ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.30 by root, Fri Nov 23 04:36:03 2007 UTC vs.
Revision 1.277 by root, Thu Dec 31 06:50:17 2009 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
9	=head1 DESCRIPTION	9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	#include <stdio.h> // for puts
		15
		16	// every watcher type has its own typedef'd struct
		17	// with the name ev_TYPE
		18	ev_io stdin_watcher;
		19	ev_timer timeout_watcher;
		20
		21	// all watcher callbacks have a similar signature
		22	// this callback is called when data is readable on stdin
		23	static void
		24	stdin_cb (EV_P_ ev_io *w, int revents)
		25	{
		26	puts ("stdin ready");
		27	// for one-shot events, one must manually stop the watcher
		28	// with its corresponding stop function.
		29	ev_io_stop (EV_A_ w);
		30
		31	// this causes all nested ev_loop's to stop iterating
		32	ev_unloop (EV_A_ EVUNLOOP_ALL);
		33	}
		34
		35	// another callback, this time for a time-out
		36	static void
		37	timeout_cb (EV_P_ ev_timer *w, int revents)
		38	{
		39	puts ("timeout");
		40	// this causes the innermost ev_loop to stop iterating
		41	ev_unloop (EV_A_ EVUNLOOP_ONE);
		42	}
		43
		44	int
		45	main (void)
		46	{
		47	// use the default event loop unless you have special needs
		48	struct ev_loop *loop = ev_default_loop (0);
		49
		50	// initialise an io watcher, then start it
		51	// this one will watch for stdin to become readable
		52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		53	ev_io_start (loop, &stdin_watcher);
		54
		55	// initialise a timer watcher, then start it
		56	// simple non-repeating 5.5 second timeout
		57	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		58	ev_timer_start (loop, &timeout_watcher);
		59
		60	// now wait for events to arrive
		61	ev_loop (loop, 0);
		62
		63	// unloop was called, so exit
		64	return 0;
		65	}
		66
		67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
		70
		71	The newest version of this document is also available as an html-formatted
		72	web page you might find easier to navigate when reading it for the first
		73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 ABOUT LIBEV
10		84
11	Libev is an event loop: you register interest in certain events (such as a	85	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	86	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	87	these event sources and provide your program with events.
14		88
15	To do this, it must take more or less complete control over your process	89	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	90	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	91	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	93	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	94	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	95	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	96	watcher.
23		97
24	=head1 FEATURES	98	=head2 FEATURES
25		99
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	100	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	101	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	102	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
		103	(for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		104	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		105	timers (C<ev_timer>), absolute timers with customised rescheduling
		106	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	107	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	108	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		109	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		110	limited support for fork events (C<ev_fork>).
		111
		112	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	113	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	114	for example).
33		115
34	=head1 CONVENTIONS	116	=head2 CONVENTIONS
35		117
36	Libev is very configurable. In this manual the default configuration	118	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	119	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	120	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	121	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	122	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	123	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	124	this argument.
43		125
44	=head1 TIME REPRESENTATION	126	=head2 TIME REPRESENTATION
45		127
46	Libev represents time as a single floating point number, representing the	128	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	129	the (fractional) number of seconds since the (POSIX) epoch (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
51		157
52	=head1 GLOBAL FUNCTIONS	158	=head1 GLOBAL FUNCTIONS
53		159
54	These functions can be called anytime, even before initialising the	160	These functions can be called anytime, even before initialising the
55	library in any way.	161	library in any way.
…		…
60		166
61	Returns the current time as libev would use it. Please note that the	167	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	168	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	169	you actually want to know.
64		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 ()>.
		176
65	=item int ev_version_major ()	177	=item int ev_version_major ()
66		178
67	=item int ev_version_minor ()	179	=item int ev_version_minor ()
68		180
69	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
70	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
71	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
72	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
73	version of the library your program was compiled against.	185	version of the library your program was compiled against.
74		186
		187	These version numbers refer to the ABI version of the library, not the
		188	release version.
		189
75	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,
76	as this indicates an incompatible change. Minor versions are usually	191	as this indicates an incompatible change. Minor versions are usually
77	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
78	not a problem.	193	not a problem.
