueue is
		563	used if available (warning, breaks stuff, best use only with your own
		564	private event loop and only if you know the OS supports your types of
		565	fds):
		566
		567	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
		568
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	569	=item struct ev_loop *ev_loop_new (unsigned int flags)
162		570
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	571	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	572	always distinct from the default loop.
165	handle signal and child watchers, and attempts to do so will be greeted by	573
166	undefined behaviour (or a failed assertion if assertions are enabled).	574	Note that this function I<is> thread-safe, and one common way to use
		575	libev with threads is indeed to create one loop per thread, and using the
		576	default loop in the "main" or "initial" thread.
		577
		578	Example: Try to create a event loop that uses epoll and nothing else.
		579
		580	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		581	if (!epoller)
		582	fatal ("no epoll found here, maybe it hides under your chair");
167		583
168	=item ev_default_destroy ()	584	=item ev_default_destroy ()
169		585
170	Destroys the default loop again (frees all memory and kernel state	586	Destroys the default loop (frees all memory and kernel state etc.). None
171	etc.). This stops all registered event watchers (by not touching them in	587	of the active event watchers will be stopped in the normal sense, so
172	any way whatsoever, although you cannot rely on this :).	588	e.g. C<ev_is_active> might still return true. It is your responsibility to
		589	either stop all watchers cleanly yourself I<before> calling this function,
		590	or cope with the fact afterwards (which is usually the easiest thing, you
		591	can just ignore the watchers and/or C<free ()> them for example).
		592
		593	Note that certain global state, such as signal state (and installed signal
		594	handlers), will not be freed by this function, and related watchers (such
		595	as signal and child watchers) would need to be stopped manually.
		596
		597	In general it is not advisable to call this function except in the
		598	rare occasion where you really need to free e.g. the signal handling
		599	pipe fds. If you need dynamically allocated loops it is better to use
		600	C<ev_loop_new> and C<ev_loop_destroy>.
173		601
174	=item ev_loop_destroy (loop)	602	=item ev_loop_destroy (loop)
175		603
176	Like C<ev_default_destroy>, but destroys an event loop created by an	604	Like C<ev_default_destroy>, but destroys an event loop created by an
177	earlier call to C<ev_loop_new>.	605	earlier call to C<ev_loop_new>.
178		606
179	=item ev_default_fork ()	607	=item ev_default_fork ()
180		608
		609	This function sets a flag that causes subsequent C<ev_loop> iterations
181	This function reinitialises the kernel state for backends that have	610	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	611	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	612	the child process (or both child and parent, but that again makes little
184	again makes little sense).	613	sense). You I<must> call it in the child before using any of the libev
		614	functions, and it will only take effect at the next C<ev_loop> iteration.
185		615
186	You I<must> call this function after forking if and only if you want to	616	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	617	process if and only if you want to use the event library in the child. If
188	have to call it.	618	you just fork+exec, you don't have to call it at all.
189		619
190	The function itself is quite fast and it's usually not a problem to call	620	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	621	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>:	622	quite nicely into a call to C<pthread_atfork>:
193		623
…		…
195		625
196	=item ev_loop_fork (loop)	626	=item ev_loop_fork (loop)
197		627
198	Like C<ev_default_fork>, but acts on an event loop created by	628	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	629	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.	630	after fork that you want to re-use in the child, and how you do this is
		631	entirely your own problem.
201		632
		633	=item int ev_is_default_loop (loop)
		634
		635	Returns true when the given loop is, in fact, the default loop, and false
		636	otherwise.
		637
		638	=item unsigned int ev_loop_count (loop)
		639
		640	Returns the count of loop iterations for the loop, which is identical to
		641	the number of times libev did poll for new events. It starts at C<0> and
		642	happily wraps around with enough iterations.
		643
		644	This value can sometimes be useful as a generation counter of sorts (it
		645	"ticks" the number of loop iterations), as it roughly corresponds with
		646	C<ev_prepare> and C<ev_check> calls.
		647
		648	=item unsigned int ev_loop_depth (loop)
		649
		650	Returns the number of times C<ev_loop> was entered minus the number of
		651	times C<ev_loop> was exited, in other words, the recursion depth.
		652
		653	Outside C<ev_loop>, this number is zero. In a callback, this number is
		654	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		655	in which case it is higher.
		656
		657	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		658	etc.), doesn't count as exit.
		659
202	=item unsigned int ev_method (loop)	660	=item unsigned int ev_backend (loop)
203		661
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	662	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	663	use.
206		664
207	=item ev_tstamp ev_now (loop)	665	=item ev_tstamp ev_now (loop)
208		666
209	Returns the current "event loop time", which is the time the event loop	667	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	668	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	669	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	670	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).	671	event occurring (or more correctly, libev finding out about it).
		672
		673	=item ev_now_update (loop)
		674
		675	Establishes the current time by querying the kernel, updating the time
		676	returned by C<ev_now ()> in the progress. This is a costly operation and
		677	is usually done automatically within C<ev_loop ()>.
		678
		679	This function is rarely useful, but when some event callback runs for a
		680	very long time without entering the event loop, updating libev's idea of
		681	the current time is a good idea.
		682
		683	See also L<The special problem of time updates> in the C<ev_timer> section.
		684
		685	=item ev_suspend (loop)
		686
		687	=item ev_resume (loop)
		688
		689	These two functions suspend and resume a loop, for use when the loop is
		690	not used for a while and timeouts should not be processed.
		691
		692	A typical use case would be an interactive program such as a game: When
		693	the user presses C<^Z> to suspend the game and resumes it an hour later it
		694	would be best to handle timeouts as if no time had actually passed while
		695	the program was suspended. This can be achieved by calling C<ev_suspend>
		696	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		697	C<ev_resume> directly afterwards to resume timer processing.
		698
		699	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		700	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		701	will be rescheduled (that is, they will lose any events that would have
		702	occured while suspended).
		703
		704	After calling C<ev_suspend> you B<must not> call I<any> function on the
		705	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		706	without a previous call to C<ev_suspend>.
		707
		708	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		709	event loop time (see C<ev_now_update>).
214		710
215	=item ev_loop (loop, int flags)	711	=item ev_loop (loop, int flags)
216		712
217	Finally, this is it, the event handler. This function usually is called	713	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	714	after you have initialised all your watchers and you want to start
219	events.	715	handling events.
220		716
221	If the flags argument is specified as 0, it will not return until either	717	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.	718	either no event watchers are active anymore or C<ev_unloop> was called.
		719
		720	Please note that an explicit C<ev_unloop> is usually better than
		721	relying on all watchers to be stopped when deciding when a program has
		722	finished (especially in interactive programs), but having a program
		723	that automatically loops as long as it has to and no longer by virtue
		724	of relying on its watchers stopping correctly, that is truly a thing of
		725	beauty.
223		726
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	727	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	728	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.	729	process in case there are no events and will return after one iteration of
		730	the loop.
227		731
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	732	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	733	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	734	will block your process until at least one new event arrives (which could
		735	be an event internal to libev itself, so there is no guarantee that a
		736	user-registered callback will be called), and will return after one
231	one iteration of the loop.	737	iteration of the loop.
232		738
233	This flags value could be used to implement alternative looping	739	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	740	with something not expressible using other libev watchers (i.e. "roll your
235	more generic mechanism.	741	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		742	usually a better approach for this kind of thing.
		743
		744	Here are the gory details of what C<ev_loop> does:
		745
		746	- Before the first iteration, call any pending watchers.
		747	* If EVFLAG_FORKCHECK was used, check for a fork.
		748	- If a fork was detected (by any means), queue and call all fork watchers.
		749	- Queue and call all prepare watchers.
		750	- If we have been forked, detach and recreate the kernel state
		751	as to not disturb the other process.
		752	- Update the kernel state with all outstanding changes.
		753	- Update the "event loop time" (ev_now ()).
		754	- Calculate for how long to sleep or block, if at all
		755	(active idle watchers, EVLOOP_NONBLOCK or not having
		756	any active watchers at all will result in not sleeping).
		757	- Sleep if the I/O and timer collect interval say so.
		758	- Block the process, waiting for any events.
		759	- Queue all outstanding I/O (fd) events.
		760	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		761	- Queue all expired timers.
		762	- Queue all expired periodics.
		763	- Unless any events are pending now, queue all idle watchers.
		764	- Queue all check watchers.
		765	- Call all queued watchers in reverse order (i.e. check watchers first).
		766	Signals and child watchers are implemented as I/O watchers, and will
		767	be handled here by queueing them when their watcher gets executed.
		768	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		769	were used, or there are no active watchers, return, otherwise
		770	continue with step *.
		771
		772	Example: Queue some jobs and then loop until no events are outstanding
		773	anymore.
		774
		775	... queue jobs here, make sure they register event watchers as long
		776	... as they still have work to do (even an idle watcher will do..)
		777	ev_loop (my_loop, 0);
		778	... jobs done or somebody called unloop. yeah!
236		779
237	=item ev_unloop (loop, how)	780	=item ev_unloop (loop, how)
238		781
239	Can be used to make a call to C<ev_loop> return early (but only after it	782	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	783	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	784	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.	785	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
		786
		787	This "unloop state" will be cleared when entering C<ev_loop> again.
		788
		789	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
243		790
244	=item ev_ref (loop)	791	=item ev_ref (loop)
245		792
246	=item ev_unref (loop)	793	=item ev_unref (loop)
247		794
248	Ref/unref can be used to add or remove a reference count on the event	795	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	796	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	797	count is nonzero, C<ev_loop> will not return on its own.
251	a watcher you never unregister that should not keep C<ev_loop> from	798
252	returning, ev_unref() after starting, and ev_ref() before stopping it. For	799	This is useful when you have a watcher that you never intend to
		800	unregister, but that nevertheless should not keep C<ev_loop> from
		801	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		802	before stopping it.
		803
253	example, libev itself uses this for its internal signal pipe: It is not	804	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	805	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	806	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	807	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>.	808	third-party libraries. Just remember to I<unref after start> and I<ref
		809	before stop> (but only if the watcher wasn't active before, or was active
		810	before, respectively. Note also that libev might stop watchers itself
		811	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		812	in the callback).
		813
		814	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		815	running when nothing else is active.
		816
		817	ev_signal exitsig;
		818	ev_signal_init (&exitsig, sig_cb, SIGINT);
		819	ev_signal_start (loop, &exitsig);
		820	evf_unref (loop);
		821
		822	Example: For some weird reason, unregister the above signal handler again.
		823
		824	ev_ref (loop);
		825	ev_signal_stop (loop, &exitsig);
		826
		827	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		828
		829	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		830
		831	These advanced functions influence the time that libev will spend waiting
		832	for events. Both time intervals are by default C<0>, meaning that libev
		833	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		834	latency.
		835
		836	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		837	allows libev to delay invocation of I/O and timer/periodic callbacks
		838	to increase efficiency of loop iterations (or to increase power-saving
		839	opportunities).
		840
		841	The idea is that sometimes your program runs just fast enough to handle
		842	one (or very few) event(s) per loop iteration. While this makes the
		843	program responsive, it also wastes a lot of CPU time to poll for new
		844	events, especially with backends like C<select ()> which have a high
		845	overhead for the actual polling but can deliver many events at once.
		846
		847	By setting a higher I<io collect interval> you allow libev to spend more
		848	time collecting I/O events, so you can handle more events per iteration,
		849	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		850	C<ev_timer>) will be not affected. Setting this to a non-null value will
		851	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		852	sleep time ensures that libev will not poll for I/O events more often then
		853	once per this interval, on average.
		854
		855	Likewise, by setting a higher I<timeout collect interval> you allow libev
		856	to spend more time collecting timeouts, at the expense of increased
		857	latency/jitter/inexactness (the watcher callback will be called
		858	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		859	value will not introduce any overhead in libev.
		860
		861	Many (busy) programs can usually benefit by setting the I/O collect
		862	interval to a value near C<0.1> or so, which is often enough for
		863	interactive servers (of course not for games), likewise for timeouts. It
		864	usually doesn't make much sense to set it to a lower value than C<0.01>,
		865	as this approaches the timing granularity of most systems. Note that if
		866	you do transactions with the outside world and you can't increase the
		867	parallelity, then this setting will limit your transaction rate (if you
		868	need to poll once per transaction and the I/O collect interval is 0.01,
		869	then you can't do more than 100 transations per second).
		870
		871	Setting the I<timeout collect interval> can improve the opportunity for
		872	saving power, as the program will "bundle" timer callback invocations that
		873	are "near" in time together, by delaying some, thus reducing the number of
		874	times the process sleeps and wakes up again. Another useful technique to
		875	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		876	they fire on, say, one-second boundaries only.
		877
		878	Example: we only need 0.1s timeout granularity, and we wish not to poll
		879	more often than 100 times per second:
		880
		881	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		882	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		883
		884	=item ev_invoke_pending (loop)
		885
		886	This call will simply invoke all pending watchers while resetting their
		887	pending state. Normally, C<ev_loop> does this automatically when required,
		888	but when overriding the invoke callback this call comes handy.
		889
		890	=item int ev_pending_count (loop)
		891
		892	Returns the number of pending watchers - zero indicates that no watchers
		893	are pending.
		894
		895	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		896
		897	This overrides the invoke pending functionality of the loop: Instead of
		898	invoking all pending watchers when there are any, C<ev_loop> will call
		899	this callback instead. This is useful, for example, when you want to
		900	invoke the actual watchers inside another context (another thread etc.).
		901
		902	If you want to reset the callback, use C<ev_invoke_pending> as new
		903	callback.
		904
		905	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		906
		907	Sometimes you want to share the same loop between multiple threads. This
		908	can be done relatively simply by putting mutex_lock/unlock calls around
		909	each call to a libev function.
		910
		911	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		912	wait for it to return. One way around this is to wake up the loop via
		913	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		914	and I<acquire> callbacks on the loop.
		915
		916	When set, then C<release> will be called just before the thread is
		917	suspended waiting for new events, and C<acquire> is called just
		918	afterwards.
		919
		920	Ideally, C<release> will just call your mutex_unlock function, and
		921	C<acquire> will just call the mutex_lock function again.
		922
		923	While event loop modifications are allowed between invocations of
		924	C<release> and C<acquire> (that's their only purpose after all), no
		925	modifications done will affect the event loop, i.e. adding watchers will
		926	have no effect on the set of file descriptors being watched, or the time
		927	waited. Use an C<ev_async> watcher to wake up C<ev_loop> when you want it
		928	to take note of any changes you made.
		929
		930	In theory, threads executing C<ev_loop> will be async-cancel safe between
		931	invocations of C<release> and C<acquire>.
		932
		933	See also the locking example in the C<THREADS> section later in this
		934	document.
		935
		936	=item ev_set_userdata (loop, void *data)
		937
		938	=item ev_userdata (loop)
		939
		940	Set and retrieve a single C<void *> associated with a loop. When
		941	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		942	C<0.>
		943
		944	These two functions can be used to associate arbitrary data with a loop,
		945	and are intended solely for the C<invoke_pending_cb>, C<release> and
		946	C<acquire> callbacks described above, but of course can be (ab-)used for
		947	any other purpose as well.
		948
		949	=item ev_loop_verify (loop)
		950
		951	This function only does something when C<EV_VERIFY> support has been
		952	compiled in, which is the default for non-minimal builds. It tries to go
		953	through all internal structures and checks them for validity. If anything
		954	is found to be inconsistent, it will print an error message to standard
		955	error and call C<abort ()>.
		956
		957	This can be used to catch bugs inside libev itself: under normal
		958	circumstances, this function will never abort as of course libev keeps its
		959	data structures consistent.
258		960
259	=back	961	=back
260		962
		963
261	=head1 ANATOMY OF A WATCHER	964	=head1 ANATOMY OF A WATCHER
		965
		966	In the following description, uppercase C<TYPE> in names stands for the
		967	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		968	watchers and C<ev_io_start> for I/O watchers.
262		969
263	A watcher is a structure that you create and register to record your	970	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	971	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:	972	become readable, you would create an C<ev_io> watcher for that:
266		973
267	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	974	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
268	{	975	{
269	ev_io_stop (w);	976	ev_io_stop (w);
270	ev_unloop (loop, EVUNLOOP_ALL);	977	ev_unloop (loop, EVUNLOOP_ALL);
271	}	978	}
272		979
273	struct ev_loop *loop = ev_default_loop (0);	980	struct ev_loop *loop = ev_default_loop (0);
		981
274	struct ev_io stdin_watcher;	982	ev_io stdin_watcher;
		983
275	ev_init (&stdin_watcher, my_cb);	984	ev_init (&stdin_watcher, my_cb);
276	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	985	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
277	ev_io_start (loop, &stdin_watcher);	986	ev_io_start (loop, &stdin_watcher);
		987
278	ev_loop (loop, 0);	988	ev_loop (loop, 0);
279		989
280	As you can see, you are responsible for allocating the memory for your	990	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,	991	watcher structures (and it is I<usually> a bad idea to do this on the
282	although this can sometimes be quite valid).	992	stack).
		993
		994	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		995	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
283		996
284	Each watcher structure must be initialised by a call to C<ev_init	997	Each watcher structure must be initialised by a call to C<ev_init
285	(watcher *, callback)>, which expects a callback to be provided. This	998	(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	999	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	1000	watchers, each time the event loop detects that the file descriptor given
288	is readable and/or writable).	1001	is readable and/or writable).
289		1002
290	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1003	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	1004	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	1005	is also a macro to combine initialisation and setting in one call: C<<
293	(watcher *, callback, ...) >>.	1006	ev_TYPE_init (watcher *, callback, ...) >>.
294		1007
295	To make the watcher actually watch out for events, you have to start it	1008	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	1009	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	1010	*) >>), and you can stop watching for events at any time by calling the
298	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1011	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
299		1012
300	As long as your watcher is active (has been started but not stopped) you	1013	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	1014	must not touch the values stored in it. Most specifically you must never
302	reinitialise it or call its set method.	1015	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		1016
309	Each and every callback receives the event loop pointer as first, the	1017	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	1018	registered watcher structure as second, and a bitset of received events as
311	third argument.	1019	third argument.
312		1020
…		…
321	=item C<EV_WRITE>	1029	=item C<EV_WRITE>
322		1030
323	The file descriptor in the C<ev_io> watcher has become readable and/or	1031	The file descriptor in the C<ev_io> watcher has become readable and/or
324	writable.	1032	writable.
325		1033
326	=item C<EV_TIMEOUT>	1034	=item C<EV_TIMER>
327		1035
328	The C<ev_timer> watcher has timed out.	1036	The C<ev_timer> watcher has timed out.
329		1037
330	=item C<EV_PERIODIC>	1038	=item C<EV_PERIODIC>
331		1039
…		…
336	The signal specified in the C<ev_signal> watcher has been received by a thread.	1044	The signal specified in the C<ev_signal> watcher has been received by a thread.
337		1045
338	=item C<EV_CHILD>	1046	=item C<EV_CHILD>
339		1047
340	The pid specified in the C<ev_child> watcher has received a status change.	1048	The pid specified in the C<ev_child> watcher has received a status change.