79		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
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	234	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
81		235
82	Sets the allocation function to use (the prototype is similar to the	236	Sets the allocation function to use (the prototype is similar - the
83	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
84	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
85	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
86	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.
87		245
88	You could override this function in high-availability programs to, say,	246	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	247	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	248	or even to sleep a while and retry until some memory is available.
91		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
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	270	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
93		271
94	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
95	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
96	indicating the system call or subsystem causing the problem. If this	274	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	275	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	276	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	277	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	278	(such as abort).
101		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
102	=back	292	=back
103		293
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	294	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
105		295
106	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>
107	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>
108	events, and dynamically created loops which do not.	298	I<function>).
109		299
110	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
111	in your main thread (or in a separate thread) and for each thread you	301	supports signals and child events, and dynamically created loops which do
112	create, you also create another event loop. Libev itself does no locking	302	not.
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		303
117	=over 4	304	=over 4
118		305
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	306	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		307
121	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
122	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
123	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
124	flags).	311	flags. If that is troubling you, check C<ev_backend ()> afterwards).
125		312
126	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
127	function.	314	function.
128		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
129	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
130	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>).
131		329
132	It supports the following flags:	330	The following flags are supported:
133		331
134	=over 4	332	=over 4
135		333
136	=item C<EVFLAG_AUTO>	334	=item C<EVFLAG_AUTO>
137		335
138	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
139	thing, believe me).	337	thing, believe me).
140		338
141	=item C<EVFLAG_NOENV>	339	=item C<EVFLAG_NOENV>
142		340
143	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
144	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
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	343	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	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
147	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
148	around bugs.	346	around bugs.
149		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 is both faster and might make
		380	it possible to get the queued signal data.
		381
		382	Signalfd will not be used by default as this changes your signal mask, and
		383	there are a lot of shoddy libraries and programs (glib's threadpool for
		384	example) that can't properly initialise their signal masks.
		385
150	=item C<EVMETHOD_SELECT> (value 1, portable select backend)	386	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		387
152	This is your standard select(2) backend. Not I<completely> standard, as	388	This is your standard select(2) backend. Not I<completely> standard, as
153	libev tries to roll its own fd_set with no limits on the number of fds,	389	libev tries to roll its own fd_set with no limits on the number of fds,
154	but if that fails, expect a fairly low limit on the number of fds when	390	but if that fails, expect a fairly low limit on the number of fds when
155	using this backend. It doesn't scale too well (O(highest_fd)), but its usually	391	using this backend. It doesn't scale too well (O(highest_fd)), but its
156	the fastest backend for a low number of fds.	392	usually the fastest backend for a low number of (low-numbered :) fds.
157		393
		394	To get good performance out of this backend you need a high amount of
		395	parallelism (most of the file descriptors should be busy). If you are
		396	writing a server, you should C<accept ()> in a loop to accept as many
		397	connections as possible during one iteration. You might also want to have
		398	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		399	readiness notifications you get per iteration.
		400
		401	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		402	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		403	C<exceptfds> set on that platform).
		404
158	=item C<EVMETHOD_POLL> (value 2, poll backend, available everywhere except on windows)	405	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
159		406
160	And this is your standard poll(2) backend. It's more complicated than	407	And this is your standard poll(2) backend. It's more complicated
161	select, but handles sparse fds better and has no artificial limit on the	408	than select, but handles sparse fds better and has no artificial
162	number of fds you can use (except it will slow down considerably with a	409	limit on the number of fds you can use (except it will slow down
163	lot of inactive fds). It scales similarly to select, i.e. O(total_fds).	410	considerably with a lot of inactive fds). It scales similarly to select,
		411	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		412	performance tips.
164		413
		414	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		415	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
		416
165	=item C<EVMETHOD_EPOLL> (value 4, Linux)	417	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		418
		419	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		420	kernels).
166		421
167	For few fds, this backend is a bit little slower than poll and select,	422	For few fds, this backend is a bit little slower than poll and select,
168	but it scales phenomenally better. While poll and select usually scale like	423	but it scales phenomenally better. While poll and select usually scale
169	O(total_fds) where n is the total number of fds (or the highest fd), epoll scales	424	like O(total_fds) where n is the total number of fds (or the highest fd),
170	either O(1) or O(active_fds).	425	epoll scales either O(1) or O(active_fds).