		1049
		1050	=item C<EV_STAT>
		1051
		1052	The path specified in the C<ev_stat> watcher changed its attributes somehow.
341		1053
342	=item C<EV_IDLE>	1054	=item C<EV_IDLE>
343		1055
344	The C<ev_idle> watcher has determined that you have nothing better to do.	1056	The C<ev_idle> watcher has determined that you have nothing better to do.
345		1057
…		…
353	received events. Callbacks of both watcher types can start and stop as	1065	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	1066	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	1067	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
356	C<ev_loop> from blocking).	1068	C<ev_loop> from blocking).
357		1069
		1070	=item C<EV_EMBED>
		1071
		1072	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1073
		1074	=item C<EV_FORK>
		1075
		1076	The event loop has been resumed in the child process after fork (see
		1077	C<ev_fork>).
		1078
		1079	=item C<EV_ASYNC>
		1080
		1081	The given async watcher has been asynchronously notified (see C<ev_async>).
		1082
		1083	=item C<EV_CUSTOM>
		1084
		1085	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1086	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1087
358	=item C<EV_ERROR>	1088	=item C<EV_ERROR>
359		1089
360	An unspecified error has occured, the watcher has been stopped. This might	1090	An unspecified error has occurred, the watcher has been stopped. This might
361	happen because the watcher could not be properly started because libev	1091	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	1092	ran out of memory, a file descriptor was found to be closed or any other
		1093	problem. Libev considers these application bugs.
		1094
363	problem. You best act on it by reporting the problem and somehow coping	1095	You best act on it by reporting the problem and somehow coping with the
364	with the watcher being stopped.	1096	watcher being stopped. Note that well-written programs should not receive
		1097	an error ever, so when your watcher receives it, this usually indicates a
		1098	bug in your program.
365		1099
366	Libev will usually signal a few "dummy" events together with an error,	1100	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	1101	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	1102	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	1103	the error from read() or write(). This will not work in multi-threaded
370	programs, though, so beware.	1104	programs, though, as the fd could already be closed and reused for another
		1105	thing, so beware.
371		1106
372	=back	1107	=back
373		1108
		1109	=head2 GENERIC WATCHER FUNCTIONS
		1110
		1111	=over 4
		1112
		1113	=item C<ev_init> (ev_TYPE *watcher, callback)
		1114
		1115	This macro initialises the generic portion of a watcher. The contents
		1116	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1117	the generic parts of the watcher are initialised, you I<need> to call
		1118	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1119	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1120	which rolls both calls into one.
		1121
		1122	You can reinitialise a watcher at any time as long as it has been stopped
		1123	(or never started) and there are no pending events outstanding.
		1124
		1125	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1126	int revents)>.
		1127
		1128	Example: Initialise an C<ev_io> watcher in two steps.
		1129
		1130	ev_io w;
		1131	ev_init (&w, my_cb);
		1132	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1133
		1134	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1135
		1136	This macro initialises the type-specific parts of a watcher. You need to
		1137	call C<ev_init> at least once before you call this macro, but you can
		1138	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1139	macro on a watcher that is active (it can be pending, however, which is a
		1140	difference to the C<ev_init> macro).
		1141
		1142	Although some watcher types do not have type-specific arguments
		1143	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1144
		1145	See C<ev_init>, above, for an example.
		1146
		1147	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1148
		1149	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1150	calls into a single call. This is the most convenient method to initialise
		1151	a watcher. The same limitations apply, of course.
		1152
		1153	Example: Initialise and set an C<ev_io> watcher in one step.
		1154
		1155	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1156
		1157	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1158
		1159	Starts (activates) the given watcher. Only active watchers will receive
		1160	events. If the watcher is already active nothing will happen.
		1161
		1162	Example: Start the C<ev_io> watcher that is being abused as example in this
		1163	whole section.
		1164
		1165	ev_io_start (EV_DEFAULT_UC, &w);
		1166
		1167	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1168
		1169	Stops the given watcher if active, and clears the pending status (whether
		1170	the watcher was active or not).
		1171
		1172	It is possible that stopped watchers are pending - for example,
		1173	non-repeating timers are being stopped when they become pending - but
		1174	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1175	pending. If you want to free or reuse the memory used by the watcher it is
		1176	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1177
		1178	=item bool ev_is_active (ev_TYPE *watcher)
		1179
		1180	Returns a true value iff the watcher is active (i.e. it has been started
		1181	and not yet been stopped). As long as a watcher is active you must not modify
		1182	it.
		1183
		1184	=item bool ev_is_pending (ev_TYPE *watcher)
		1185
		1186	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1187	events but its callback has not yet been invoked). As long as a watcher
		1188	is pending (but not active) you must not call an init function on it (but
		1189	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1190	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1191	it).
		1192
		1193	=item callback ev_cb (ev_TYPE *watcher)
		1194
		1195	Returns the callback currently set on the watcher.
		1196
		1197	=item ev_cb_set (ev_TYPE *watcher, callback)
		1198
		1199	Change the callback. You can change the callback at virtually any time
		1200	(modulo threads).
		1201
		1202	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1203
		1204	=item int ev_priority (ev_TYPE *watcher)
		1205
		1206	Set and query the priority of the watcher. The priority is a small
		1207	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1208	(default: C<-2>). Pending watchers with higher priority will be invoked
		1209	before watchers with lower priority, but priority will not keep watchers
		1210	from being executed (except for C<ev_idle> watchers).
		1211
		1212	If you need to suppress invocation when higher priority events are pending
		1213	you need to look at C<ev_idle> watchers, which provide this functionality.
		1214
		1215	You I<must not> change the priority of a watcher as long as it is active or
		1216	pending.
		1217
		1218	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1219	fine, as long as you do not mind that the priority value you query might
		1220	or might not have been clamped to the valid range.
		1221
		1222	The default priority used by watchers when no priority has been set is
		1223	always C<0>, which is supposed to not be too high and not be too low :).
		1224
		1225	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1226	priorities.
		1227
		1228	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1229
		1230	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1231	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1232	can deal with that fact, as both are simply passed through to the
		1233	callback.
		1234
		1235	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1236
		1237	If the watcher is pending, this function clears its pending status and
		1238	returns its C<revents> bitset (as if its callback was invoked). If the
		1239	watcher isn't pending it does nothing and returns C<0>.
		1240
		1241	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1242	callback to be invoked, which can be accomplished with this function.
		1243
		1244	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1245
		1246	Feeds the given event set into the event loop, as if the specified event
		1247	had happened for the specified watcher (which must be a pointer to an
		1248	initialised but not necessarily started event watcher). Obviously you must
		1249	not free the watcher as long as it has pending events.
		1250
		1251	Stopping the watcher, letting libev invoke it, or calling
		1252	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1253	not started in the first place.
		1254
		1255	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1256	functions that do not need a watcher.
		1257
		1258	=back
		1259
		1260
374	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1261	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375		1262
376	Each watcher has, by default, a member C<void *data> that you can change	1263	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	1264	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	1265	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	1266	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	1267	member, you can also "subclass" the watcher type and provide your own
381	data:	1268	data:
382		1269
383	struct my_io	1270	struct my_io
384	{	1271	{
385	struct ev_io io;	1272	ev_io io;
386	int otherfd;	1273	int otherfd;
387	void *somedata;	1274	void *somedata;
388	struct whatever *mostinteresting;	1275	struct whatever *mostinteresting;
389	}	1276	};
		1277
		1278	...
		1279	struct my_io w;
		1280	ev_io_init (&w.io, my_cb, fd, EV_READ);
390		1281
391	And since your callback will be called with a pointer to the watcher, you	1282	And since your callback will be called with a pointer to the watcher, you
392	can cast it back to your own type:	1283	can cast it back to your own type:
393		1284
394	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1285	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
395	{	1286	{
396	struct my_io *w = (struct my_io *)w_;	1287	struct my_io *w = (struct my_io *)w_;
397	...	1288	...
398	}	1289	}
399		1290
400	More interesting and less C-conformant ways of catsing your callback type	1291	More interesting and less C-conformant ways of casting your callback type
401	have been omitted....	1292	instead have been omitted.
		1293
		1294	Another common scenario is to use some data structure with multiple
		1295	embedded watchers:
		1296
		1297	struct my_biggy
		1298	{
		1299	int some_data;
		1300	ev_timer t1;
		1301	ev_timer t2;
		1302	}
		1303
		1304	In this case getting the pointer to C<my_biggy> is a bit more
		1305	complicated: Either you store the address of your C<my_biggy> struct
		1306	in the C<data> member of the watcher (for woozies), or you need to use
		1307	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1308	programmers):
		1309
		1310	#include <stddef.h>
		1311
		1312	static void
		1313	t1_cb (EV_P_ ev_timer *w, int revents)
		1314	{
		1315	struct my_biggy big = (struct my_biggy *)
		1316	(((char *)w) - offsetof (struct my_biggy, t1));
		1317	}
		1318
		1319	static void
		1320	t2_cb (EV_P_ ev_timer *w, int revents)
		1321	{
		1322	struct my_biggy big = (struct my_biggy *)
		1323	(((char *)w) - offsetof (struct my_biggy, t2));
		1324	}
		1325
		1326	=head2 WATCHER PRIORITY MODELS
		1327
		1328	Many event loops support I<watcher priorities>, which are usually small
		1329	integers that influence the ordering of event callback invocation
		1330	between watchers in some way, all else being equal.
		1331
		1332	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1333	description for the more technical details such as the actual priority
		1334	range.
		1335
		1336	There are two common ways how these these priorities are being interpreted
		1337	by event loops:
		1338
		1339	In the more common lock-out model, higher priorities "lock out" invocation
		1340	of lower priority watchers, which means as long as higher priority
		1341	watchers receive events, lower priority watchers are not being invoked.
		1342
		1343	The less common only-for-ordering model uses priorities solely to order
		1344	callback invocation within a single event loop iteration: Higher priority
		1345	watchers are invoked before lower priority ones, but they all get invoked
		1346	before polling for new events.
		1347
		1348	Libev uses the second (only-for-ordering) model for all its watchers
		1349	except for idle watchers (which use the lock-out model).
		1350
		1351	The rationale behind this is that implementing the lock-out model for
		1352	watchers is not well supported by most kernel interfaces, and most event
		1353	libraries will just poll for the same events again and again as long as
		1354	their callbacks have not been executed, which is very inefficient in the
		1355	common case of one high-priority watcher locking out a mass of lower
		1356	priority ones.
		1357
		1358	Static (ordering) priorities are most useful when you have two or more
		1359	watchers handling the same resource: a typical usage example is having an
		1360	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1361	timeouts. Under load, data might be received while the program handles
		1362	other jobs, but since timers normally get invoked first, the timeout
		1363	handler will be executed before checking for data. In that case, giving
		1364	the timer a lower priority than the I/O watcher ensures that I/O will be
		1365	handled first even under adverse conditions (which is usually, but not
		1366	always, what you want).
		1367
		1368	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1369	will only be executed when no same or higher priority watchers have
		1370	received events, they can be used to implement the "lock-out" model when
		1371	required.
		1372
		1373	For example, to emulate how many other event libraries handle priorities,
		1374	you can associate an C<ev_idle> watcher to each such watcher, and in
		1375	the normal watcher callback, you just start the idle watcher. The real
		1376	processing is done in the idle watcher callback. This causes libev to
		1377	continously poll and process kernel event data for the watcher, but when
		1378	the lock-out case is known to be rare (which in turn is rare :), this is
		1379	workable.
		1380
		1381	Usually, however, the lock-out model implemented that way will perform
		1382	miserably under the type of load it was designed to handle. In that case,
		1383	it might be preferable to stop the real watcher before starting the
		1384	idle watcher, so the kernel will not have to process the event in case
		1385	the actual processing will be delayed for considerable time.
		1386
		1387	Here is an example of an I/O watcher that should run at a strictly lower
		1388	priority than the default, and which should only process data when no
		1389	other events are pending:
		1390
		1391	ev_idle idle; // actual processing watcher
		1392	ev_io io; // actual event watcher
		1393
		1394	static void
		1395	io_cb (EV_P_ ev_io *w, int revents)
		1396	{
		1397	// stop the I/O watcher, we received the event, but
		1398	// are not yet ready to handle it.
		1399	ev_io_stop (EV_A_ w);
		1400
		1401	// start the idle watcher to ahndle the actual event.
		1402	// it will not be executed as long as other watchers
		1403	// with the default priority are receiving events.
		1404	ev_idle_start (EV_A_ &idle);
		1405	}
		1406
		1407	static void
		1408	idle_cb (EV_P_ ev_idle *w, int revents)
		1409	{
		1410	// actual processing
		1411	read (STDIN_FILENO, ...);
		1412
		1413	// have to start the I/O watcher again, as
		1414	// we have handled the event
		1415	ev_io_start (EV_P_ &io);
		1416	}
		1417
		1418	// initialisation
		1419	ev_idle_init (&idle, idle_cb);
		1420	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1421	ev_io_start (EV_DEFAULT_ &io);
		1422
		1423	In the "real" world, it might also be beneficial to start a timer, so that
		1424	low-priority connections can not be locked out forever under load. This
		1425	enables your program to keep a lower latency for important connections
		1426	during short periods of high load, while not completely locking out less
		1427	important ones.
402		1428
403		1429
404	=head1 WATCHER TYPES	1430	=head1 WATCHER TYPES
405		1431
406	This section describes each watcher in detail, but will not repeat	1432	This section describes each watcher in detail, but will not repeat
407	information given in the last section.	1433	information given in the last section. Any initialisation/set macros,
		1434	functions and members specific to the watcher type are explained.
408		1435
		1436	Members are additionally marked with either I<[read-only]>, meaning that,
		1437	while the watcher is active, you can look at the member and expect some
		1438	sensible content, but you must not modify it (you can modify it while the
		1439	watcher is stopped to your hearts content), or I<[read-write]>, which
		1440	means you can expect it to have some sensible content while the watcher
		1441	is active, but you can also modify it. Modifying it may not do something
		1442	sensible or take immediate effect (or do anything at all), but libev will
		1443	not crash or malfunction in any way.
		1444
		1445
409	=head2 C<ev_io> - is this file descriptor readable or writable	1446	=head2 C<ev_io> - is this file descriptor readable or writable?
410		1447
411	I/O watchers check whether a file descriptor is readable or writable	1448	I/O watchers check whether a file descriptor is readable or writable
412	in each iteration of the event loop (This behaviour is called	1449	in each iteration of the event loop, or, more precisely, when reading
413	level-triggering because you keep receiving events as long as the	1450	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	1451	some data. This behaviour is called level-triggering because you keep
415	act on the event and neither want to receive future events).	1452	receiving events as long as the condition persists. Remember you can stop
		1453	the watcher if you don't want to act on the event and neither want to
		1454	receive future events.
416		1455
417	In general you can register as many read and/or write event watchers oer	1456	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	1457	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	1458	descriptors to non-blocking mode is also usually a good idea (but not
420	required if you know what you are doing).	1459	required if you know what you are doing).
421		1460
422	You have to be careful with dup'ed file descriptors, though. Some backends	1461	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	1462	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	1463	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
425	to the same file/socket etc. description.	1464	descriptors for which non-blocking operation makes no sense (such as
		1465	files) - libev doesn't guarentee any specific behaviour in that case.
426		1466
427	If you must do this, then force the use of a known-to-be-good backend	1467	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	1468	receive "spurious" readiness notifications, that is your callback might
429	EVMETHOD_POLL).	1469	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1470	because there is no data. Not only are some backends known to create a
		1471	lot of those (for example Solaris ports), it is very easy to get into
		1472	this situation even with a relatively standard program structure. Thus
		1473	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1474	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1475
		1476	If you cannot run the fd in non-blocking mode (for example you should
		1477	not play around with an Xlib connection), then you have to separately
		1478	re-test whether a file descriptor is really ready with a known-to-be good
		1479	interface such as poll (fortunately in our Xlib example, Xlib already
		1480	does this on its own, so its quite safe to use). Some people additionally
		1481	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1482	indefinitely.
		1483
		1484	But really, best use non-blocking mode.
		1485
		1486	=head3 The special problem of disappearing file descriptors
		1487
		1488	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1489	descriptor (either due to calling C<close> explicitly or any other means,
		1490	such as C<dup2>). The reason is that you register interest in some file
		1491	descriptor, but when it goes away, the operating system will silently drop
		1492	this interest. If another file descriptor with the same number then is
		1493	registered with libev, there is no efficient way to see that this is, in
		1494	fact, a different file descriptor.
		1495
		1496	To avoid having to explicitly tell libev about such cases, libev follows
		1497	the following policy: Each time C<ev_io_set> is being called, libev
		1498	will assume that this is potentially a new file descriptor, otherwise
		1499	it is assumed that the file descriptor stays the same. That means that
		1500	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1501	descriptor even if the file descriptor number itself did not change.
		1502
		1503	This is how one would do it normally anyway, the important point is that
		1504	the libev application should not optimise around libev but should leave
		1505	optimisations to libev.
		1506
		1507	=head3 The special problem of dup'ed file descriptors
		1508
		1509	Some backends (e.g. epoll), cannot register events for file descriptors,
		1510	but only events for the underlying file descriptions. That means when you
		1511	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1512	events for them, only one file descriptor might actually receive events.
		1513
		1514	There is no workaround possible except not registering events
		1515	for potentially C<dup ()>'ed file descriptors, or to resort to
		1516	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1517
		1518	=head3 The special problem of fork
		1519
		1520	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1521	useless behaviour. Libev fully supports fork, but needs to be told about
		1522	it in the child.
		1523
		1524	To support fork in your programs, you either have to call
		1525	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1526	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1527	C<EVBACKEND_POLL>.
		1528
		1529	=head3 The special problem of SIGPIPE
		1530
		1531	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1532	when writing to a pipe whose other end has been closed, your program gets
		1533	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1534	this is sensible behaviour, for daemons, this is usually undesirable.
		1535
		1536	So when you encounter spurious, unexplained daemon exits, make sure you
		1537	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1538	somewhere, as that would have given you a big clue).
		1539
		1540	=head3 The special problem of accept()ing when you can't
		1541
		1542	Many implementations of the POSIX C<accept> function (for example,
		1543	found in port-2004 Linux) have the peculiar behaviour of not removing a
		1544	connection from the pending queue in all error cases.
		1545
		1546	For example, larger servers often run out of file descriptors (because
		1547	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1548	rejecting the connection, leading to libev signalling readiness on
		1549	the next iteration again (the connection still exists after all), and
		1550	typically causing the program to loop at 100% CPU usage.
		1551
		1552	Unfortunately, the set of errors that cause this issue differs between
		1553	operating systems, there is usually little the app can do to remedy the
		1554	situation, and no known thread-safe method of removing the connection to
		1555	cope with overload is known (to me).
		1556
		1557	One of the easiest ways to handle this situation is to just ignore it
		1558	- when the program encounters an overload, it will just loop until the
		1559	situation is over. While this is a form of busy waiting, no OS offers an
		1560	event-based way to handle this situation, so it's the best one can do.
		1561
		1562	A better way to handle the situation is to log any errors other than
		1563	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1564	messages, and continue as usual, which at least gives the user an idea of
		1565	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1566	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1567	usage.
		1568
		1569	If your program is single-threaded, then you could also keep a dummy file
		1570	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1571	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1572	close that fd, and create a new dummy fd. This will gracefully refuse
		1573	clients under typical overload conditions.
		1574
		1575	The last way to handle it is to simply log the error and C<exit>, as
		1576	is often done with C<malloc> failures, but this results in an easy
		1577	opportunity for a DoS attack.
		1578
		1579	=head3 Watcher-Specific Functions
430		1580
431	=over 4	1581	=over 4
432		1582
433	=item ev_io_init (ev_io *, callback, int fd, int events)	1583	=item ev_io_init (ev_io *, callback, int fd, int events)
434		1584
435	=item ev_io_set (ev_io *, int fd, int events)	1585	=item ev_io_set (ev_io *, int fd, int events)
436		1586
437	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1587	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 |	1588	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
439	EV_WRITE> to receive the given events.	1589	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1590
		1591	=item int fd [read-only]
		1592
		1593	The file descriptor being watched.
		1594
		1595	=item int events [read-only]
		1596
		1597	The events being watched.
440		1598
441	=back	1599	=back
442		1600
		1601	=head3 Examples
		1602
		1603	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1604	readable, but only once. Since it is likely line-buffered, you could
		1605	attempt to read a whole line in the callback.
		1606
		1607	static void
		1608	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1609	{
		1610	ev_io_stop (loop, w);
		1611	.. read from stdin here (or from w->fd) and handle any I/O errors
		1612	}
		1613
		1614	...
		1615	struct ev_loop *loop = ev_default_init (0);
		1616	ev_io stdin_readable;
		1617	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1618	ev_io_start (loop, &stdin_readable);
		1619	ev_loop (loop, 0);
		1620
		1621
443	=head2 C<ev_timer> - relative and optionally recurring timeouts	1622	=head2 C<ev_timer> - relative and optionally repeating timeouts
444		1623
445	Timer watchers are simple relative timers that generate an event after a	1624	Timer watchers are simple relative timers that generate an event after a
446	given time, and optionally repeating in regular intervals after that.	1625	given time, and optionally repeating in regular intervals after that.
447		1626
448	The timers are based on real time, that is, if you register an event that	1627	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	1628	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	1629	year, it will still time out after (roughly) one hour. "Roughly" because
451	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1630	detecting time jumps is hard, and some inaccuracies are unavoidable (the
452	monotonic clock option helps a lot here).	1631	monotonic clock option helps a lot here).
		1632
		1633	The callback is guaranteed to be invoked only I<after> its timeout has
		1634	passed (not I<at>, so on systems with very low-resolution clocks this
		1635	might introduce a small delay). If multiple timers become ready during the
		1636	same loop iteration then the ones with earlier time-out values are invoked
		1637	before ones of the same priority with later time-out values (but this is
		1638	no longer true when a callback calls C<ev_loop> recursively).
		1639
		1640	=head3 Be smart about timeouts
		1641
		1642	Many real-world problems involve some kind of timeout, usually for error
		1643	recovery. A typical example is an HTTP request - if the other side hangs,
		1644	you want to raise some error after a while.
		1645
		1646	What follows are some ways to handle this problem, from obvious and
		1647	inefficient to smart and efficient.
		1648
		1649	In the following, a 60 second activity timeout is assumed - a timeout that
		1650	gets reset to 60 seconds each time there is activity (e.g. each time some
		1651	data or other life sign was received).
		1652
		1653	=over 4
		1654
		1655	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1656
		1657	This is the most obvious, but not the most simple way: In the beginning,
		1658	start the watcher:
		1659
		1660	ev_timer_init (timer, callback, 60., 0.);
		1661	ev_timer_start (loop, timer);
		1662
		1663	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1664	and start it again:
		1665
		1666	ev_timer_stop (loop, timer);
		1667	ev_timer_set (timer, 60., 0.);
		1668	ev_timer_start (loop, timer);
		1669
		1670	This is relatively simple to implement, but means that each time there is
		1671	some activity, libev will first have to remove the timer from its internal
		1672	data structure and then add it again. Libev tries to be fast, but it's
		1673	still not a constant-time operation.
		1674
		1675	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1676
		1677	This is the easiest way, and involves using C<ev_timer_again> instead of
		1678	C<ev_timer_start>.
		1679
		1680	To implement this, configure an C<ev_timer> with a C<repeat> value
		1681	of C<60> and then call C<ev_timer_again> at start and each time you
		1682	successfully read or write some data. If you go into an idle state where
		1683	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1684	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1685
		1686	That means you can ignore both the C<ev_timer_start> function and the
		1687	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1688	member and C<ev_timer_again>.
		1689
		1690	At start:
		1691
		1692	ev_init (timer, callback);
		1693	timer->repeat = 60.;
		1694	ev_timer_again (loop, timer);
		1695
		1696	Each time there is some activity:
		1697
		1698	ev_timer_again (loop, timer);
		1699
		1700	It is even possible to change the time-out on the fly, regardless of
		1701	whether the watcher is active or not:
		1702
		1703	timer->repeat = 30.;
		1704	ev_timer_again (loop, timer);
		1705
		1706	This is slightly more efficient then stopping/starting the timer each time
		1707	you want to modify its timeout value, as libev does not have to completely
		1708	remove and re-insert the timer from/into its internal data structure.
		1709
		1710	It is, however, even simpler than the "obvious" way to do it.
		1711
		1712	=item 3. Let the timer time out, but then re-arm it as required.
		1713
		1714	This method is more tricky, but usually most efficient: Most timeouts are
		1715	relatively long compared to the intervals between other activity - in
		1716	our example, within 60 seconds, there are usually many I/O events with
		1717	associated activity resets.
		1718
		1719	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1720	but remember the time of last activity, and check for a real timeout only
		1721	within the callback:
		1722
		1723	ev_tstamp last_activity; // time of last activity
		1724
		1725	static void
		1726	callback (EV_P_ ev_timer *w, int revents)
		1727	{
		1728	ev_tstamp now = ev_now (EV_A);
		1729	ev_tstamp timeout = last_activity + 60.;
		1730
		1731	// if last_activity + 60. is older than now, we did time out
		1732	if (timeout < now)
		1733	{
		1734	// timeout occured, take action
		1735	}
		1736	else
		1737	{
		1738	// callback was invoked, but there was some activity, re-arm
		1739	// the watcher to fire in last_activity + 60, which is
		1740	// guaranteed to be in the future, so "again" is positive:
		1741	w->repeat = timeout - now;
		1742	ev_timer_again (EV_A_ w);
		1743	}
		1744	}
		1745
		1746	To summarise the callback: first calculate the real timeout (defined
		1747	as "60 seconds after the last activity"), then check if that time has
		1748	been reached, which means something I<did>, in fact, time out. Otherwise
		1749	the callback was invoked too early (C<timeout> is in the future), so
		1750	re-schedule the timer to fire at that future time, to see if maybe we have
		1751	a timeout then.
		1752
		1753	Note how C<ev_timer_again> is used, taking advantage of the
		1754	C<ev_timer_again> optimisation when the timer is already running.
		1755
		1756	This scheme causes more callback invocations (about one every 60 seconds
		1757	minus half the average time between activity), but virtually no calls to
		1758	libev to change the timeout.
		1759
		1760	To start the timer, simply initialise the watcher and set C<last_activity>
		1761	to the current time (meaning we just have some activity :), then call the
		1762	callback, which will "do the right thing" and start the timer:
		1763
		1764	ev_init (timer, callback);
		1765	last_activity = ev_now (loop);
		1766	callback (loop, timer, EV_TIMER);
		1767
		1768	And when there is some activity, simply store the current time in
		1769	C<last_activity>, no libev calls at all:
		1770
		1771	last_actiivty = ev_now (loop);
		1772
		1773	This technique is slightly more complex, but in most cases where the
		1774	time-out is unlikely to be triggered, much more efficient.
		1775
		1776	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1777	callback :) - just change the timeout and invoke the callback, which will
		1778	fix things for you.
		1779
		1780	=item 4. Wee, just use a double-linked list for your timeouts.
		1781
		1782	If there is not one request, but many thousands (millions...), all
		1783	employing some kind of timeout with the same timeout value, then one can
		1784	do even better:
		1785
		1786	When starting the timeout, calculate the timeout value and put the timeout
		1787	at the I<end> of the list.
		1788
		1789	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1790	the list is expected to fire (for example, using the technique #3).
		1791
		1792	When there is some activity, remove the timer from the list, recalculate
		1793	the timeout, append it to the end of the list again, and make sure to
		1794	update the C<ev_timer> if it was taken from the beginning of the list.
		1795
		1796	This way, one can manage an unlimited number of timeouts in O(1) time for
		1797	starting, stopping and updating the timers, at the expense of a major
		1798	complication, and having to use a constant timeout. The constant timeout
		1799	ensures that the list stays sorted.
		1800
		1801	=back
		1802
		1803	So which method the best?
		1804
		1805	Method #2 is a simple no-brain-required solution that is adequate in most
		1806	situations. Method #3 requires a bit more thinking, but handles many cases
		1807	better, and isn't very complicated either. In most case, choosing either
		1808	one is fine, with #3 being better in typical situations.
		1809
		1810	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1811	rather complicated, but extremely efficient, something that really pays
		1812	off after the first million or so of active timers, i.e. it's usually
		1813	overkill :)
		1814
		1815	=head3 The special problem of time updates
		1816
		1817	Establishing the current time is a costly operation (it usually takes at
		1818	least two system calls): EV therefore updates its idea of the current
		1819	time only before and after C<ev_loop> collects new events, which causes a
		1820	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1821	lots of events in one iteration.
453		1822
454	The relative timeouts are calculated relative to the C<ev_now ()>	1823	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	1824	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	1825	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	1826	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:	1827	timeout on the current time, use something like this to adjust for this:
459		1828
460	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1829	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
461		1830
		1831	If the event loop is suspended for a long time, you can also force an
		1832	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1833	()>.
		1834
		1835	=head3 The special problems of suspended animation
		1836
		1837	When you leave the server world it is quite customary to hit machines that
		1838	can suspend/hibernate - what happens to the clocks during such a suspend?
		1839
		1840	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1841	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1842	to run until the system is suspended, but they will not advance while the
		1843	system is suspended. That means, on resume, it will be as if the program
		1844	was frozen for a few seconds, but the suspend time will not be counted
		1845	towards C<ev_timer> when a monotonic clock source is used. The real time
		1846	clock advanced as expected, but if it is used as sole clocksource, then a
		1847	long suspend would be detected as a time jump by libev, and timers would
		1848	be adjusted accordingly.
		1849
		1850	I would not be surprised to see different behaviour in different between
		1851	operating systems, OS versions or even different hardware.
		1852
		1853	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1854	time jump in the monotonic clocks and the realtime clock. If the program
		1855	is suspended for a very long time, and monotonic clock sources are in use,
		1856	then you can expect C<ev_timer>s to expire as the full suspension time
		1857	will be counted towards the timers. When no monotonic clock source is in
		1858	use, then libev will again assume a timejump and adjust accordingly.
		1859
		1860	It might be beneficial for this latter case to call C<ev_suspend>
		1861	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1862	deterministic behaviour in this case (you can do nothing against
		1863	C<SIGSTOP>).
		1864
		1865	=head3 Watcher-Specific Functions and Data Members
		1866
462	=over 4	1867	=over 4
463		1868
464	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1869	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
465		1870
466	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1871	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
467		1872
468	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1873	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	1874	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	1875	reached. If it is positive, then the timer will automatically be
471	later, again, and again, until stopped manually.	1876	configured to trigger again C<repeat> seconds later, again, and again,
		1877	until stopped manually.
472		1878
473	The timer itself will do a best-effort at avoiding drift, that is, if you	1879	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	1880	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	1881	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	1882	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.	1883	do stuff) the timer will not fire more than once per event loop iteration.
478		1884
479	=item ev_timer_again (loop)	1885	=item ev_timer_again (loop, ev_timer *)
480		1886
481	This will act as if the timer timed out and restart it again if it is	1887	This will act as if the timer timed out and restart it again if it is
482	repeating. The exact semantics are:	1888	repeating. The exact semantics are:
483		1889
		1890	If the timer is pending, its pending status is cleared.
		1891
484	If the timer is started but nonrepeating, stop it.	1892	If the timer is started but non-repeating, stop it (as if it timed out).
485		1893
486	If the timer is repeating, either start it if necessary (with the repeat	1894	If the timer is repeating, either start it if necessary (with the
487	value), or reset the running timer to the repeat value.	1895	C<repeat> value), or reset the running timer to the C<repeat> value.
488		1896
489	This sounds a bit complicated, but here is a useful and typical	1897	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	1898	usage example.
491	timeout, that is, you want to be called when there have been, say, 60	1899
492	seconds of inactivity on the socket. The easiest way to do this is to	1900	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
493	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1901
494	time you successfully read or write some data. If you go into an idle	1902	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	1903	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.	1904	the timeout value currently configured.
		1905
		1906	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1907	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		1908	will return C<4>. When the timer expires and is restarted, it will return
		1909	roughly C<7> (likely slightly less as callback invocation takes some time,
		1910	too), and so on.
		1911
		1912	=item ev_tstamp repeat [read-write]
		1913
		1914	The current C<repeat> value. Will be used each time the watcher times out
		1915	or C<ev_timer_again> is called, and determines the next timeout (if any),
		1916	which is also when any modifications are taken into account.
497		1917
498	=back	1918	=back
499		1919
		1920	=head3 Examples
		1921
		1922	Example: Create a timer that fires after 60 seconds.
		1923
		1924	static void
		1925	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1926	{
		1927	.. one minute over, w is actually stopped right here
		1928	}
		1929
		1930	ev_timer mytimer;
		1931	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1932	ev_timer_start (loop, &mytimer);
		1933
		1934	Example: Create a timeout timer that times out after 10 seconds of
		1935	inactivity.
		1936
		1937	static void
		1938	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1939	{
		1940	.. ten seconds without any activity
		1941	}
		1942
		1943	ev_timer mytimer;
		1944	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1945	ev_timer_again (&mytimer); /* start timer */
		1946	ev_loop (loop, 0);
		1947
		1948	// and in some piece of code that gets executed on any "activity":
		1949	// reset the timeout to start ticking again at 10 seconds
		1950	ev_timer_again (&mytimer);
		1951
		1952
500	=head2 C<ev_periodic> - to cron or not to cron	1953	=head2 C<ev_periodic> - to cron or not to cron?
501		1954
502	Periodic watchers are also timers of a kind, but they are very versatile	1955	Periodic watchers are also timers of a kind, but they are very versatile
503	(and unfortunately a bit complex).	1956	(and unfortunately a bit complex).
504		1957
505	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1958	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	1959	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	1960	(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 ()	1961	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	1962	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	1963	wrist-watch).
511	roughly 10 seconds later and of course not if you reset your system time
512	again).
513		1964
514	They can also be used to implement vastly more complex timers, such as	1965	You can tell a periodic watcher to trigger after some specific point
		1966	in time: for example, if you tell a periodic watcher to trigger "in 10
		1967	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1968	not a delay) and then reset your system clock to January of the previous
		1969	year, then it will take a year or more to trigger the event (unlike an
		1970	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1971	it, as it uses a relative timeout).
		1972
		1973	C<ev_periodic> watchers can also be used to implement vastly more complex
515	triggering an event on eahc midnight, local time.	1974	timers, such as triggering an event on each "midnight, local time", or
		1975	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1976	those cannot react to time jumps.
		1977
		1978	As with timers, the callback is guaranteed to be invoked only when the
		1979	point in time where it is supposed to trigger has passed. If multiple
		1980	timers become ready during the same loop iteration then the ones with
		1981	earlier time-out values are invoked before ones with later time-out values
		1982	(but this is no longer true when a callback calls C<ev_loop> recursively).
		1983
		1984	=head3 Watcher-Specific Functions and Data Members
516		1985
517	=over 4	1986	=over 4
518		1987
519	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1988	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
520		1989
521	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1990	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
522		1991
523	Lots of arguments, lets sort it out... There are basically three modes of	1992	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:	1993	operation, and we will explain them from simplest to most complex:
525
526		1994
527	=over 4	1995	=over 4
528		1996
529	=item * absolute timer (interval = reschedule_cb = 0)	1997	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
530		1998
531	In this configuration the watcher triggers an event at the wallclock time	1999	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,	2000	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	2001	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.	2002	will be stopped and invoked when the system clock reaches or surpasses
		2003	this point in time.
535		2004
536	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2005	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
537		2006
538	In this mode the watcher will always be scheduled to time out at the next	2007	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	2008	C<offset + N * interval> time (for some integer N, which can also be
540	of any time jumps.	2009	negative) and then repeat, regardless of any time jumps. The C<offset>
		2010	argument is merely an offset into the C<interval> periods.
541		2011
542	This can be used to create timers that do not drift with respect to system	2012	This can be used to create timers that do not drift with respect to the
543	time:	2013	system clock, for example, here is an C<ev_periodic> that triggers each
		2014	hour, on the hour (with respect to UTC):
544		2015
545	ev_periodic_set (&periodic, 0., 3600., 0);	2016	ev_periodic_set (&periodic, 0., 3600., 0);
546		2017
547	This doesn't mean there will always be 3600 seconds in between triggers,	2018	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	2019	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	2020	full hour (UTC), or more correctly, when the system time is evenly divisible
550	by 3600.	2021	by 3600.
551		2022
552	Another way to think about it (for the mathematically inclined) is that	2023	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	2024	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.	2025	time where C<time = offset (mod interval)>, regardless of any time jumps.
555		2026
		2027	For numerical stability it is preferable that the C<offset> value is near
		2028	C<ev_now ()> (the current time), but there is no range requirement for
		2029	this value, and in fact is often specified as zero.
		2030
		2031	Note also that there is an upper limit to how often a timer can fire (CPU
		2032	speed for example), so if C<interval> is very small then timing stability
		2033	will of course deteriorate. Libev itself tries to be exact to be about one
		2034	millisecond (if the OS supports it and the machine is fast enough).
		2035
556	=item * manual reschedule mode (reschedule_cb = callback)	2036	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
557		2037
558	In this mode the values for C<interval> and C<at> are both being	2038	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	2039	ignored. Instead, each time the periodic watcher gets scheduled, the
560	reschedule callback will be called with the watcher as first, and the	2040	reschedule callback will be called with the watcher as first, and the
561	current time as second argument.	2041	current time as second argument.
562		2042
563	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	2043	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	2044	or make ANY other event loop modifications whatsoever, unless explicitly
		2045	allowed by documentation here>.