171		426
		427	The epoll mechanism deserves honorable mention as the most misdesigned
		428	of the more advanced event mechanisms: mere annoyances include silently
		429	dropping file descriptors, requiring a system call per change per file
		430	descriptor (and unnecessary guessing of parameters), problems with dup and
		431	so on. The biggest issue is fork races, however - if a program forks then
		432	I<both> parent and child process have to recreate the epoll set, which can
		433	take considerable time (one syscall per file descriptor) and is of course
		434	hard to detect.
		435
		436	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		437	of course I<doesn't>, and epoll just loves to report events for totally
		438	I<different> file descriptors (even already closed ones, so one cannot
		439	even remove them from the set) than registered in the set (especially
		440	on SMP systems). Libev tries to counter these spurious notifications by
		441	employing an additional generation counter and comparing that against the
		442	events to filter out spurious ones, recreating the set when required.
		443
172	While stopping and starting an I/O watcher in the same iteration will	444	While stopping, setting and starting an I/O watcher in the same iteration
173	result in some caching, there is still a syscall per such incident	445	will result in some caching, there is still a system call per such
174	(because the fd could point to a different file description now), so its	446	incident (because the same I<file descriptor> could point to a different
175	best to avoid that. Also, dup()ed file descriptors might not work very	447	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
176	well if you register events for both fds.	448	file descriptors might not work very well if you register events for both
		449	file descriptors.
177		450
		451	Best performance from this backend is achieved by not unregistering all
		452	watchers for a file descriptor until it has been closed, if possible,
		453	i.e. keep at least one watcher active per fd at all times. Stopping and
		454	starting a watcher (without re-setting it) also usually doesn't cause
		455	extra overhead. A fork can both result in spurious notifications as well
		456	as in libev having to destroy and recreate the epoll object, which can
		457	take considerable time and thus should be avoided.
		458
		459	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		460	faster than epoll for maybe up to a hundred file descriptors, depending on
		461	the usage. So sad.
		462
		463	While nominally embeddable in other event loops, this feature is broken in
		464	all kernel versions tested so far.
		465
		466	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		467	C<EVBACKEND_POLL>.
		468
178	=item C<EVMETHOD_KQUEUE> (value 8, most BSD clones)	469	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
179		470
180	Kqueue deserves special mention, as at the time of this writing, it	471	Kqueue deserves special mention, as at the time of this writing, it
181	was broken on all BSDs except NetBSD (usually it doesn't work with	472	was broken on all BSDs except NetBSD (usually it doesn't work reliably
182	anything but sockets and pipes, except on Darwin, where of course its	473	with anything but sockets and pipes, except on Darwin, where of course
183	completely useless). For this reason its not being "autodetected" unless	474	it's completely useless). Unlike epoll, however, whose brokenness
184	you explicitly specify the flags (i.e. you don't use EVFLAG_AUTO).	475	is by design, these kqueue bugs can (and eventually will) be fixed
		476	without API changes to existing programs. For this reason it's not being
		477	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		478	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		479	system like NetBSD.
		480
		481	You still can embed kqueue into a normal poll or select backend and use it
		482	only for sockets (after having made sure that sockets work with kqueue on
		483	the target platform). See C<ev_embed> watchers for more info.
185		484
186	It scales in the same way as the epoll backend, but the interface to the	485	It scales in the same way as the epoll backend, but the interface to the
187	kernel is more efficient (which says nothing about its actual speed, of	486	kernel is more efficient (which says nothing about its actual speed, of
188	course). While starting and stopping an I/O watcher does not cause an	487	course). While stopping, setting and starting an I/O watcher does never
189	extra syscall as with epoll, it still adds up to four event changes per	488	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
190	incident, so its best to avoid that.	489	two event changes per incident. Support for C<fork ()> is very bad (but
		490	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		491	cases
191		492
		493	This backend usually performs well under most conditions.
		494
		495	While nominally embeddable in other event loops, this doesn't work
		496	everywhere, so you might need to test for this. And since it is broken
		497	almost everywhere, you should only use it when you have a lot of sockets
		498	(for which it usually works), by embedding it into another event loop
		499	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		500	also broken on OS X)) and, did I mention it, using it only for sockets.