		2046
565	to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards.	2047	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2048	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2049	only event loop modification you are allowed to do).
566		2050
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,	2051	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
571	ev_tstamp now)>, e.g.:	2052	*w, ev_tstamp now)>, e.g.:
572		2053
		2054	static ev_tstamp
573	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2055	my_rescheduler (ev_periodic *w, ev_tstamp now)
574	{	2056	{
575	return now + 60.;	2057	return now + 60.;
576	}	2058	}
577		2059
578	It must return the next time to trigger, based on the passed time value	2060	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	2061	(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	2062	will usually be called just before the callback will be triggered, but
581	might be called at other times, too.	2063	might be called at other times, too.
582		2064
		2065	NOTE: I<< This callback must always return a time that is higher than or
		2066	equal to the passed C<now> value >>.
		2067
583	This can be used to create very complex timers, such as a timer that	2068	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	2069	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	2070	next midnight after C<now> and return the timestamp value for this. How
586	is, again, up to you (but it is not trivial).	2071	you do this is, again, up to you (but it is not trivial, which is the main
		2072	reason I omitted it as an example).
587		2073
588	=back	2074	=back
589		2075
590	=item ev_periodic_again (loop, ev_periodic *)	2076	=item ev_periodic_again (loop, ev_periodic *)
591		2077
592	Simply stops and restarts the periodic watcher again. This is only useful	2078	Simply stops and restarts the periodic watcher again. This is only useful
593	when you changed some parameters or the reschedule callback would return	2079	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	2080	a different time than the last time it was called (e.g. in a crond like
595	program when the crontabs have changed).	2081	program when the crontabs have changed).
596		2082
		2083	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2084
		2085	When active, returns the absolute time that the watcher is supposed
		2086	to trigger next. This is not the same as the C<offset> argument to
		2087	C<ev_periodic_set>, but indeed works even in interval and manual
		2088	rescheduling modes.
		2089
		2090	=item ev_tstamp offset [read-write]
		2091
		2092	When repeating, this contains the offset value, otherwise this is the
		2093	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2094	although libev might modify this value for better numerical stability).
		2095
		2096	Can be modified any time, but changes only take effect when the periodic
		2097	timer fires or C<ev_periodic_again> is being called.
		2098
		2099	=item ev_tstamp interval [read-write]
		2100
		2101	The current interval value. Can be modified any time, but changes only
		2102	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2103	called.
		2104
		2105	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2106
		2107	The current reschedule callback, or C<0>, if this functionality is
		2108	switched off. Can be changed any time, but changes only take effect when
		2109	the periodic timer fires or C<ev_periodic_again> is being called.
		2110
597	=back	2111	=back
598		2112
		2113	=head3 Examples
		2114
		2115	Example: Call a callback every hour, or, more precisely, whenever the
		2116	system time is divisible by 3600. The callback invocation times have
		2117	potentially a lot of jitter, but good long-term stability.
		2118
		2119	static void
		2120	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		2121	{
		2122	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2123	}
		2124
		2125	ev_periodic hourly_tick;
		2126	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2127	ev_periodic_start (loop, &hourly_tick);
		2128
		2129	Example: The same as above, but use a reschedule callback to do it:
		2130
		2131	#include <math.h>
		2132
		2133	static ev_tstamp
		2134	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2135	{
		2136	return now + (3600. - fmod (now, 3600.));
		2137	}
		2138
		2139	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2140
		2141	Example: Call a callback every hour, starting now:
		2142
		2143	ev_periodic hourly_tick;
		2144	ev_periodic_init (&hourly_tick, clock_cb,
		2145	fmod (ev_now (loop), 3600.), 3600., 0);
		2146	ev_periodic_start (loop, &hourly_tick);
		2147
		2148
599	=head2 C<ev_signal> - signal me when a signal gets signalled	2149	=head2 C<ev_signal> - signal me when a signal gets signalled!
600		2150
601	Signal watchers will trigger an event when the process receives a specific	2151	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	2152	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	2153	will try it's best to deliver signals synchronously, i.e. as part of the
604	normal event processing, like any other event.	2154	normal event processing, like any other event.
605		2155
		2156	If you want signals to be delivered truly asynchronously, just use
		2157	C<sigaction> as you would do without libev and forget about sharing
		2158	the signal. You can even use C<ev_async> from a signal handler to
		2159	synchronously wake up an event loop.
		2160
606	You can configure as many watchers as you like per signal. Only when the	2161	You can configure as many watchers as you like for the same signal, but
		2162	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2163	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2164	C<SIGINT> in both the default loop and another loop at the same time. At
		2165	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2166
607	first watcher gets started will libev actually register a signal watcher	2167	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	2168	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	2169	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	2170
611	SIG_DFL (regardless of what it was set to before).	2171	If possible and supported, libev will install its handlers with
		2172	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2173	not be unduly interrupted. If you have a problem with system calls getting
		2174	interrupted by signals you can block all signals in an C<ev_check> watcher
		2175	and unblock them in an C<ev_prepare> watcher.
		2176
		2177	=head3 The special problem of inheritance over fork/execve/pthread_create
		2178
		2179	Both the signal mask (C<sigprocmask>) and the signal disposition
		2180	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2181	stopping it again), that is, libev might or might not block the signal,
		2182	and might or might not set or restore the installed signal handler.
		2183
		2184	While this does not matter for the signal disposition (libev never
		2185	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2186	C<execve>), this matters for the signal mask: many programs do not expect
		2187	certain signals to be blocked.
		2188
		2189	This means that before calling C<exec> (from the child) you should reset
		2190	the signal mask to whatever "default" you expect (all clear is a good
		2191	choice usually).
		2192
		2193	The simplest way to ensure that the signal mask is reset in the child is
		2194	to install a fork handler with C<pthread_atfork> that resets it. That will
		2195	catch fork calls done by libraries (such as the libc) as well.
		2196
		2197	In current versions of libev, the signal will not be blocked indefinitely
		2198	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2199	the window of opportunity for problems, it will not go away, as libev
		2200	I<has> to modify the signal mask, at least temporarily.
		2201
		2202	So I can't stress this enough: I<If you do not reset your signal mask when
		2203	you expect it to be empty, you have a race condition in your code>. This
		2204	is not a libev-specific thing, this is true for most event libraries.
		2205
		2206	=head3 Watcher-Specific Functions and Data Members
612		2207
613	=over 4	2208	=over 4
614		2209
615	=item ev_signal_init (ev_signal *, callback, int signum)	2210	=item ev_signal_init (ev_signal *, callback, int signum)
616		2211
617	=item ev_signal_set (ev_signal *, int signum)	2212	=item ev_signal_set (ev_signal *, int signum)
618		2213
619	Configures the watcher to trigger on the given signal number (usually one	2214	Configures the watcher to trigger on the given signal number (usually one
620	of the C<SIGxxx> constants).	2215	of the C<SIGxxx> constants).
621		2216
		2217	=item int signum [read-only]
		2218
		2219	The signal the watcher watches out for.
		2220
622	=back	2221	=back
623		2222
		2223	=head3 Examples
		2224
		2225	Example: Try to exit cleanly on SIGINT.
		2226
		2227	static void
		2228	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2229	{
		2230	ev_unloop (loop, EVUNLOOP_ALL);
		2231	}
		2232
		2233	ev_signal signal_watcher;
		2234	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2235	ev_signal_start (loop, &signal_watcher);
		2236
		2237
624	=head2 C<ev_child> - wait for pid status changes	2238	=head2 C<ev_child> - watch out for process status changes
625		2239
626	Child watchers trigger when your process receives a SIGCHLD in response to	2240	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).	2241	some child status changes (most typically when a child of yours dies or
		2242	exits). It is permissible to install a child watcher I<after> the child
		2243	has been forked (which implies it might have already exited), as long
		2244	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2245	forking and then immediately registering a watcher for the child is fine,
		2246	but forking and registering a watcher a few event loop iterations later or
		2247	in the next callback invocation is not.
		2248
		2249	Only the default event loop is capable of handling signals, and therefore
		2250	you can only register child watchers in the default event loop.
		2251
		2252	Due to some design glitches inside libev, child watchers will always be
		2253	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2254	libev)
		2255
		2256	=head3 Process Interaction
		2257
		2258	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2259	initialised. This is necessary to guarantee proper behaviour even if the
		2260	first child watcher is started after the child exits. The occurrence
		2261	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2262	synchronously as part of the event loop processing. Libev always reaps all
		2263	children, even ones not watched.
		2264
		2265	=head3 Overriding the Built-In Processing
		2266
		2267	Libev offers no special support for overriding the built-in child
		2268	processing, but if your application collides with libev's default child
		2269	handler, you can override it easily by installing your own handler for
		2270	C<SIGCHLD> after initialising the default loop, and making sure the
		2271	default loop never gets destroyed. You are encouraged, however, to use an
		2272	event-based approach to child reaping and thus use libev's support for
		2273	that, so other libev users can use C<ev_child> watchers freely.
		2274
		2275	=head3 Stopping the Child Watcher
		2276
		2277	Currently, the child watcher never gets stopped, even when the
		2278	child terminates, so normally one needs to stop the watcher in the
		2279	callback. Future versions of libev might stop the watcher automatically
		2280	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2281	problem).
		2282
		2283	=head3 Watcher-Specific Functions and Data Members
628		2284
629	=over 4	2285	=over 4
630		2286
631	=item ev_child_init (ev_child *, callback, int pid)	2287	=item ev_child_init (ev_child *, callback, int pid, int trace)
632		2288
633	=item ev_child_set (ev_child *, int pid)	2289	=item ev_child_set (ev_child *, int pid, int trace)
634		2290
635	Configures the watcher to wait for status changes of process C<pid> (or	2291	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	2292	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	2293	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	2294	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	2295	C<waitpid> documentation). The C<rpid> member contains the pid of the
640	process causing the status change.	2296	process causing the status change. C<trace> must be either C<0> (only
		2297	activate the watcher when the process terminates) or C<1> (additionally
		2298	activate the watcher when the process is stopped or continued).
		2299
		2300	=item int pid [read-only]
		2301
		2302	The process id this watcher watches out for, or C<0>, meaning any process id.
		2303
		2304	=item int rpid [read-write]
		2305
		2306	The process id that detected a status change.
		2307
		2308	=item int rstatus [read-write]
		2309
		2310	The process exit/trace status caused by C<rpid> (see your systems
		2311	C<waitpid> and C<sys/wait.h> documentation for details).
641		2312
642	=back	2313	=back
643		2314
		2315	=head3 Examples
		2316
		2317	Example: C<fork()> a new process and install a child handler to wait for
		2318	its completion.
		2319
		2320	ev_child cw;
		2321
		2322	static void
		2323	child_cb (EV_P_ ev_child *w, int revents)
		2324	{
		2325	ev_child_stop (EV_A_ w);
		2326	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2327	}
		2328
		2329	pid_t pid = fork ();
		2330
		2331	if (pid < 0)
		2332	// error
		2333	else if (pid == 0)
		2334	{
		2335	// the forked child executes here
		2336	exit (1);
		2337	}
		2338	else
		2339	{
		2340	ev_child_init (&cw, child_cb, pid, 0);
		2341	ev_child_start (EV_DEFAULT_ &cw);
		2342	}
		2343
		2344
		2345	=head2 C<ev_stat> - did the file attributes just change?
		2346
		2347	This watches a file system path for attribute changes. That is, it calls
		2348	C<stat> on that path in regular intervals (or when the OS says it changed)
		2349	and sees if it changed compared to the last time, invoking the callback if
		2350	it did.
		2351
		2352	The path does not need to exist: changing from "path exists" to "path does
		2353	not exist" is a status change like any other. The condition "path does not
		2354	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2355	C<st_nlink> field being zero (which is otherwise always forced to be at
		2356	least one) and all the other fields of the stat buffer having unspecified
		2357	contents.
		2358
		2359	The path I<must not> end in a slash or contain special components such as
		2360	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2361	your working directory changes, then the behaviour is undefined.
		2362
		2363	Since there is no portable change notification interface available, the
		2364	portable implementation simply calls C<stat(2)> regularly on the path
		2365	to see if it changed somehow. You can specify a recommended polling
		2366	interval for this case. If you specify a polling interval of C<0> (highly
		2367	recommended!) then a I<suitable, unspecified default> value will be used
		2368	(which you can expect to be around five seconds, although this might
		2369	change dynamically). Libev will also impose a minimum interval which is
		2370	currently around C<0.1>, but that's usually overkill.
		2371
		2372	This watcher type is not meant for massive numbers of stat watchers,
		2373	as even with OS-supported change notifications, this can be
		2374	resource-intensive.
		2375
		2376	At the time of this writing, the only OS-specific interface implemented
		2377	is the Linux inotify interface (implementing kqueue support is left as an
		2378	exercise for the reader. Note, however, that the author sees no way of
		2379	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2380
		2381	=head3 ABI Issues (Largefile Support)
		2382
		2383	Libev by default (unless the user overrides this) uses the default
		2384	compilation environment, which means that on systems with large file
		2385	support disabled by default, you get the 32 bit version of the stat
		2386	structure. When using the library from programs that change the ABI to
		2387	use 64 bit file offsets the programs will fail. In that case you have to
		2388	compile libev with the same flags to get binary compatibility. This is
		2389	obviously the case with any flags that change the ABI, but the problem is
		2390	most noticeably displayed with ev_stat and large file support.
		2391
		2392	The solution for this is to lobby your distribution maker to make large
		2393	file interfaces available by default (as e.g. FreeBSD does) and not
		2394	optional. Libev cannot simply switch on large file support because it has
		2395	to exchange stat structures with application programs compiled using the
		2396	default compilation environment.
		2397
		2398	=head3 Inotify and Kqueue
		2399
		2400	When C<inotify (7)> support has been compiled into libev and present at
		2401	runtime, it will be used to speed up change detection where possible. The
		2402	inotify descriptor will be created lazily when the first C<ev_stat>
		2403	watcher is being started.
		2404
		2405	Inotify presence does not change the semantics of C<ev_stat> watchers
		2406	except that changes might be detected earlier, and in some cases, to avoid
		2407	making regular C<stat> calls. Even in the presence of inotify support
		2408	there are many cases where libev has to resort to regular C<stat> polling,
		2409	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2410	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2411	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2412	xfs are fully working) libev usually gets away without polling.
		2413
		2414	There is no support for kqueue, as apparently it cannot be used to
		2415	implement this functionality, due to the requirement of having a file
		2416	descriptor open on the object at all times, and detecting renames, unlinks
		2417	etc. is difficult.
		2418
		2419	=head3 C<stat ()> is a synchronous operation
		2420
		2421	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2422	the process. The exception are C<ev_stat> watchers - those call C<stat
		2423	()>, which is a synchronous operation.
		2424
		2425	For local paths, this usually doesn't matter: unless the system is very
		2426	busy or the intervals between stat's are large, a stat call will be fast,
		2427	as the path data is usually in memory already (except when starting the
		2428	watcher).
		2429
		2430	For networked file systems, calling C<stat ()> can block an indefinite
		2431	time due to network issues, and even under good conditions, a stat call
		2432	often takes multiple milliseconds.
		2433
		2434	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2435	paths, although this is fully supported by libev.
		2436
		2437	=head3 The special problem of stat time resolution
		2438
		2439	The C<stat ()> system call only supports full-second resolution portably,
		2440	and even on systems where the resolution is higher, most file systems
		2441	still only support whole seconds.
		2442
		2443	That means that, if the time is the only thing that changes, you can
		2444	easily miss updates: on the first update, C<ev_stat> detects a change and
		2445	calls your callback, which does something. When there is another update
		2446	within the same second, C<ev_stat> will be unable to detect unless the
		2447	stat data does change in other ways (e.g. file size).
		2448
		2449	The solution to this is to delay acting on a change for slightly more
		2450	than a second (or till slightly after the next full second boundary), using
		2451	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2452	ev_timer_again (loop, w)>).
		2453
		2454	The C<.02> offset is added to work around small timing inconsistencies
		2455	of some operating systems (where the second counter of the current time
		2456	might be be delayed. One such system is the Linux kernel, where a call to
		2457	C<gettimeofday> might return a timestamp with a full second later than
		2458	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2459	update file times then there will be a small window where the kernel uses
		2460	the previous second to update file times but libev might already execute
		2461	the timer callback).
		2462
		2463	=head3 Watcher-Specific Functions and Data Members
		2464
		2465	=over 4
		2466
		2467	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2468
		2469	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2470
		2471	Configures the watcher to wait for status changes of the given
		2472	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2473	be detected and should normally be specified as C<0> to let libev choose
		2474	a suitable value. The memory pointed to by C<path> must point to the same
		2475	path for as long as the watcher is active.
		2476
		2477	The callback will receive an C<EV_STAT> event when a change was detected,
		2478	relative to the attributes at the time the watcher was started (or the
		2479	last change was detected).
		2480
		2481	=item ev_stat_stat (loop, ev_stat *)
		2482
		2483	Updates the stat buffer immediately with new values. If you change the
		2484	watched path in your callback, you could call this function to avoid
		2485	detecting this change (while introducing a race condition if you are not
		2486	the only one changing the path). Can also be useful simply to find out the
		2487	new values.
		2488
		2489	=item ev_statdata attr [read-only]
		2490
		2491	The most-recently detected attributes of the file. Although the type is
		2492	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2493	suitable for your system, but you can only rely on the POSIX-standardised
		2494	members to be present. If the C<st_nlink> member is C<0>, then there was
		2495	some error while C<stat>ing the file.
		2496
		2497	=item ev_statdata prev [read-only]
		2498
		2499	The previous attributes of the file. The callback gets invoked whenever
		2500	C<prev> != C<attr>, or, more precisely, one or more of these members
		2501	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2502	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2503
		2504	=item ev_tstamp interval [read-only]
		2505
		2506	The specified interval.
		2507
		2508	=item const char *path [read-only]
		2509
		2510	The file system path that is being watched.
		2511
		2512	=back
		2513
		2514	=head3 Examples
		2515
		2516	Example: Watch C</etc/passwd> for attribute changes.
		2517
		2518	static void
		2519	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2520	{
		2521	/* /etc/passwd changed in some way */
		2522	if (w->attr.st_nlink)
		2523	{
		2524	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2525	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2526	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2527	}
		2528	else
		2529	/* you shalt not abuse printf for puts */
		2530	puts ("wow, /etc/passwd is not there, expect problems. "
		2531	"if this is windows, they already arrived\n");
		2532	}
		2533
		2534	...
		2535	ev_stat passwd;
		2536
		2537	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2538	ev_stat_start (loop, &passwd);
		2539
		2540	Example: Like above, but additionally use a one-second delay so we do not
		2541	miss updates (however, frequent updates will delay processing, too, so
		2542	one might do the work both on C<ev_stat> callback invocation I<and> on
		2543	C<ev_timer> callback invocation).
		2544
		2545	static ev_stat passwd;
		2546	static ev_timer timer;
		2547
		2548	static void
		2549	timer_cb (EV_P_ ev_timer *w, int revents)
		2550	{
		2551	ev_timer_stop (EV_A_ w);
		2552
		2553	/* now it's one second after the most recent passwd change */
		2554	}
		2555
		2556	static void
		2557	stat_cb (EV_P_ ev_stat *w, int revents)
		2558	{
		2559	/* reset the one-second timer */
		2560	ev_timer_again (EV_A_ &timer);
		2561	}
		2562
		2563	...