		501
		502	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		503	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		504	C<NOTE_EOF>.
		505
192	=item C<EVMETHOD_DEVPOLL> (value 16, Solaris 8)	506	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
193		507
194	This is not implemented yet (and might never be).	508	This is not implemented yet (and might never be, unless you send me an
		509	implementation). According to reports, C</dev/poll> only supports sockets
		510	and is not embeddable, which would limit the usefulness of this backend
		511	immensely.
195		512
196	=item C<EVMETHOD_PORT> (value 32, Solaris 10)	513	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
197		514
198	This uses the Solaris 10 port mechanism. As with everything on Solaris,	515	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
199	it's really slow, but it still scales very well (O(active_fds)).	516	it's really slow, but it still scales very well (O(active_fds)).
200		517
		518	Please note that Solaris event ports can deliver a lot of spurious
		519	notifications, so you need to use non-blocking I/O or other means to avoid
		520	blocking when no data (or space) is available.
		521
		522	While this backend scales well, it requires one system call per active
		523	file descriptor per loop iteration. For small and medium numbers of file
		524	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		525	might perform better.
		526
		527	On the positive side, with the exception of the spurious readiness
		528	notifications, this backend actually performed fully to specification
		529	in all tests and is fully embeddable, which is a rare feat among the
		530	OS-specific backends (I vastly prefer correctness over speed hacks).
		531
		532	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		533	C<EVBACKEND_POLL>.
		534
201	=item C<EVMETHOD_ALL>	535	=item C<EVBACKEND_ALL>
202		536
203	Try all backends (even potentially broken ones that wouldn't be tried	537	Try all backends (even potentially broken ones that wouldn't be tried
204	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as	538	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
205	C<EVMETHOD_ALL & ~EVMETHOD_KQUEUE>.	539	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		540
		541	It is definitely not recommended to use this flag.
206		542
207	=back	543	=back
208		544
209	If one or more of these are ored into the flags value, then only these	545	If one or more of the backend flags are or'ed into the flags value,
210	backends will be tried (in the reverse order as given here). If none are	546	then only these backends will be tried (in the reverse order as listed
211	specified, most compiled-in backend will be tried, usually in reverse	547	here). If none are specified, all backends in C<ev_recommended_backends
212	order of their flag values :)	548	()> will be tried.
		549
		550	Example: This is the most typical usage.
		551
		552	if (!ev_default_loop (0))
		553	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		554
		555	Example: Restrict libev to the select and poll backends, and do not allow
		556	environment settings to be taken into account:
		557
		558	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		559
		560	Example: Use whatever libev has to offer, but make sure that kqueue is
		561	used if available (warning, breaks stuff, best use only with your own
		562	private event loop and only if you know the OS supports your types of
		563	fds):
		564
		565	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
213		566
214	=item struct ev_loop *ev_loop_new (unsigned int flags)	567	=item struct ev_loop *ev_loop_new (unsigned int flags)
215		568
216	Similar to C<ev_default_loop>, but always creates a new event loop that is	569	Similar to C<ev_default_loop>, but always creates a new event loop that is
217	always distinct from the default loop. Unlike the default loop, it cannot	570	always distinct from the default loop. Unlike the default loop, it cannot
218	handle signal and child watchers, and attempts to do so will be greeted by	571	handle signal and child watchers, and attempts to do so will be greeted by
219	undefined behaviour (or a failed assertion if assertions are enabled).	572	undefined behaviour (or a failed assertion if assertions are enabled).
220		573
		574	Note that this function I<is> thread-safe, and the recommended 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");
		583
221	=item ev_default_destroy ()	584	=item ev_default_destroy ()
222		585
223	Destroys the default loop again (frees all memory and kernel state	586	Destroys the default loop again (frees all memory and kernel state
224	etc.). This stops all registered event watchers (by not touching them in	587	etc.). None of the active event watchers will be stopped in the normal
225	any way whatsoever, although you cannot rely on this :).	588	sense, so e.g. C<ev_is_active> might still return true. It is your
		589	responsibility to either stop all watchers cleanly yourself I<before>
		590	calling this function, or cope with the fact afterwards (which is usually
		591	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		592	for example).