		2564	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2565	ev_stat_start (loop, &passwd);
		2566	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2567
		2568
644	=head2 C<ev_idle> - when you've got nothing better to do	2569	=head2 C<ev_idle> - when you've got nothing better to do...
645		2570
646	Idle watchers trigger events when there are no other events are pending	2571	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	2572	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,	2573	as receiving "events").
649	imagine) it will not be triggered. But when your process is idle all idle	2574
650	watchers are being called again and again, once per event loop iteration -	2575	That is, as long as your process is busy handling sockets or timeouts
		2576	(or even signals, imagine) of the same or higher priority it will not be
		2577	triggered. But when your process is idle (or only lower-priority watchers
		2578	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	2579	iteration - until stopped, that is, or your process receives more events
652	busy.	2580	and becomes busy again with higher priority stuff.
653		2581
654	The most noteworthy effect is that as long as any idle watchers are	2582	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.	2583	active, the process will not block when waiting for new events.
656		2584
657	Apart from keeping your process non-blocking (which is a useful	2585	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	2586	effect on its own sometimes), idle watchers are a good place to do
659	"pseudo-background processing", or delay processing stuff to after the	2587	"pseudo-background processing", or delay processing stuff to after the
660	event loop has handled all outstanding events.	2588	event loop has handled all outstanding events.
661		2589
		2590	=head3 Watcher-Specific Functions and Data Members
		2591
662	=over 4	2592	=over 4
663		2593
664	=item ev_idle_init (ev_signal *, callback)	2594	=item ev_idle_init (ev_idle *, callback)
665		2595
666	Initialises and configures the idle watcher - it has no parameters of any	2596	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,	2597	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
668	believe me.	2598	believe me.
669		2599
670	=back	2600	=back
671		2601
		2602	=head3 Examples
		2603
		2604	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2605	callback, free it. Also, use no error checking, as usual.
		2606
		2607	static void
		2608	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2609	{
		2610	free (w);
		2611	// now do something you wanted to do when the program has
		2612	// no longer anything immediate to do.
		2613	}
		2614
		2615	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2616	ev_idle_init (idle_watcher, idle_cb);
		2617	ev_idle_start (loop, idle_watcher);
		2618
		2619
672	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2620	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
673		2621
674	Prepare and check watchers are usually (but not always) used in tandem:	2622	Prepare and check watchers are usually (but not always) used in pairs:
675	Prepare watchers get invoked before the process blocks and check watchers	2623	prepare watchers get invoked before the process blocks and check watchers
676	afterwards.	2624	afterwards.
677		2625
		2626	You I<must not> call C<ev_loop> or similar functions that enter
		2627	the current event loop from either C<ev_prepare> or C<ev_check>
		2628	watchers. Other loops than the current one are fine, however. The
		2629	rationale behind this is that you do not need to check for recursion in
		2630	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2631	C<ev_check> so if you have one watcher of each kind they will always be
		2632	called in pairs bracketing the blocking call.
		2633
678	Their main purpose is to integrate other event mechanisms into libev. This	2634	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	2635	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.	2636	variable changes, implement your own watchers, integrate net-snmp or a
		2637	coroutine library and lots more. They are also occasionally useful if
		2638	you cache some data and want to flush it before blocking (for example,
		2639	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2640	watcher).
681		2641
682	This is done by examining in each prepare call which file descriptors need	2642	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	2643	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	2644	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	2645	libraries provide exactly this functionality). Then, in the check watcher,
686	any events that occured (by checking the pending status of all watchers	2646	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	2647	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,	2648	I/O and timer callbacks will never actually be called (but must be valid
689	because you never know, you know?).	2649	nevertheless, because you never know, you know?).
690		2650
691	As another example, the Perl Coro module uses these hooks to integrate	2651	As another example, the Perl Coro module uses these hooks to integrate
692	coroutines into libev programs, by yielding to other active coroutines	2652	coroutines into libev programs, by yielding to other active coroutines
693	during each prepare and only letting the process block if no coroutines	2653	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	2654	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,	2655	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	2656	of lower priority, but only once, using idle watchers to keep the event
697	coroutines exist, thus mapping low-priority coroutines to idle/background	2657	loop from blocking if lower-priority coroutines are active, thus mapping
698	tasks).	2658	low-priority coroutines to idle/background tasks).
		2659
		2660	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2661	priority, to ensure that they are being run before any other watchers
		2662	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2663
		2664	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2665	activate ("feed") events into libev. While libev fully supports this, they
		2666	might get executed before other C<ev_check> watchers did their job. As
		2667	C<ev_check> watchers are often used to embed other (non-libev) event
		2668	loops those other event loops might be in an unusable state until their
		2669	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2670	others).
		2671
		2672	=head3 Watcher-Specific Functions and Data Members
699		2673
700	=over 4	2674	=over 4
701		2675
702	=item ev_prepare_init (ev_prepare *, callback)	2676	=item ev_prepare_init (ev_prepare *, callback)
703		2677
704	=item ev_check_init (ev_check *, callback)	2678	=item ev_check_init (ev_check *, callback)
705		2679
706	Initialises and configures the prepare or check watcher - they have no	2680	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>	2681	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.	2682	macros, but using them is utterly, utterly, utterly and completely
		2683	pointless.
709		2684
710	=back	2685	=back
711		2686
		2687	=head3 Examples
		2688
		2689	There are a number of principal ways to embed other event loops or modules
		2690	into libev. Here are some ideas on how to include libadns into libev
		2691	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2692	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2693	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2694	Glib event loop).
		2695
		2696	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2697	and in a check watcher, destroy them and call into libadns. What follows
		2698	is pseudo-code only of course. This requires you to either use a low
		2699	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2700	the callbacks for the IO/timeout watchers might not have been called yet.
		2701
		2702	static ev_io iow [nfd];
		2703	static ev_timer tw;
		2704
		2705	static void
		2706	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2707	{
		2708	}
		2709
		2710	// create io watchers for each fd and a timer before blocking
		2711	static void
		2712	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2713	{
		2714	int timeout = 3600000;
		2715	struct pollfd fds [nfd];
		2716	// actual code will need to loop here and realloc etc.
		2717	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2718
		2719	/* the callback is illegal, but won't be called as we stop during check */
		2720	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2721	ev_timer_start (loop, &tw);
		2722
		2723	// create one ev_io per pollfd
		2724	for (int i = 0; i < nfd; ++i)
		2725	{
		2726	ev_io_init (iow + i, io_cb, fds [i].fd,
		2727	((fds [i].events & POLLIN ? EV_READ : 0)
		2728	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2729
		2730	fds [i].revents = 0;
		2731	ev_io_start (loop, iow + i);
		2732	}
		2733	}
		2734
		2735	// stop all watchers after blocking
		2736	static void
		2737	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2738	{
		2739	ev_timer_stop (loop, &tw);
		2740
		2741	for (int i = 0; i < nfd; ++i)
		2742	{
		2743	// set the relevant poll flags
		2744	// could also call adns_processreadable etc. here
		2745	struct pollfd *fd = fds + i;
		2746	int revents = ev_clear_pending (iow + i);
		2747	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2748	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2749
		2750	// now stop the watcher
		2751	ev_io_stop (loop, iow + i);
		2752	}
		2753
		2754	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2755	}
		2756
		2757	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2758	in the prepare watcher and would dispose of the check watcher.
		2759
		2760	Method 3: If the module to be embedded supports explicit event
		2761	notification (libadns does), you can also make use of the actual watcher
		2762	callbacks, and only destroy/create the watchers in the prepare watcher.
		2763
		2764	static void
		2765	timer_cb (EV_P_ ev_timer *w, int revents)
		2766	{
		2767	adns_state ads = (adns_state)w->data;
		2768	update_now (EV_A);
		2769
		2770	adns_processtimeouts (ads, &tv_now);
		2771	}
		2772
		2773	static void
		2774	io_cb (EV_P_ ev_io *w, int revents)
		2775	{
		2776	adns_state ads = (adns_state)w->data;
		2777	update_now (EV_A);
		2778
		2779	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2780	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2781	}
		2782
		2783	// do not ever call adns_afterpoll
		2784
		2785	Method 4: Do not use a prepare or check watcher because the module you
		2786	want to embed is not flexible enough to support it. Instead, you can
		2787	override their poll function. The drawback with this solution is that the
		2788	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2789	this approach, effectively embedding EV as a client into the horrible
		2790	libglib event loop.
		2791
		2792	static gint
		2793	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2794	{
		2795	int got_events = 0;
		2796
		2797	for (n = 0; n < nfds; ++n)
		2798	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2799
		2800	if (timeout >= 0)
		2801	// create/start timer
		2802
		2803	// poll
		2804	ev_loop (EV_A_ 0);
		2805
		2806	// stop timer again
		2807	if (timeout >= 0)
		2808	ev_timer_stop (EV_A_ &to);
		2809
		2810	// stop io watchers again - their callbacks should have set
		2811	for (n = 0; n < nfds; ++n)
		2812	ev_io_stop (EV_A_ iow [n]);
		2813
		2814	return got_events;
		2815	}
		2816
		2817
		2818	=head2 C<ev_embed> - when one backend isn't enough...
		2819
		2820	This is a rather advanced watcher type that lets you embed one event loop
		2821	into another (currently only C<ev_io> events are supported in the embedded
		2822	loop, other types of watchers might be handled in a delayed or incorrect
		2823	fashion and must not be used).
		2824
		2825	There are primarily two reasons you would want that: work around bugs and
		2826	prioritise I/O.
		2827
		2828	As an example for a bug workaround, the kqueue backend might only support
		2829	sockets on some platform, so it is unusable as generic backend, but you
		2830	still want to make use of it because you have many sockets and it scales
		2831	so nicely. In this case, you would create a kqueue-based loop and embed
		2832	it into your default loop (which might use e.g. poll). Overall operation
		2833	will be a bit slower because first libev has to call C<poll> and then
		2834	C<kevent>, but at least you can use both mechanisms for what they are
		2835	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2836
		2837	As for prioritising I/O: under rare circumstances you have the case where
		2838	some fds have to be watched and handled very quickly (with low latency),
		2839	and even priorities and idle watchers might have too much overhead. In
		2840	this case you would put all the high priority stuff in one loop and all
		2841	the rest in a second one, and embed the second one in the first.
		2842
		2843	As long as the watcher is active, the callback will be invoked every
		2844	time there might be events pending in the embedded loop. The callback
		2845	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2846	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2847	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2848	to give the embedded loop strictly lower priority for example).
		2849
		2850	You can also set the callback to C<0>, in which case the embed watcher
		2851	will automatically execute the embedded loop sweep whenever necessary.
		2852
		2853	Fork detection will be handled transparently while the C<ev_embed> watcher
		2854	is active, i.e., the embedded loop will automatically be forked when the
		2855	embedding loop forks. In other cases, the user is responsible for calling
		2856	C<ev_loop_fork> on the embedded loop.
		2857
		2858	Unfortunately, not all backends are embeddable: only the ones returned by
		2859	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2860	portable one.
		2861
		2862	So when you want to use this feature you will always have to be prepared
		2863	that you cannot get an embeddable loop. The recommended way to get around
		2864	this is to have a separate variables for your embeddable loop, try to
		2865	create it, and if that fails, use the normal loop for everything.
		2866
		2867	=head3 C<ev_embed> and fork
		2868
		2869	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2870	automatically be applied to the embedded loop as well, so no special
		2871	fork handling is required in that case. When the watcher is not running,
		2872	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2873	as applicable.
		2874
		2875	=head3 Watcher-Specific Functions and Data Members
		2876
		2877	=over 4
		2878
		2879	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2880
		2881	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2882
		2883	Configures the watcher to embed the given loop, which must be
		2884	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2885	invoked automatically, otherwise it is the responsibility of the callback
		2886	to invoke it (it will continue to be called until the sweep has been done,
		2887	if you do not want that, you need to temporarily stop the embed watcher).
		2888
		2889	=item ev_embed_sweep (loop, ev_embed *)
		2890
		2891	Make a single, non-blocking sweep over the embedded loop. This works
		2892	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2893	appropriate way for embedded loops.
		2894
		2895	=item struct ev_loop *other [read-only]
		2896
		2897	The embedded event loop.
		2898
		2899	=back
		2900
		2901	=head3 Examples
		2902
		2903	Example: Try to get an embeddable event loop and embed it into the default
		2904	event loop. If that is not possible, use the default loop. The default
		2905	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2906	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2907	used).
		2908
		2909	struct ev_loop *loop_hi = ev_default_init (0);
		2910	struct ev_loop *loop_lo = 0;
		2911	ev_embed embed;
		2912
		2913	// see if there is a chance of getting one that works
		2914	// (remember that a flags value of 0 means autodetection)
		2915	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2916	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2917	: 0;
		2918
		2919	// if we got one, then embed it, otherwise default to loop_hi
		2920	if (loop_lo)
		2921	{
		2922	ev_embed_init (&embed, 0, loop_lo);
		2923	ev_embed_start (loop_hi, &embed);
		2924	}
		2925	else
		2926	loop_lo = loop_hi;
		2927
		2928	Example: Check if kqueue is available but not recommended and create
		2929	a kqueue backend for use with sockets (which usually work with any
		2930	kqueue implementation). Store the kqueue/socket-only event loop in
		2931	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2932
		2933	struct ev_loop *loop = ev_default_init (0);
		2934	struct ev_loop *loop_socket = 0;
		2935	ev_embed embed;
		2936
		2937	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2938	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2939	{
		2940	ev_embed_init (&embed, 0, loop_socket);
		2941	ev_embed_start (loop, &embed);
		2942	}
		2943
		2944	if (!loop_socket)
		2945	loop_socket = loop;
		2946
		2947	// now use loop_socket for all sockets, and loop for everything else
		2948
		2949
		2950	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2951
		2952	Fork watchers are called when a C<fork ()> was detected (usually because
		2953	whoever is a good citizen cared to tell libev about it by calling
		2954	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2955	event loop blocks next and before C<ev_check> watchers are being called,
		2956	and only in the child after the fork. If whoever good citizen calling
		2957	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2958	handlers will be invoked, too, of course.
		2959
		2960	=head3 The special problem of life after fork - how is it possible?
		2961
		2962	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2963	up/change the process environment, followed by a call to C<exec()>. This
		2964	sequence should be handled by libev without any problems.
		2965
		2966	This changes when the application actually wants to do event handling
		2967	in the child, or both parent in child, in effect "continuing" after the
		2968	fork.
		2969
		2970	The default mode of operation (for libev, with application help to detect
		2971	forks) is to duplicate all the state in the child, as would be expected
		2972	when I<either> the parent I<or> the child process continues.
		2973
		2974	When both processes want to continue using libev, then this is usually the
		2975	wrong result. In that case, usually one process (typically the parent) is
		2976	supposed to continue with all watchers in place as before, while the other
		2977	process typically wants to start fresh, i.e. without any active watchers.
		2978
		2979	The cleanest and most efficient way to achieve that with libev is to
		2980	simply create a new event loop, which of course will be "empty", and
		2981	use that for new watchers. This has the advantage of not touching more
		2982	memory than necessary, and thus avoiding the copy-on-write, and the
		2983	disadvantage of having to use multiple event loops (which do not support
		2984	signal watchers).
		2985
		2986	When this is not possible, or you want to use the default loop for
		2987	other reasons, then in the process that wants to start "fresh", call
		2988	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2989	the default loop will "orphan" (not stop) all registered watchers, so you
		2990	have to be careful not to execute code that modifies those watchers. Note
		2991	also that in that case, you have to re-register any signal watchers.
		2992
		2993	=head3 Watcher-Specific Functions and Data Members
		2994
		2995	=over 4
		2996
		2997	=item ev_fork_init (ev_signal *, callback)
		2998
		2999	Initialises and configures the fork watcher - it has no parameters of any
		3000	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3001	believe me.
		3002
		3003	=back
		3004
		3005
		3006	=head2 C<ev_async> - how to wake up another event loop
		3007
		3008	In general, you cannot use an C<ev_loop> from multiple threads or other
		3009	asynchronous sources such as signal handlers (as opposed to multiple event
		3010	loops - those are of course safe to use in different threads).
		3011
		3012	Sometimes, however, you need to wake up another event loop you do not
		3013	control, for example because it belongs to another thread. This is what
		3014	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		3015	can signal it by calling C<ev_async_send>, which is thread- and signal
		3016	safe.
		3017
		3018	This functionality is very similar to C<ev_signal> watchers, as signals,
		3019	too, are asynchronous in nature, and signals, too, will be compressed
		3020	(i.e. the number of callback invocations may be less than the number of
		3021	C<ev_async_sent> calls).
		3022
		3023	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		3024	just the default loop.
		3025
		3026	=head3 Queueing
		3027
		3028	C<ev_async> does not support queueing of data in any way. The reason
		3029	is that the author does not know of a simple (or any) algorithm for a
		3030	multiple-writer-single-reader queue that works in all cases and doesn't
		3031	need elaborate support such as pthreads or unportable memory access
		3032	semantics.
		3033
		3034	That means that if you want to queue data, you have to provide your own
		3035	queue. But at least I can tell you how to implement locking around your
		3036	queue:
		3037
		3038	=over 4
		3039
		3040	=item queueing from a signal handler context
		3041
		3042	To implement race-free queueing, you simply add to the queue in the signal
		3043	handler but you block the signal handler in the watcher callback. Here is
		3044	an example that does that for some fictitious SIGUSR1 handler:
		3045
		3046	static ev_async mysig;
		3047
		3048	static void
		3049	sigusr1_handler (void)
		3050	{
		3051	sometype data;
		3052
		3053	// no locking etc.
		3054	queue_put (data);
		3055	ev_async_send (EV_DEFAULT_ &mysig);
		3056	}
		3057
		3058	static void
		3059	mysig_cb (EV_P_ ev_async *w, int revents)
		3060	{
		3061	sometype data;
		3062	sigset_t block, prev;
		3063
		3064	sigemptyset (&block);
		3065	sigaddset (&block, SIGUSR1);
		3066	sigprocmask (SIG_BLOCK, &block, &prev);
		3067
		3068	while (queue_get (&data))
		3069	process (data);
		3070
		3071	if (sigismember (&prev, SIGUSR1)
		3072	sigprocmask (SIG_UNBLOCK, &block, 0);
		3073	}
		3074
		3075	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3076	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3077	either...).
		3078
		3079	=item queueing from a thread context
		3080
		3081	The strategy for threads is different, as you cannot (easily) block
		3082	threads but you can easily preempt them, so to queue safely you need to
		3083	employ a traditional mutex lock, such as in this pthread example:
		3084
		3085	static ev_async mysig;
		3086	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3087
		3088	static void
		3089	otherthread (void)
		3090	{
		3091	// only need to lock the actual queueing operation
		3092	pthread_mutex_lock (&mymutex);
		3093	queue_put (data);
		3094	pthread_mutex_unlock (&mymutex);
		3095
		3096	ev_async_send (EV_DEFAULT_ &mysig);
		3097	}
		3098
		3099	static void
		3100	mysig_cb (EV_P_ ev_async *w, int revents)
		3101	{
		3102	pthread_mutex_lock (&mymutex);
		3103
		3104	while (queue_get (&data))
		3105	process (data);
		3106
		3107	pthread_mutex_unlock (&mymutex);
		3108	}
		3109
		3110	=back
		3111
		3112
		3113	=head3 Watcher-Specific Functions and Data Members
		3114
		3115	=over 4
		3116
		3117	=item ev_async_init (ev_async *, callback)
		3118
		3119	Initialises and configures the async watcher - it has no parameters of any
		3120	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3121	trust me.
		3122
		3123	=item ev_async_send (loop, ev_async *)
		3124
		3125	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3126	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3127	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3128	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3129	section below on what exactly this means).
		3130
		3131	Note that, as with other watchers in libev, multiple events might get
		3132	compressed into a single callback invocation (another way to look at this
		3133	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3134	reset when the event loop detects that).
		3135
		3136	This call incurs the overhead of a system call only once per event loop
		3137	iteration, so while the overhead might be noticeable, it doesn't apply to
		3138	repeated calls to C<ev_async_send> for the same event loop.
		3139
		3140	=item bool = ev_async_pending (ev_async *)
		3141
		3142	Returns a non-zero value when C<ev_async_send> has been called on the
		3143	watcher but the event has not yet been processed (or even noted) by the
		3144	event loop.
		3145
		3146	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3147	the loop iterates next and checks for the watcher to have become active,
		3148	it will reset the flag again. C<ev_async_pending> can be used to very
		3149	quickly check whether invoking the loop might be a good idea.
		3150
		3151	Not that this does I<not> check whether the watcher itself is pending,
		3152	only whether it has been requested to make this watcher pending: there
		3153	is a time window between the event loop checking and resetting the async
		3154	notification, and the callback being invoked.
		3155
		3156	=back
		3157
		3158
712	=head1 OTHER FUNCTIONS	3159	=head1 OTHER FUNCTIONS
713		3160
714	There are some other functions of possible interest. Described. Here. Now.	3161	There are some other functions of possible interest. Described. Here. Now.
715		3162
716	=over 4	3163	=over 4
717		3164
718	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3165	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
719		3166
720	This function combines a simple timer and an I/O watcher, calls your	3167	This function combines a simple timer and an I/O watcher, calls your
721	callback on whichever event happens first and automatically stop both	3168	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	3169	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	3170	or timeout without having to allocate/configure/start/stop/free one or
724	more watchers yourself.	3171	more watchers yourself.
725		3172
726	If C<fd> is less than 0, then no I/O watcher will be started and events	3173	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	3174	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
728	C<events> set will be craeted and started.	3175	the given C<fd> and C<events> set will be created and started.
729		3176
730	If C<timeout> is less than 0, then no timeout watcher will be	3177	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	3178	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	3179	repeat = 0) will be started. C<0> is a valid timeout.
733	dubious value.
734		3180
735	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3181	The callback has the type C<void (*cb)(int revents, void *arg)> and is
736	passed an events set like normal event callbacks (with a combination of	3182	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>	3183	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
738	value passed to C<ev_once>:	3184	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3185	a timeout and an io event at the same time - you probably should give io
		3186	events precedence.
739		3187
		3188	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3189
740	static void stdin_ready (int revents, void *arg)	3190	static void stdin_ready (int revents, void *arg)
741	{	3191	{
742	if (revents & EV_TIMEOUT)
743	/* doh, nothing entered */;
744	else if (revents & EV_READ)	3192	if (revents & EV_READ)
745	/* stdin might have data for us, joy! */;	3193	/* stdin might have data for us, joy! */;
		3194	else if (revents & EV_TIMER)
		3195	/* doh, nothing entered */;
746	}	3196	}
747		3197
748	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3198	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
749
750	=item ev_feed_event (loop, watcher, int events)
751
752	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
754	initialised but not necessarily started event watcher).
755		3199
756	=item ev_feed_fd_event (loop, int fd, int revents)	3200	=item ev_feed_fd_event (loop, int fd, int revents)
757		3201
758	Feed an event on the given fd, as if a file descriptor backend detected	3202	Feed an event on the given fd, as if a file descriptor backend detected
759	the given events it.	3203	the given events it.
760		3204
761	=item ev_feed_signal_event (loop, int signum)	3205	=item ev_feed_signal_event (loop, int signum)
762		3206
763	Feed an event as if the given signal occured (loop must be the default loop!).	3207	Feed an event as if the given signal occurred (C<loop> must be the default
		3208	loop!).
764		3209
765	=back	3210	=back
766		3211
		3212
		3213	=head1 LIBEVENT EMULATION
		3214
		3215	Libev offers a compatibility emulation layer for libevent. It cannot
		3216	emulate the internals of libevent, so here are some usage hints:
		3217
		3218	=over 4
		3219
		3220	=item * Use it by including <event.h>, as usual.
		3221
		3222	=item * The following members are fully supported: ev_base, ev_callback,
		3223	ev_arg, ev_fd, ev_res, ev_events.
		3224
		3225	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		3226	maintained by libev, it does not work exactly the same way as in libevent (consider
		3227	it a private API).
		3228
		3229	=item * Priorities are not currently supported. Initialising priorities
		3230	will fail and all watchers will have the same priority, even though there
		3231	is an ev_pri field.
		3232
		3233	=item * In libevent, the last base created gets the signals, in libev, the
		3234	first base created (== the default loop) gets the signals.
		3235
		3236	=item * Other members are not supported.
		3237
		3238	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		3239	to use the libev header file and library.
		3240
		3241	=back
		3242
		3243	=head1 C++ SUPPORT
		3244
		3245	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3246	you to use some convenience methods to start/stop watchers and also change
		3247	the callback model to a model using method callbacks on objects.
		3248
		3249	To use it,
		3250
		3251	#include <ev++.h>
		3252
		3253	This automatically includes F<ev.h> and puts all of its definitions (many
		3254	of them macros) into the global namespace. All C++ specific things are
		3255	put into the C<ev> namespace. It should support all the same embedding
		3256	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3257
		3258	Care has been taken to keep the overhead low. The only data member the C++
		3259	classes add (compared to plain C-style watchers) is the event loop pointer
		3260	that the watcher is associated with (or no additional members at all if
		3261	you disable C<EV_MULTIPLICITY> when embedding libev).
		3262
		3263	Currently, functions, and static and non-static member functions can be
		3264	used as callbacks. Other types should be easy to add as long as they only
		3265	need one additional pointer for context. If you need support for other
		3266	types of functors please contact the author (preferably after implementing
		3267	it).
		3268
		3269	Here is a list of things available in the C<ev> namespace:
		3270
		3271	=over 4
		3272
		3273	=item C<ev::READ>, C<ev::WRITE> etc.
		3274
		3275	These are just enum values with the same values as the C<EV_READ> etc.
		3276	macros from F<ev.h>.
		3277
		3278	=item C<ev::tstamp>, C<ev::now>
		3279
		3280	Aliases to the same types/functions as with the C<ev_> prefix.
		3281
		3282	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3283
		3284	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3285	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3286	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3287	defines by many implementations.
		3288
		3289	All of those classes have these methods:
		3290
		3291	=over 4
		3292
		3293	=item ev::TYPE::TYPE ()
		3294
		3295	=item ev::TYPE::TYPE (loop)
		3296
		3297	=item ev::TYPE::~TYPE
		3298
		3299	The constructor (optionally) takes an event loop to associate the watcher
		3300	with. If it is omitted, it will use C<EV_DEFAULT>.
		3301
		3302	The constructor calls C<ev_init> for you, which means you have to call the
		3303	C<set> method before starting it.
		3304
		3305	It will not set a callback, however: You have to call the templated C<set>
		3306	method to set a callback before you can start the watcher.
		3307
		3308	(The reason why you have to use a method is a limitation in C++ which does
		3309	not allow explicit template arguments for constructors).
		3310
		3311	The destructor automatically stops the watcher if it is active.
		3312
		3313	=item w->set<class, &class::method> (object *)
		3314
		3315	This method sets the callback method to call. The method has to have a
		3316	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3317	first argument and the C<revents> as second. The object must be given as
		3318	parameter and is stored in the C<data> member of the watcher.
		3319
		3320	This method synthesizes efficient thunking code to call your method from
		3321	the C callback that libev requires. If your compiler can inline your
		3322	callback (i.e. it is visible to it at the place of the C<set> call and
		3323	your compiler is good :), then the method will be fully inlined into the
		3324	thunking function, making it as fast as a direct C callback.
		3325
		3326	Example: simple class declaration and watcher initialisation
		3327
		3328	struct myclass
		3329	{
		3330	void io_cb (ev::io &w, int revents) { }
		3331	}
		3332
		3333	myclass obj;
		3334	ev::io iow;
		3335	iow.set <myclass, &myclass::io_cb> (&obj);
		3336
		3337	=item w->set (object *)
		3338
		3339	This is an B<experimental> feature that might go away in a future version.
		3340
		3341	This is a variation of a method callback - leaving out the method to call
		3342	will default the method to C<operator ()>, which makes it possible to use
		3343	functor objects without having to manually specify the C<operator ()> all
		3344	the time. Incidentally, you can then also leave out the template argument
		3345	list.
		3346
		3347	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3348	int revents)>.
		3349
		3350	See the method-C<set> above for more details.
		3351
		3352	Example: use a functor object as callback.
		3353
		3354	struct myfunctor
		3355	{
		3356	void operator() (ev::io &w, int revents)
		3357	{
		3358	...
		3359	}
		3360	}
		3361
		3362	myfunctor f;
		3363
		3364	ev::io w;
		3365	w.set (&f);
		3366
		3367	=item w->set<function> (void *data = 0)
		3368
		3369	Also sets a callback, but uses a static method or plain function as
		3370	callback. The optional C<data> argument will be stored in the watcher's
		3371	C<data> member and is free for you to use.
		3372
		3373	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3374
		3375	See the method-C<set> above for more details.
		3376
		3377	Example: Use a plain function as callback.
		3378
		3379	static void io_cb (ev::io &w, int revents) { }
		3380	iow.set <io_cb> ();
		3381
		3382	=item w->set (loop)
		3383
		3384	Associates a different C<struct ev_loop> with this watcher. You can only
		3385	do this when the watcher is inactive (and not pending either).
		3386
		3387	=item w->set ([arguments])
		3388
		3389	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		3390	called at least once. Unlike the C counterpart, an active watcher gets
		3391	automatically stopped and restarted when reconfiguring it with this
		3392	method.
		3393
		3394	=item w->start ()
		3395
		3396	Starts the watcher. Note that there is no C<loop> argument, as the
		3397	constructor already stores the event loop.
		3398
		3399	=item w->stop ()
		3400
		3401	Stops the watcher if it is active. Again, no C<loop> argument.
		3402
		3403	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3404
		3405	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3406	C<ev_TYPE_again> function.
		3407
		3408	=item w->sweep () (C<ev::embed> only)
		3409
		3410	Invokes C<ev_embed_sweep>.
		3411
		3412	=item w->update () (C<ev::stat> only)
		3413
		3414	Invokes C<ev_stat_stat>.
		3415
		3416	=back
		3417
		3418	=back
		3419
		3420	Example: Define a class with an IO and idle watcher, start one of them in
		3421	the constructor.
		3422
		3423	class myclass
		3424	{
		3425	ev::io io ; void io_cb (ev::io &w, int revents);
		3426	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3427
		3428	myclass (int fd)
		3429	{
		3430	io .set <myclass, &myclass::io_cb > (this);
		3431	idle.set <myclass, &myclass::idle_cb> (this);
		3432
		3433	io.start (fd, ev::READ);
		3434	}
		3435	};
		3436
		3437
		3438	=head1 OTHER LANGUAGE BINDINGS
		3439
		3440	Libev does not offer other language bindings itself, but bindings for a
		3441	number of languages exist in the form of third-party packages. If you know
		3442	any interesting language binding in addition to the ones listed here, drop
		3443	me a note.
		3444
		3445	=over 4
		3446
		3447	=item Perl
		3448
		3449	The EV module implements the full libev API and is actually used to test
		3450	libev. EV is developed together with libev. Apart from the EV core module,
		3451	there are additional modules that implement libev-compatible interfaces
		3452	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3453	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3454	and C<EV::Glib>).
		3455
		3456	It can be found and installed via CPAN, its homepage is at
		3457	L<http://software.schmorp.de/pkg/EV>.
		3458
		3459	=item Python
		3460
		3461	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3462	seems to be quite complete and well-documented.
		3463
		3464	=item Ruby
		3465
		3466	Tony Arcieri has written a ruby extension that offers access to a subset
		3467	of the libev API and adds file handle abstractions, asynchronous DNS and
		3468	more on top of it. It can be found via gem servers. Its homepage is at
		3469	L<http://rev.rubyforge.org/>.
		3470
		3471	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3472	makes rev work even on mingw.
		3473
		3474	=item Haskell
		3475
		3476	A haskell binding to libev is available at
		3477	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3478
		3479	=item D
		3480
		3481	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3482	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3483
		3484	=item Ocaml
		3485
		3486	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3487	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3488
		3489	=item Lua
		3490
		3491	Brian Maher has written a partial interface to libev for lua (at the
		3492	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		3493	L<http://github.com/brimworks/lua-ev>.
		3494
		3495	=back
		3496
		3497
		3498	=head1 MACRO MAGIC
		3499
		3500	Libev can be compiled with a variety of options, the most fundamental
		3501	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3502	functions and callbacks have an initial C<struct ev_loop *> argument.
		3503
		3504	To make it easier to write programs that cope with either variant, the
		3505	following macros are defined:
		3506
		3507	=over 4
		3508
		3509	=item C<EV_A>, C<EV_A_>
		3510
		3511	This provides the loop I<argument> for functions, if one is required ("ev
		3512	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3513	C<EV_A_> is used when other arguments are following. Example:
		3514
		3515	ev_unref (EV_A);
		3516	ev_timer_add (EV_A_ watcher);
		3517	ev_loop (EV_A_ 0);
		3518
		3519	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3520	which is often provided by the following macro.
		3521
		3522	=item C<EV_P>, C<EV_P_>
		3523
		3524	This provides the loop I<parameter> for functions, if one is required ("ev
		3525	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3526	C<EV_P_> is used when other parameters are following. Example:
		3527
		3528	// this is how ev_unref is being declared
		3529	static void ev_unref (EV_P);
		3530
		3531	// this is how you can declare your typical callback
		3532	static void cb (EV_P_ ev_timer *w, int revents)
		3533
		3534	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3535	suitable for use with C<EV_A>.
		3536
		3537	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3538
		3539	Similar to the other two macros, this gives you the value of the default
		3540	loop, if multiple loops are supported ("ev loop default").
		3541
		3542	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3543
		3544	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3545	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3546	is undefined when the default loop has not been initialised by a previous
		3547	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3548
		3549	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3550	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3551
		3552	=back
		3553
		3554	Example: Declare and initialise a check watcher, utilising the above
		3555	macros so it will work regardless of whether multiple loops are supported
		3556	or not.
		3557
		3558	static void
		3559	check_cb (EV_P_ ev_timer *w, int revents)
		3560	{
		3561	ev_check_stop (EV_A_ w);
		3562	}
		3563
		3564	ev_check check;
		3565	ev_check_init (&check, check_cb);
		3566	ev_check_start (EV_DEFAULT_ &check);
		3567	ev_loop (EV_DEFAULT_ 0);
		3568
		3569	=head1 EMBEDDING
		3570
		3571	Libev can (and often is) directly embedded into host
		3572	applications. Examples of applications that embed it include the Deliantra
		3573	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3574	and rxvt-unicode.
		3575
		3576	The goal is to enable you to just copy the necessary files into your
		3577	source directory without having to change even a single line in them, so
		3578	you can easily upgrade by simply copying (or having a checked-out copy of
		3579	libev somewhere in your source tree).
		3580
		3581	=head2 FILESETS
		3582
		3583	Depending on what features you need you need to include one or more sets of files
		3584	in your application.
		3585
		3586	=head3 CORE EVENT LOOP
		3587
		3588	To include only the libev core (all the C<ev_*> functions), with manual
		3589	configuration (no autoconf):
		3590
		3591	#define EV_STANDALONE 1
		3592	#include "ev.c"
		3593
		3594	This will automatically include F<ev.h>, too, and should be done in a
		3595	single C source file only to provide the function implementations. To use
		3596	it, do the same for F<ev.h> in all files wishing to use this API (best
		3597	done by writing a wrapper around F<ev.h> that you can include instead and
		3598	where you can put other configuration options):
		3599
		3600	#define EV_STANDALONE 1
		3601	#include "ev.h"
		3602
		3603	Both header files and implementation files can be compiled with a C++
		3604	compiler (at least, that's a stated goal, and breakage will be treated
		3605	as a bug).
		3606
		3607	You need the following files in your source tree, or in a directory
		3608	in your include path (e.g. in libev/ when using -Ilibev):
		3609
		3610	ev.h
		3611	ev.c
		3612	ev_vars.h
		3613	ev_wrap.h
		3614
		3615	ev_win32.c required on win32 platforms only
		3616
		3617	ev_select.c only when select backend is enabled (which is enabled by default)
		3618	ev_poll.c only when poll backend is enabled (disabled by default)
		3619	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3620	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3621	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3622
		3623	F<ev.c> includes the backend files directly when enabled, so you only need
		3624	to compile this single file.
		3625
		3626	=head3 LIBEVENT COMPATIBILITY API
		3627
		3628	To include the libevent compatibility API, also include:
		3629
		3630	#include "event.c"
		3631
		3632	in the file including F<ev.c>, and:
		3633
		3634	#include "event.h"
		3635
		3636	in the files that want to use the libevent API. This also includes F<ev.h>.
		3637
		3638	You need the following additional files for this:
		3639
		3640	event.h
		3641	event.c
		3642
		3643	=head3 AUTOCONF SUPPORT
		3644
		3645	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3646	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3647	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3648	include F<config.h> and configure itself accordingly.
		3649
		3650	For this of course you need the m4 file:
		3651
		3652	libev.m4
		3653
		3654	=head2 PREPROCESSOR SYMBOLS/MACROS
		3655
		3656	Libev can be configured via a variety of preprocessor symbols you have to
		3657	define before including (or compiling) any of its files. The default in
		3658	the absence of autoconf is documented for every option.