		593
		594	Note that certain global state, such as signal state (and installed signal
		595	handlers), will not be freed by this function, and related watchers (such
		596	as signal and child watchers) would need to be stopped manually.
		597
		598	In general it is not advisable to call this function except in the
		599	rare occasion where you really need to free e.g. the signal handling
		600	pipe fds. If you need dynamically allocated loops it is better to use
		601	C<ev_loop_new> and C<ev_loop_destroy>.
226		602
227	=item ev_loop_destroy (loop)	603	=item ev_loop_destroy (loop)
228		604
229	Like C<ev_default_destroy>, but destroys an event loop created by an	605	Like C<ev_default_destroy>, but destroys an event loop created by an
230	earlier call to C<ev_loop_new>.	606	earlier call to C<ev_loop_new>.
231		607
232	=item ev_default_fork ()	608	=item ev_default_fork ()
233		609
		610	This function sets a flag that causes subsequent C<ev_loop> iterations
234	This function reinitialises the kernel state for backends that have	611	to reinitialise the kernel state for backends that have one. Despite the
235	one. Despite the name, you can call it anytime, but it makes most sense	612	name, you can call it anytime, but it makes most sense after forking, in
236	after forking, in either the parent or child process (or both, but that	613	the child process (or both child and parent, but that again makes little
237	again makes little sense).	614	sense). You I<must> call it in the child before using any of the libev
		615	functions, and it will only take effect at the next C<ev_loop> iteration.
238		616
239	You I<must> call this function in the child process after forking if and	617	On the other hand, you only need to call this function in the child
240	only if you want to use the event library in both processes. If you just	618	process if and only if you want to use the event library in the child. If
241	fork+exec, you don't have to call it.	619	you just fork+exec, you don't have to call it at all.
242		620
243	The function itself is quite fast and it's usually not a problem to call	621	The function itself is quite fast and it's usually not a problem to call
244	it just in case after a fork. To make this easy, the function will fit in	622	it just in case after a fork. To make this easy, the function will fit in
245	quite nicely into a call to C<pthread_atfork>:	623	quite nicely into a call to C<pthread_atfork>:
246		624
…		…
248		626
249	=item ev_loop_fork (loop)	627	=item ev_loop_fork (loop)
250		628
251	Like C<ev_default_fork>, but acts on an event loop created by	629	Like C<ev_default_fork>, but acts on an event loop created by
252	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	630	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
253	after fork, and how you do this is entirely your own problem.	631	after fork that you want to re-use in the child, and how you do this is
		632	entirely your own problem.
254		633
		634	=item int ev_is_default_loop (loop)
		635
		636	Returns true when the given loop is, in fact, the default loop, and false
		637	otherwise.
		638
		639	=item unsigned int ev_loop_count (loop)
		640
		641	Returns the count of loop iterations for the loop, which is identical to
		642	the number of times libev did poll for new events. It starts at C<0> and
		643	happily wraps around with enough iterations.
		644
		645	This value can sometimes be useful as a generation counter of sorts (it
		646	"ticks" the number of loop iterations), as it roughly corresponds with
		647	C<ev_prepare> and C<ev_check> calls.
		648
		649	=item unsigned int ev_loop_depth (loop)
		650
		651	Returns the number of times C<ev_loop> was entered minus the number of
		652	times C<ev_loop> was exited, in other words, the recursion depth.
		653
		654	Outside C<ev_loop>, this number is zero. In a callback, this number is
		655	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		656	in which case it is higher.
		657
		658	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		659	etc.), doesn't count as exit.
		660
255	=item unsigned int ev_method (loop)	661	=item unsigned int ev_backend (loop)
256		662
257	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	663	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
258	use.	664	use.