		3659
		3660	Symbols marked with "(h)" do not change the ABI, and can have different
		3661	values when compiling libev vs. including F<ev.h>, so it is permissible
		3662	to redefine them before including F<ev.h> without breakign compatibility
		3663	to a compiled library. All other symbols change the ABI, which means all
		3664	users of libev and the libev code itself must be compiled with compatible
		3665	settings.
		3666
		3667	=over 4
		3668
		3669	=item EV_STANDALONE (h)
		3670
		3671	Must always be C<1> if you do not use autoconf configuration, which
		3672	keeps libev from including F<config.h>, and it also defines dummy
		3673	implementations for some libevent functions (such as logging, which is not
		3674	supported). It will also not define any of the structs usually found in
		3675	F<event.h> that are not directly supported by the libev core alone.
		3676
		3677	In standalone mode, libev will still try to automatically deduce the
		3678	configuration, but has to be more conservative.
		3679
		3680	=item EV_USE_MONOTONIC
		3681
		3682	If defined to be C<1>, libev will try to detect the availability of the
		3683	monotonic clock option at both compile time and runtime. Otherwise no
		3684	use of the monotonic clock option will be attempted. If you enable this,
		3685	you usually have to link against librt or something similar. Enabling it
		3686	when the functionality isn't available is safe, though, although you have
		3687	to make sure you link against any libraries where the C<clock_gettime>
		3688	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3689
		3690	=item EV_USE_REALTIME
		3691
		3692	If defined to be C<1>, libev will try to detect the availability of the
		3693	real-time clock option at compile time (and assume its availability
		3694	at runtime if successful). Otherwise no use of the real-time clock
		3695	option will be attempted. This effectively replaces C<gettimeofday>
		3696	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3697	correctness. See the note about libraries in the description of
		3698	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3699	C<EV_USE_CLOCK_SYSCALL>.
		3700
		3701	=item EV_USE_CLOCK_SYSCALL
		3702
		3703	If defined to be C<1>, libev will try to use a direct syscall instead
		3704	of calling the system-provided C<clock_gettime> function. This option
		3705	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3706	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3707	programs needlessly. Using a direct syscall is slightly slower (in
		3708	theory), because no optimised vdso implementation can be used, but avoids
		3709	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3710	higher, as it simplifies linking (no need for C<-lrt>).
		3711
		3712	=item EV_USE_NANOSLEEP
		3713
		3714	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3715	and will use it for delays. Otherwise it will use C<select ()>.
		3716
		3717	=item EV_USE_EVENTFD
		3718
		3719	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3720	available and will probe for kernel support at runtime. This will improve
		3721	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3722	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3723	2.7 or newer, otherwise disabled.
		3724
		3725	=item EV_USE_SELECT
		3726
		3727	If undefined or defined to be C<1>, libev will compile in support for the
		3728	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3729	other method takes over, select will be it. Otherwise the select backend
		3730	will not be compiled in.
		3731
		3732	=item EV_SELECT_USE_FD_SET
		3733
		3734	If defined to C<1>, then the select backend will use the system C<fd_set>
		3735	structure. This is useful if libev doesn't compile due to a missing
		3736	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3737	on exotic systems. This usually limits the range of file descriptors to
		3738	some low limit such as 1024 or might have other limitations (winsocket
		3739	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3740	configures the maximum size of the C<fd_set>.
		3741
		3742	=item EV_SELECT_IS_WINSOCKET
		3743
		3744	When defined to C<1>, the select backend will assume that
		3745	select/socket/connect etc. don't understand file descriptors but
		3746	wants osf handles on win32 (this is the case when the select to
		3747	be used is the winsock select). This means that it will call
		3748	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3749	it is assumed that all these functions actually work on fds, even
		3750	on win32. Should not be defined on non-win32 platforms.
		3751
		3752	=item EV_FD_TO_WIN32_HANDLE(fd)
		3753
		3754	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3755	file descriptors to socket handles. When not defining this symbol (the
		3756	default), then libev will call C<_get_osfhandle>, which is usually
		3757	correct. In some cases, programs use their own file descriptor management,
		3758	in which case they can provide this function to map fds to socket handles.
		3759
		3760	=item EV_WIN32_HANDLE_TO_FD(handle)
		3761
		3762	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3763	using the standard C<_open_osfhandle> function. For programs implementing
		3764	their own fd to handle mapping, overwriting this function makes it easier
		3765	to do so. This can be done by defining this macro to an appropriate value.
		3766
		3767	=item EV_WIN32_CLOSE_FD(fd)
		3768
		3769	If programs implement their own fd to handle mapping on win32, then this
		3770	macro can be used to override the C<close> function, useful to unregister
		3771	file descriptors again. Note that the replacement function has to close
		3772	the underlying OS handle.
		3773
		3774	=item EV_USE_POLL
		3775
		3776	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3777	backend. Otherwise it will be enabled on non-win32 platforms. It
		3778	takes precedence over select.
		3779
		3780	=item EV_USE_EPOLL
		3781
		3782	If defined to be C<1>, libev will compile in support for the Linux
		3783	C<epoll>(7) backend. Its availability will be detected at runtime,
		3784	otherwise another method will be used as fallback. This is the preferred
		3785	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3786	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3787
		3788	=item EV_USE_KQUEUE
		3789
		3790	If defined to be C<1>, libev will compile in support for the BSD style
		3791	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3792	otherwise another method will be used as fallback. This is the preferred
		3793	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3794	supports some types of fds correctly (the only platform we found that
		3795	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3796	not be used unless explicitly requested. The best way to use it is to find
		3797	out whether kqueue supports your type of fd properly and use an embedded
		3798	kqueue loop.
		3799
		3800	=item EV_USE_PORT
		3801
		3802	If defined to be C<1>, libev will compile in support for the Solaris
		3803	10 port style backend. Its availability will be detected at runtime,
		3804	otherwise another method will be used as fallback. This is the preferred
		3805	backend for Solaris 10 systems.
		3806
		3807	=item EV_USE_DEVPOLL
		3808
		3809	Reserved for future expansion, works like the USE symbols above.
		3810
		3811	=item EV_USE_INOTIFY
		3812
		3813	If defined to be C<1>, libev will compile in support for the Linux inotify
		3814	interface to speed up C<ev_stat> watchers. Its actual availability will
		3815	be detected at runtime. If undefined, it will be enabled if the headers
		3816	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3817
		3818	=item EV_ATOMIC_T
		3819
		3820	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3821	access is atomic with respect to other threads or signal contexts. No such
		3822	type is easily found in the C language, so you can provide your own type
		3823	that you know is safe for your purposes. It is used both for signal handler "locking"
		3824	as well as for signal and thread safety in C<ev_async> watchers.
		3825
		3826	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3827	(from F<signal.h>), which is usually good enough on most platforms.
		3828
		3829	=item EV_H (h)
		3830
		3831	The name of the F<ev.h> header file used to include it. The default if
		3832	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3833	used to virtually rename the F<ev.h> header file in case of conflicts.
		3834
		3835	=item EV_CONFIG_H (h)
		3836
		3837	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3838	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3839	C<EV_H>, above.
		3840
		3841	=item EV_EVENT_H (h)
		3842
		3843	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3844	of how the F<event.h> header can be found, the default is C<"event.h">.
		3845
		3846	=item EV_PROTOTYPES (h)
		3847
		3848	If defined to be C<0>, then F<ev.h> will not define any function
		3849	prototypes, but still define all the structs and other symbols. This is
		3850	occasionally useful if you want to provide your own wrapper functions
		3851	around libev functions.
		3852
		3853	=item EV_MULTIPLICITY
		3854
		3855	If undefined or defined to C<1>, then all event-loop-specific functions
		3856	will have the C<struct ev_loop *> as first argument, and you can create
		3857	additional independent event loops. Otherwise there will be no support
		3858	for multiple event loops and there is no first event loop pointer
		3859	argument. Instead, all functions act on the single default loop.
		3860
		3861	=item EV_MINPRI
		3862
		3863	=item EV_MAXPRI
		3864
		3865	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3866	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3867	provide for more priorities by overriding those symbols (usually defined
		3868	to be C<-2> and C<2>, respectively).
		3869
		3870	When doing priority-based operations, libev usually has to linearly search
		3871	all the priorities, so having many of them (hundreds) uses a lot of space
		3872	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3873	fine.
		3874
		3875	If your embedding application does not need any priorities, defining these
		3876	both to C<0> will save some memory and CPU.
		3877
		3878	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		3879	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		3880	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		3881
		3882	If undefined or defined to be C<1> (and the platform supports it), then
		3883	the respective watcher type is supported. If defined to be C<0>, then it
		3884	is not. Disabling watcher types mainly saves codesize.
		3885
		3886	=item EV_FEATURES
		3887
		3888	If you need to shave off some kilobytes of code at the expense of some
		3889	speed (but with the full API), you can define this symbol to request
		3890	certain subsets of functionality. The default is to enable all features
		3891	that can be enabled on the platform.
		3892
		3893	A typical way to use this symbol is to define it to C<0> (or to a bitset
		3894	with some broad features you want) and then selectively re-enable
		3895	additional parts you want, for example if you want everything minimal,
		3896	but multiple event loop support, async and child watchers and the poll
		3897	backend, use this:
		3898
		3899	#define EV_FEATURES 0
		3900	#define EV_MULTIPLICITY 1
		3901	#define EV_USE_POLL 1
		3902	#define EV_CHILD_ENABLE 1
		3903	#define EV_ASYNC_ENABLE 1
		3904
		3905	The actual value is a bitset, it can be a combination of the following
		3906	values:
		3907
		3908	=over 4
		3909
		3910	=item C<1> - faster/larger code
		3911
		3912	Use larger code to speed up some operations.
		3913
		3914	Currently this is used to override some inlining decisions (enlarging the roughly
		3915	30% code size on amd64.
		3916
		3917	When optimising for size, use of compiler flags such as C<-Os> with
		3918	gcc recommended, as well as C<-DNDEBUG>, as libev contains a number of
		3919	assertions.
		3920
		3921	=item C<2> - faster/larger data structures
		3922
		3923	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		3924	hash table sizes and so on. This will usually further increase codesize
		3925	and can additionally have an effect on the size of data structures at
		3926	runtime.
		3927
		3928	=item C<4> - full API configuration
		3929
		3930	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		3931	enables multiplicity (C<EV_MULTIPLICITY>=1).
		3932
		3933	=item C<8> - full API
		3934
		3935	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		3936	details on which parts of the API are still available without this
		3937	feature, and do not complain if this subset changes over time.
		3938
		3939	=item C<16> - enable all optional watcher types
		3940
		3941	Enables all optional watcher types. If you want to selectively enable
		3942	only some watcher types other than I/O and timers (e.g. prepare,
		3943	embed, async, child...) you can enable them manually by defining
		3944	C<EV_watchertype_ENABLE> to C<1> instead.
		3945
		3946	=item C<32> - enable all backends
		3947
		3948	This enables all backends - without this feature, you need to enable at
		3949	least one backend manually (C<EV_USE_SELECT> is a good choice).
		3950
		3951	=item C<64> - enable OS-specific "helper" APIs
		3952
		3953	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		3954	default.
		3955
		3956	=back
		3957
		3958	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		3959	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		3960	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		3961	watchers, timers and monotonic clock support.
		3962
		3963	With an intelligent-enough linker (gcc+binutils are intelligent enough
		3964	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		3965	your program might be left out as well - a binary starting a timer and an
		3966	I/O watcher then might come out at only 5Kb.
		3967
		3968	=item EV_AVOID_STDIO
		3969
		3970	If this is set to C<1> at compiletime, then libev will avoid using stdio
		3971	functions (printf, scanf, perror etc.). This will increase the codesize
		3972	somewhat, but if your program doesn't otherwise depend on stdio and your
		3973	libc allows it, this avoids linking in the stdio library which is quite
		3974	big.
		3975
		3976	Note that error messages might become less precise when this option is
		3977	enabled.
		3978
		3979	=item EV_NSIG
		3980
		3981	The highest supported signal number, +1 (or, the number of
		3982	signals): Normally, libev tries to deduce the maximum number of signals
		3983	automatically, but sometimes this fails, in which case it can be
		3984	specified. Also, using a lower number than detected (C<32> should be
		3985	good for about any system in existance) can save some memory, as libev
		3986	statically allocates some 12-24 bytes per signal number.
		3987
		3988	=item EV_PID_HASHSIZE
		3989
		3990	C<ev_child> watchers use a small hash table to distribute workload by
		3991	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		3992	usually more than enough. If you need to manage thousands of children you
		3993	might want to increase this value (I<must> be a power of two).
		3994
		3995	=item EV_INOTIFY_HASHSIZE
		3996
		3997	C<ev_stat> watchers use a small hash table to distribute workload by
		3998	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		3999	disabled), usually more than enough. If you need to manage thousands of
		4000	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4001	power of two).
		4002
		4003	=item EV_USE_4HEAP
		4004
		4005	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4006	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4007	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4008	faster performance with many (thousands) of watchers.
		4009
		4010	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4011	will be C<0>.
		4012
		4013	=item EV_HEAP_CACHE_AT
		4014
		4015	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4016	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4017	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4018	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4019	but avoids random read accesses on heap changes. This improves performance
		4020	noticeably with many (hundreds) of watchers.
		4021
		4022	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4023	will be C<0>.
		4024
		4025	=item EV_VERIFY
		4026
		4027	Controls how much internal verification (see C<ev_loop_verify ()>) will
		4028	be done: If set to C<0>, no internal verification code will be compiled
		4029	in. If set to C<1>, then verification code will be compiled in, but not
		4030	called. If set to C<2>, then the internal verification code will be
		4031	called once per loop, which can slow down libev. If set to C<3>, then the
		4032	verification code will be called very frequently, which will slow down
		4033	libev considerably.
		4034
		4035	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4036	will be C<0>.
		4037
		4038	=item EV_COMMON
		4039
		4040	By default, all watchers have a C<void *data> member. By redefining
		4041	this macro to a something else you can include more and other types of
		4042	members. You have to define it each time you include one of the files,
		4043	though, and it must be identical each time.
		4044
		4045	For example, the perl EV module uses something like this:
		4046
		4047	#define EV_COMMON \
		4048	SV *self; /* contains this struct */ \
		4049	SV *cb_sv, *fh /* note no trailing ";" */
		4050
		4051	=item EV_CB_DECLARE (type)
		4052
		4053	=item EV_CB_INVOKE (watcher, revents)
		4054
		4055	=item ev_set_cb (ev, cb)
		4056
		4057	Can be used to change the callback member declaration in each watcher,
		4058	and the way callbacks are invoked and set. Must expand to a struct member
		4059	definition and a statement, respectively. See the F<ev.h> header file for
		4060	their default definitions. One possible use for overriding these is to
		4061	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		4062	method calls instead of plain function calls in C++.
		4063
		4064	=back
		4065
		4066	=head2 EXPORTED API SYMBOLS
		4067
		4068	If you need to re-export the API (e.g. via a DLL) and you need a list of
		4069	exported symbols, you can use the provided F<Symbol.*> files which list
		4070	all public symbols, one per line:
		4071
		4072	Symbols.ev for libev proper
		4073	Symbols.event for the libevent emulation
		4074
		4075	This can also be used to rename all public symbols to avoid clashes with
		4076	multiple versions of libev linked together (which is obviously bad in
		4077	itself, but sometimes it is inconvenient to avoid this).
		4078
		4079	A sed command like this will create wrapper C<#define>'s that you need to
		4080	include before including F<ev.h>:
		4081
		4082	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		4083
		4084	This would create a file F<wrap.h> which essentially looks like this:
		4085
		4086	#define ev_backend myprefix_ev_backend
		4087	#define ev_check_start myprefix_ev_check_start
		4088	#define ev_check_stop myprefix_ev_check_stop
		4089	...
		4090
		4091	=head2 EXAMPLES
		4092
		4093	For a real-world example of a program the includes libev
		4094	verbatim, you can have a look at the EV perl module
		4095	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4096	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4097	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4098	will be compiled. It is pretty complex because it provides its own header
		4099	file.
		4100
		4101	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4102	that everybody includes and which overrides some configure choices:
		4103
		4104	#define EV_FEATURES 8
		4105	#define EV_USE_SELECT 1
		4106	#define EV_PREPARE_ENABLE 1
		4107	#define EV_IDLE_ENABLE 1
		4108	#define EV_SIGNAL_ENABLE 1
		4109	#define EV_CHILD_ENABLE 1
		4110	#define EV_USE_STDEXCEPT 0
		4111	#define EV_CONFIG_H <config.h>
		4112
		4113	#include "ev++.h"
		4114
		4115	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4116
		4117	#include "ev_cpp.h"
		4118	#include "ev.c"
		4119
		4120	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4121
		4122	=head2 THREADS AND COROUTINES
		4123
		4124	=head3 THREADS
		4125
		4126	All libev functions are reentrant and thread-safe unless explicitly
		4127	documented otherwise, but libev implements no locking itself. This means
		4128	that you can use as many loops as you want in parallel, as long as there
		4129	are no concurrent calls into any libev function with the same loop
		4130	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4131	of course): libev guarantees that different event loops share no data
		4132	structures that need any locking.
		4133
		4134	Or to put it differently: calls with different loop parameters can be done
		4135	concurrently from multiple threads, calls with the same loop parameter
		4136	must be done serially (but can be done from different threads, as long as
		4137	only one thread ever is inside a call at any point in time, e.g. by using
		4138	a mutex per loop).
		4139
		4140	Specifically to support threads (and signal handlers), libev implements
		4141	so-called C<ev_async> watchers, which allow some limited form of
		4142	concurrency on the same event loop, namely waking it up "from the
		4143	outside".
		4144
		4145	If you want to know which design (one loop, locking, or multiple loops
		4146	without or something else still) is best for your problem, then I cannot
		4147	help you, but here is some generic advice:
		4148
		4149	=over 4
		4150
		4151	=item * most applications have a main thread: use the default libev loop
		4152	in that thread, or create a separate thread running only the default loop.
		4153
		4154	This helps integrating other libraries or software modules that use libev
		4155	themselves and don't care/know about threading.
		4156
		4157	=item * one loop per thread is usually a good model.
		4158
		4159	Doing this is almost never wrong, sometimes a better-performance model
		4160	exists, but it is always a good start.
		4161
		4162	=item * other models exist, such as the leader/follower pattern, where one
		4163	loop is handed through multiple threads in a kind of round-robin fashion.
		4164
		4165	Choosing a model is hard - look around, learn, know that usually you can do
		4166	better than you currently do :-)
		4167
		4168	=item * often you need to talk to some other thread which blocks in the
		4169	event loop.
		4170
		4171	C<ev_async> watchers can be used to wake them up from other threads safely
		4172	(or from signal contexts...).
		4173
		4174	An example use would be to communicate signals or other events that only
		4175	work in the default loop by registering the signal watcher with the
		4176	default loop and triggering an C<ev_async> watcher from the default loop
		4177	watcher callback into the event loop interested in the signal.
		4178
		4179	=back
		4180
		4181	=head4 THREAD LOCKING EXAMPLE
		4182
		4183	Here is a fictitious example of how to run an event loop in a different
		4184	thread than where callbacks are being invoked and watchers are
		4185	created/added/removed.
		4186
		4187	For a real-world example, see the C<EV::Loop::Async> perl module,
		4188	which uses exactly this technique (which is suited for many high-level
		4189	languages).
		4190
		4191	The example uses a pthread mutex to protect the loop data, a condition
		4192	variable to wait for callback invocations, an async watcher to notify the
		4193	event loop thread and an unspecified mechanism to wake up the main thread.
		4194
		4195	First, you need to associate some data with the event loop:
		4196
		4197	typedef struct {
		4198	mutex_t lock; /* global loop lock */
		4199	ev_async async_w;
		4200	thread_t tid;
		4201	cond_t invoke_cv;
		4202	} userdata;
		4203
		4204	void prepare_loop (EV_P)
		4205	{
		4206	// for simplicity, we use a static userdata struct.
		4207	static userdata u;
		4208
		4209	ev_async_init (&u->async_w, async_cb);
		4210	ev_async_start (EV_A_ &u->async_w);
		4211
		4212	pthread_mutex_init (&u->lock, 0);
		4213	pthread_cond_init (&u->invoke_cv, 0);
		4214
		4215	// now associate this with the loop
		4216	ev_set_userdata (EV_A_ u);
		4217	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4218	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4219
		4220	// then create the thread running ev_loop
		4221	pthread_create (&u->tid, 0, l_run, EV_A);
		4222	}
		4223
		4224	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4225	solely to wake up the event loop so it takes notice of any new watchers
		4226	that might have been added:
		4227
		4228	static void
		4229	async_cb (EV_P_ ev_async *w, int revents)
		4230	{
		4231	// just used for the side effects
		4232	}
		4233
		4234	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4235	protecting the loop data, respectively.
		4236
		4237	static void
		4238	l_release (EV_P)
		4239	{
		4240	userdata *u = ev_userdata (EV_A);
		4241	pthread_mutex_unlock (&u->lock);
		4242	}
		4243
		4244	static void
		4245	l_acquire (EV_P)
		4246	{
		4247	userdata *u = ev_userdata (EV_A);
		4248	pthread_mutex_lock (&u->lock);
		4249	}
		4250
		4251	The event loop thread first acquires the mutex, and then jumps straight
		4252	into C<ev_loop>:
		4253
		4254	void *
		4255	l_run (void *thr_arg)
		4256	{
		4257	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4258
		4259	l_acquire (EV_A);
		4260	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4261	ev_loop (EV_A_ 0);
		4262	l_release (EV_A);
		4263
		4264	return 0;
		4265	}
		4266
		4267	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4268	signal the main thread via some unspecified mechanism (signals? pipe
		4269	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4270	have been called (in a while loop because a) spurious wakeups are possible
		4271	and b) skipping inter-thread-communication when there are no pending
		4272	watchers is very beneficial):
		4273
		4274	static void
		4275	l_invoke (EV_P)
		4276	{
		4277	userdata *u = ev_userdata (EV_A);
		4278
		4279	while (ev_pending_count (EV_A))
		4280	{
		4281	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4282	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4283	}
		4284	}
		4285
		4286	Now, whenever the main thread gets told to invoke pending watchers, it
		4287	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4288	thread to continue:
		4289
		4290	static void
		4291	real_invoke_pending (EV_P)
		4292	{
		4293	userdata *u = ev_userdata (EV_A);
		4294
		4295	pthread_mutex_lock (&u->lock);
		4296	ev_invoke_pending (EV_A);
		4297	pthread_cond_signal (&u->invoke_cv);
		4298	pthread_mutex_unlock (&u->lock);
		4299	}
		4300
		4301	Whenever you want to start/stop a watcher or do other modifications to an
		4302	event loop, you will now have to lock:
		4303
		4304	ev_timer timeout_watcher;
		4305	userdata *u = ev_userdata (EV_A);
		4306
		4307	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4308
		4309	pthread_mutex_lock (&u->lock);
		4310	ev_timer_start (EV_A_ &timeout_watcher);
		4311	ev_async_send (EV_A_ &u->async_w);
		4312	pthread_mutex_unlock (&u->lock);
		4313
		4314	Note that sending the C<ev_async> watcher is required because otherwise
		4315	an event loop currently blocking in the kernel will have no knowledge
		4316	about the newly added timer. By waking up the loop it will pick up any new
		4317	watchers in the next event loop iteration.
		4318
		4319	=head3 COROUTINES
		4320
		4321	Libev is very accommodating to coroutines ("cooperative threads"):
		4322	libev fully supports nesting calls to its functions from different
		4323	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		4324	different coroutines, and switch freely between both coroutines running
		4325	the loop, as long as you don't confuse yourself). The only exception is
		4326	that you must not do this from C<ev_periodic> reschedule callbacks.
		4327
		4328	Care has been taken to ensure that libev does not keep local state inside
		4329	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		4330	they do not call any callbacks.
		4331
		4332	=head2 COMPILER WARNINGS
		4333
		4334	Depending on your compiler and compiler settings, you might get no or a
		4335	lot of warnings when compiling libev code. Some people are apparently
		4336	scared by this.
		4337
		4338	However, these are unavoidable for many reasons. For one, each compiler
		4339	has different warnings, and each user has different tastes regarding
		4340	warning options. "Warn-free" code therefore cannot be a goal except when
		4341	targeting a specific compiler and compiler-version.
		4342
		4343	Another reason is that some compiler warnings require elaborate
		4344	workarounds, or other changes to the code that make it less clear and less
		4345	maintainable.
		4346
		4347	And of course, some compiler warnings are just plain stupid, or simply
		4348	wrong (because they don't actually warn about the condition their message
		4349	seems to warn about). For example, certain older gcc versions had some
		4350	warnings that resulted an extreme number of false positives. These have
		4351	been fixed, but some people still insist on making code warn-free with
		4352	such buggy versions.
		4353
		4354	While libev is written to generate as few warnings as possible,
		4355	"warn-free" code is not a goal, and it is recommended not to build libev
		4356	with any compiler warnings enabled unless you are prepared to cope with
		4357	them (e.g. by ignoring them). Remember that warnings are just that:
		4358	warnings, not errors, or proof of bugs.
		4359
		4360
		4361	=head2 VALGRIND
		4362
		4363	Valgrind has a special section here because it is a popular tool that is
		4364	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4365
		4366	If you think you found a bug (memory leak, uninitialised data access etc.)
		4367	in libev, then check twice: If valgrind reports something like:
		4368
		4369	==2274== definitely lost: 0 bytes in 0 blocks.
		4370	==2274== possibly lost: 0 bytes in 0 blocks.
		4371	==2274== still reachable: 256 bytes in 1 blocks.
		4372
		4373	Then there is no memory leak, just as memory accounted to global variables
		4374	is not a memleak - the memory is still being referenced, and didn't leak.
		4375
		4376	Similarly, under some circumstances, valgrind might report kernel bugs
		4377	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4378	although an acceptable workaround has been found here), or it might be
		4379	confused.
		4380
		4381	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4382	make it into some kind of religion.
		4383
		4384	If you are unsure about something, feel free to contact the mailing list
		4385	with the full valgrind report and an explanation on why you think this
		4386	is a bug in libev (best check the archives, too :). However, don't be
		4387	annoyed when you get a brisk "this is no bug" answer and take the chance
		4388	of learning how to interpret valgrind properly.
		4389
		4390	If you need, for some reason, empty reports from valgrind for your project
		4391	I suggest using suppression lists.
		4392
		4393
		4394	=head1 PORTABILITY NOTES
		4395
		4396	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4397
		4398	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4399	requires, and its I/O model is fundamentally incompatible with the POSIX
		4400	model. Libev still offers limited functionality on this platform in
		4401	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4402	descriptors. This only applies when using Win32 natively, not when using
		4403	e.g. cygwin.
		4404
		4405	Lifting these limitations would basically require the full
		4406	re-implementation of the I/O system. If you are into these kinds of
		4407	things, then note that glib does exactly that for you in a very portable
		4408	way (note also that glib is the slowest event library known to man).
		4409
		4410	There is no supported compilation method available on windows except
		4411	embedding it into other applications.
		4412
		4413	Sensible signal handling is officially unsupported by Microsoft - libev
		4414	tries its best, but under most conditions, signals will simply not work.
		4415
		4416	Not a libev limitation but worth mentioning: windows apparently doesn't
		4417	accept large writes: instead of resulting in a partial write, windows will
		4418	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4419	so make sure you only write small amounts into your sockets (less than a
		4420	megabyte seems safe, but this apparently depends on the amount of memory
		4421	available).
		4422
		4423	Due to the many, low, and arbitrary limits on the win32 platform and
		4424	the abysmal performance of winsockets, using a large number of sockets
		4425	is not recommended (and not reasonable). If your program needs to use
		4426	more than a hundred or so sockets, then likely it needs to use a totally
		4427	different implementation for windows, as libev offers the POSIX readiness
		4428	notification model, which cannot be implemented efficiently on windows
		4429	(due to Microsoft monopoly games).
		4430
		4431	A typical way to use libev under windows is to embed it (see the embedding
		4432	section for details) and use the following F<evwrap.h> header file instead
		4433	of F<ev.h>:
		4434
		4435	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4436	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4437
		4438	#include "ev.h"
		4439
		4440	And compile the following F<evwrap.c> file into your project (make sure
		4441	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4442
		4443	#include "evwrap.h"
		4444	#include "ev.c"
		4445
		4446	=over 4
		4447
		4448	=item The winsocket select function
		4449
		4450	The winsocket C<select> function doesn't follow POSIX in that it
		4451	requires socket I<handles> and not socket I<file descriptors> (it is
		4452	also extremely buggy). This makes select very inefficient, and also
		4453	requires a mapping from file descriptors to socket handles (the Microsoft
		4454	C runtime provides the function C<_open_osfhandle> for this). See the
		4455	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4456	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4457
		4458	The configuration for a "naked" win32 using the Microsoft runtime
		4459	libraries and raw winsocket select is:
		4460
		4461	#define EV_USE_SELECT 1
		4462	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4463
		4464	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4465	complexity in the O(n²) range when using win32.
		4466
		4467	=item Limited number of file descriptors
		4468
		4469	Windows has numerous arbitrary (and low) limits on things.
		4470
		4471	Early versions of winsocket's select only supported waiting for a maximum
		4472	of C<64> handles (probably owning to the fact that all windows kernels
		4473	can only wait for C<64> things at the same time internally; Microsoft
		4474	recommends spawning a chain of threads and wait for 63 handles and the
		4475	previous thread in each. Sounds great!).
		4476
		4477	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4478	to some high number (e.g. C<2048>) before compiling the winsocket select
		4479	call (which might be in libev or elsewhere, for example, perl and many
		4480	other interpreters do their own select emulation on windows).
		4481
		4482	Another limit is the number of file descriptors in the Microsoft runtime
		4483	libraries, which by default is C<64> (there must be a hidden I<64>
		4484	fetish or something like this inside Microsoft). You can increase this
		4485	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4486	(another arbitrary limit), but is broken in many versions of the Microsoft
		4487	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4488	(depending on windows version and/or the phase of the moon). To get more,
		4489	you need to wrap all I/O functions and provide your own fd management, but
		4490	the cost of calling select (O(n²)) will likely make this unworkable.
		4491
		4492	=back
		4493
		4494	=head2 PORTABILITY REQUIREMENTS
		4495
		4496	In addition to a working ISO-C implementation and of course the
		4497	backend-specific APIs, libev relies on a few additional extensions:
		4498
		4499	=over 4
		4500
		4501	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4502	calling conventions regardless of C<ev_watcher_type *>.
		4503
		4504	Libev assumes not only that all watcher pointers have the same internal
		4505	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4506	assumes that the same (machine) code can be used to call any watcher
		4507	callback: The watcher callbacks have different type signatures, but libev
		4508	calls them using an C<ev_watcher *> internally.
		4509
		4510	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4511
		4512	The type C<sig_atomic_t volatile> (or whatever is defined as
		4513	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4514	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4515	believed to be sufficiently portable.
		4516
		4517	=item C<sigprocmask> must work in a threaded environment
		4518
		4519	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4520	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4521	pthread implementations will either allow C<sigprocmask> in the "main
		4522	thread" or will block signals process-wide, both behaviours would
		4523	be compatible with libev. Interaction between C<sigprocmask> and
		4524	C<pthread_sigmask> could complicate things, however.
		4525
		4526	The most portable way to handle signals is to block signals in all threads
		4527	except the initial one, and run the default loop in the initial thread as
		4528	well.
		4529
		4530	=item C<long> must be large enough for common memory allocation sizes
		4531
		4532	To improve portability and simplify its API, libev uses C<long> internally
		4533	instead of C<size_t> when allocating its data structures. On non-POSIX
		4534	systems (Microsoft...) this might be unexpectedly low, but is still at
		4535	least 31 bits everywhere, which is enough for hundreds of millions of
		4536	watchers.
		4537
		4538	=item C<double> must hold a time value in seconds with enough accuracy
		4539
		4540	The type C<double> is used to represent timestamps. It is required to
		4541	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		4542	enough for at least into the year 4000. This requirement is fulfilled by
		4543	implementations implementing IEEE 754, which is basically all existing
		4544	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4545	2200.
		4546
		4547	=back
		4548
		4549	If you know of other additional requirements drop me a note.
		4550
		4551
		4552	=head1 ALGORITHMIC COMPLEXITIES
		4553
		4554	In this section the complexities of (many of) the algorithms used inside
		4555	libev will be documented. For complexity discussions about backends see
		4556	the documentation for C<ev_default_init>.
		4557
		4558	All of the following are about amortised time: If an array needs to be
		4559	extended, libev needs to realloc and move the whole array, but this
		4560	happens asymptotically rarer with higher number of elements, so O(1) might
		4561	mean that libev does a lengthy realloc operation in rare cases, but on
		4562	average it is much faster and asymptotically approaches constant time.
		4563
		4564	=over 4
		4565
		4566	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4567
		4568	This means that, when you have a watcher that triggers in one hour and
		4569	there are 100 watchers that would trigger before that, then inserting will
		4570	have to skip roughly seven (C<ld 100>) of these watchers.
		4571
		4572	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4573
		4574	That means that changing a timer costs less than removing/adding them,
		4575	as only the relative motion in the event queue has to be paid for.
		4576
		4577	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4578
		4579	These just add the watcher into an array or at the head of a list.
		4580
		4581	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4582
		4583	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4584
		4585	These watchers are stored in lists, so they need to be walked to find the
		4586	correct watcher to remove. The lists are usually short (you don't usually
		4587	have many watchers waiting for the same fd or signal: one is typical, two
		4588	is rare).
		4589
		4590	=item Finding the next timer in each loop iteration: O(1)
		4591
		4592	By virtue of using a binary or 4-heap, the next timer is always found at a
		4593	fixed position in the storage array.
		4594
		4595	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4596
		4597	A change means an I/O watcher gets started or stopped, which requires
		4598	libev to recalculate its status (and possibly tell the kernel, depending
		4599	on backend and whether C<ev_io_set> was used).
		4600
		4601	=item Activating one watcher (putting it into the pending state): O(1)
		4602
		4603	=item Priority handling: O(number_of_priorities)
		4604
		4605	Priorities are implemented by allocating some space for each
		4606	priority. When doing priority-based operations, libev usually has to
		4607	linearly search all the priorities, but starting/stopping and activating
		4608	watchers becomes O(1) with respect to priority handling.
		4609
		4610	=item Sending an ev_async: O(1)
		4611
		4612	=item Processing ev_async_send: O(number_of_async_watchers)
		4613
		4614	=item Processing signals: O(max_signal_number)
		4615
		4616	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4617	calls in the current loop iteration. Checking for async and signal events
		4618	involves iterating over all running async watchers or all signal numbers.
		4619
		4620	=back
		4621
		4622
		4623	=head1 PORTING FROM 3.X TO 4.X
		4624
		4625	The major version 4 introduced some minor incompatible changes to the API.
		4626
		4627	=over 4
		4628
		4629	=item C<EV_TIMEOUT> replaced by C<EV_TIMER> in C<revents>
		4630
		4631	This is a simple rename - all other watcher types use their name
		4632	as revents flag, and now C<ev_timer> does, too.
		4633
		4634	Both C<EV_TIMER> and C<EV_TIMEOUT> symbols were present in 3.x versions
		4635	and continue to be present for the forseeable future, so this is mostly a
		4636	documentation change.
		4637
		4638	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		4639
		4640	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		4641	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		4642	and work, but the library code will of course be larger.
		4643
		4644	=back
		4645
		4646
		4647	=head1 GLOSSARY
		4648
		4649	=over 4
		4650
		4651	=item active
		4652
		4653	A watcher is active as long as it has been started (has been attached to
		4654	an event loop) but not yet stopped (disassociated from the event loop).
		4655
		4656	=item application
		4657
		4658	In this document, an application is whatever is using libev.
		4659
		4660	=item callback
		4661
		4662	The address of a function that is called when some event has been
		4663	detected. Callbacks are being passed the event loop, the watcher that
		4664	received the event, and the actual event bitset.
		4665
		4666	=item callback invocation
		4667
		4668	The act of calling the callback associated with a watcher.
		4669
		4670	=item event
		4671
		4672	A change of state of some external event, such as data now being available
		4673	for reading on a file descriptor, time having passed or simply not having
		4674	any other events happening anymore.
		4675
		4676	In libev, events are represented as single bits (such as C<EV_READ> or
		4677	C<EV_TIMER>).
		4678
		4679	=item event library
		4680
		4681	A software package implementing an event model and loop.
		4682
		4683	=item event loop
		4684
		4685	An entity that handles and processes external events and converts them
		4686	into callback invocations.
		4687
		4688	=item event model
		4689
		4690	The model used to describe how an event loop handles and processes
		4691	watchers and events.
		4692
		4693	=item pending
		4694
		4695	A watcher is pending as soon as the corresponding event has been detected,
		4696	and stops being pending as soon as the watcher will be invoked or its
		4697	pending status is explicitly cleared by the application.
		4698
		4699	A watcher can be pending, but not active. Stopping a watcher also clears
		4700	its pending status.
		4701
		4702	=item real time
		4703
		4704	The physical time that is observed. It is apparently strictly monotonic :)
		4705
		4706	=item wall-clock time
		4707
		4708	The time and date as shown on clocks. Unlike real time, it can actually
		4709	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4710	clock.
		4711
		4712	=item watcher
		4713
		4714	A data structure that describes interest in certain events. Watchers need
		4715	to be started (attached to an event loop) before they can receive events.
		4716
		4717	=item watcher invocation
		4718
		4719	The act of calling the callback associated with a watcher.
		4720
		4721	=back
		4722
767	=head1 AUTHOR	4723	=head1 AUTHOR
768		4724
769	Marc Lehmann <libev@schmorp.de>.	4725	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
770		4726

Diff Legend

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