259		665
260	=item ev_tstamp ev_now (loop)	666	=item ev_tstamp ev_now (loop)
261		667
262	Returns the current "event loop time", which is the time the event loop	668	Returns the current "event loop time", which is the time the event loop
263	got events and started processing them. This timestamp does not change	669	received events and started processing them. This timestamp does not
264	as long as callbacks are being processed, and this is also the base time	670	change as long as callbacks are being processed, and this is also the base
265	used for relative timers. You can treat it as the timestamp of the event	671	time used for relative timers. You can treat it as the timestamp of the
266	occuring (or more correctly, the mainloop finding out about it).	672	event occurring (or more correctly, libev finding out about it).
		673
		674	=item ev_now_update (loop)
		675
		676	Establishes the current time by querying the kernel, updating the time
		677	returned by C<ev_now ()> in the progress. This is a costly operation and
		678	is usually done automatically within C<ev_loop ()>.
		679
		680	This function is rarely useful, but when some event callback runs for a
		681	very long time without entering the event loop, updating libev's idea of
		682	the current time is a good idea.
		683
		684	See also L<The special problem of time updates> in the C<ev_timer> section.
		685
		686	=item ev_suspend (loop)
		687
		688	=item ev_resume (loop)
		689
		690	These two functions suspend and resume a loop, for use when the loop is
		691	not used for a while and timeouts should not be processed.
		692
		693	A typical use case would be an interactive program such as a game: When
		694	the user presses C<^Z> to suspend the game and resumes it an hour later it
		695	would be best to handle timeouts as if no time had actually passed while
		696	the program was suspended. This can be achieved by calling C<ev_suspend>
		697	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		698	C<ev_resume> directly afterwards to resume timer processing.
		699
		700	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		701	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		702	will be rescheduled (that is, they will lose any events that would have
		703	occured while suspended).
		704
		705	After calling C<ev_suspend> you B<must not> call I<any> function on the
		706	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		707	without a previous call to C<ev_suspend>.
		708
		709	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		710	event loop time (see C<ev_now_update>).
267		711
268	=item ev_loop (loop, int flags)	712	=item ev_loop (loop, int flags)
269		713
270	Finally, this is it, the event handler. This function usually is called	714	Finally, this is it, the event handler. This function usually is called
271	after you initialised all your watchers and you want to start handling	715	after you have initialised all your watchers and you want to start
272	events.	716	handling events.
273		717
274	If the flags argument is specified as 0, it will not return until either	718	If the flags argument is specified as C<0>, it will not return until
275	no event watchers are active anymore or C<ev_unloop> was called.	719	either no event watchers are active anymore or C<ev_unloop> was called.
		720
		721	Please note that an explicit C<ev_unloop> is usually better than
		722	relying on all watchers to be stopped when deciding when a program has
		723	finished (especially in interactive programs), but having a program
		724	that automatically loops as long as it has to and no longer by virtue
		725	of relying on its watchers stopping correctly, that is truly a thing of
		726	beauty.
276		727
277	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	728	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
278	those events and any outstanding ones, but will not block your process in	729	those events and any already outstanding ones, but will not block your
279	case there are no events and will return after one iteration of the loop.	730	process in case there are no events and will return after one iteration of
		731	the loop.
280		732
281	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	733	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
282	neccessary) and will handle those and any outstanding ones. It will block	734	necessary) and will handle those and any already outstanding ones. It
283	your process until at least one new event arrives, and will return after	735	will block your process until at least one new event arrives (which could
		736	be an event internal to libev itself, so there is no guarantee that a
		737	user-registered callback will be called), and will return after one
284	one iteration of the loop.	738	iteration of the loop.
285		739
286	This flags value could be used to implement alternative looping	740	This is useful if you are waiting for some external event in conjunction
287	constructs, but the C<prepare> and C<check> watchers provide a better and	741	with something not expressible using other libev watchers (i.e. "roll your
288	more generic mechanism.	742	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		743	usually a better approach for this kind of thing.
289		744
290	Here are the gory details of what ev_loop does:	745	Here are the gory details of what C<ev_loop> does:
291		746
292	1. If there are no active watchers (reference count is zero), return.	747	- Before the first iteration, call any pending watchers.
		748	* If EVFLAG_FORKCHECK was used, check for a fork.
		749	- If a fork was detected (by any means), queue and call all fork watchers.
293	2. Queue and immediately call all prepare watchers.	750	- Queue and call all prepare watchers.
294	3. If we have been forked, recreate the kernel state.	751	- If we have been forked, detach and recreate the kernel state
		752	as to not disturb the other process.
295	4. Update the kernel state with all outstanding changes.	753	- Update the kernel state with all outstanding changes.
296	5. Update the "event loop time".	754	- Update the "event loop time" (ev_now ()).
297	6. Calculate for how long to block.	755	- Calculate for how long to sleep or block, if at all
		756	(active idle watchers, EVLOOP_NONBLOCK or not having
		757	any active watchers at all will result in not sleeping).
		758	- Sleep if the I/O and timer collect interval say so.
298	7. Block the process, waiting for events.	759	- Block the process, waiting for any events.
		760	- Queue all outstanding I/O (fd) events.
299	8. Update the "event loop time" and do time jump handling.	761	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
300	9. Queue all outstanding timers.	762	- Queue all expired timers.
301	10. Queue all outstanding periodics.	763	- Queue all expired periodics.
302	11. If no events are pending now, queue all idle watchers.	764	- Unless any events are pending now, queue all idle watchers.
303	12. Queue all check watchers.	765	- Queue all check watchers.
304	13. Call all queued watchers in reverse order (i.e. check watchers first).	766	- Call all queued watchers in reverse order (i.e. check watchers first).
		767	Signals and child watchers are implemented as I/O watchers, and will
		768	be handled here by queueing them when their watcher gets executed.
305	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	769	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
306	was used, return, otherwise continue with step #1.	770	were used, or there are no active watchers, return, otherwise
		771	continue with step *.
		772
		773	Example: Queue some jobs and then loop until no events are outstanding
		774	anymore.
		775
		776	... queue jobs here, make sure they register event watchers as long
		777	... as they still have work to do (even an idle watcher will do..)
		778	ev_loop (my_loop, 0);
		779	... jobs done or somebody called unloop. yeah!
307		780
308	=item ev_unloop (loop, how)	781	=item ev_unloop (loop, how)
309		782
310	Can be used to make a call to C<ev_loop> return early (but only after it	783	Can be used to make a call to C<ev_loop> return early (but only after it
311	has processed all outstanding events). The C<how> argument must be either	784	has processed all outstanding events). The C<how> argument must be either
312	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	785	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
313	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	786	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
314		787
		788	This "unloop state" will be cleared when entering C<ev_loop> again.
		789
		790	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
		791
315	=item ev_ref (loop)	792	=item ev_ref (loop)
316		793
317	=item ev_unref (loop)	794	=item ev_unref (loop)
318		795
319	Ref/unref can be used to add or remove a reference count on the event	796	Ref/unref can be used to add or remove a reference count on the event
320	loop: Every watcher keeps one reference, and as long as the reference	797	loop: Every watcher keeps one reference, and as long as the reference
321	count is nonzero, C<ev_loop> will not return on its own. If you have	798	count is nonzero, C<ev_loop> will not return on its own.
322	a watcher you never unregister that should not keep C<ev_loop> from	799
323	returning, ev_unref() after starting, and ev_ref() before stopping it. For	800	This is useful when you have a watcher that you never intend to
		801	unregister, but that nevertheless should not keep C<ev_loop> from
		802	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		803	before stopping it.
		804
324	example, libev itself uses this for its internal signal pipe: It is not	805	As an example, libev itself uses this for its internal signal pipe: It
325	visible to the libev user and should not keep C<ev_loop> from exiting if	806	is not visible to the libev user and should not keep C<ev_loop> from
326	no event watchers registered by it are active. It is also an excellent	807	exiting if no event watchers registered by it are active. It is also an
327	way to do this for generic recurring timers or from within third-party	808	excellent way to do this for generic recurring timers or from within
328	libraries. Just remember to I<unref after start> and I<ref before stop>.	809	third-party libraries. Just remember to I<unref after start> and I<ref
		810	before stop> (but only if the watcher wasn't active before, or was active
		811	before, respectively. Note also that libev might stop watchers itself
		812	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		813	in the callback).
		814
		815	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		816	running when nothing else is active.
		817
		818	ev_signal exitsig;
		819	ev_signal_init (&exitsig, sig_cb, SIGINT);
		820	ev_signal_start (loop, &exitsig);
		821	evf_unref (loop);
		822
		823	Example: For some weird reason, unregister the above signal handler again.
		824
		825	ev_ref (loop);
		826	ev_signal_stop (loop, &exitsig);
		827
		828	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		829
		830	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		831
		832	These advanced functions influence the time that libev will spend waiting
		833	for events. Both time intervals are by default C<0>, meaning that libev
		834	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		835	latency.
		836
		837	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		838	allows libev to delay invocation of I/O and timer/periodic callbacks
		839	to increase efficiency of loop iterations (or to increase power-saving
		840	opportunities).
		841
		842	The idea is that sometimes your program runs just fast enough to handle
		843	one (or very few) event(s) per loop iteration. While this makes the
		844	program responsive, it also wastes a lot of CPU time to poll for new
		845	events, especially with backends like C<select ()> which have a high
		846	overhead for the actual polling but can deliver many events at once.
		847
		848	By setting a higher I<io collect interval> you allow libev to spend more
		849	time collecting I/O events, so you can handle more events per iteration,
		850	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		851	C<ev_timer>) will be not affected. Setting this to a non-null value will
		852	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		853	sleep time ensures that libev will not poll for I/O events more often then
		854	once per this interval, on average.
		855
		856	Likewise, by setting a higher I<timeout collect interval> you allow libev
		857	to spend more time collecting timeouts, at the expense of increased
		858	latency/jitter/inexactness (the watcher callback will be called
		859	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		860	value will not introduce any overhead in libev.
		861
		862	Many (busy) programs can usually benefit by setting the I/O collect
		863	interval to a value near C<0.1> or so, which is often enough for
		864	interactive servers (of course not for games), likewise for timeouts. It
		865	usually doesn't make much sense to set it to a lower value than C<0.01>,
		866	as this approaches the timing granularity of most systems. Note that if
		867	you do transactions with the outside world and you can't increase the
		868	parallelity, then this setting will limit your transaction rate (if you
		869	need to poll once per transaction and the I/O collect interval is 0.01,
		870	then you can't do more than 100 transations per second).
		871
		872	Setting the I<timeout collect interval> can improve the opportunity for
		873	saving power, as the program will "bundle" timer callback invocations that
		874	are "near" in time together, by delaying some, thus reducing the number of
		875	times the process sleeps and wakes up again. Another useful technique to
		876	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		877	they fire on, say, one-second boundaries only.
		878
		879	Example: we only need 0.1s timeout granularity, and we wish not to poll
		880	more often than 100 times per second:
		881
		882	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		883	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		884
		885	=item ev_invoke_pending (loop)
		886
		887	This call will simply invoke all pending watchers while resetting their
		888	pending state. Normally, C<ev_loop> does this automatically when required,
		889	but when overriding the invoke callback this call comes handy.
		890
		891	=item int ev_pending_count (loop)
		892
		893	Returns the number of pending watchers - zero indicates that no watchers
		894	are pending.
		895
		896	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		897
		898	This overrides the invoke pending functionality of the loop: Instead of
		899	invoking all pending watchers when there are any, C<ev_loop> will call
		900	this callback instead. This is useful, for example, when you want to
		901	invoke the actual watchers inside another context (another thread etc.).
		902
		903	If you want to reset the callback, use C<ev_invoke_pending> as new
		904	callback.
		905
		906	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		907
		908	Sometimes you want to share the same loop between multiple threads. This
		909	can be done relatively simply by putting mutex_lock/unlock calls around
		910	each call to a libev function.
		911
		912	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		913	wait for it to return. One way around this is to wake up the loop via
		914	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		915	and I<acquire> callbacks on the loop.
		916
		917	When set, then C<release> will be called just before the thread is
		918	suspended waiting for new events, and C<acquire> is called just
		919	afterwards.
		920
		921	Ideally, C<release> will just call your mutex_unlock function, and
		922	C<acquire> will just call the mutex_lock function again.
		923
		924	While event loop modifications are allowed between invocations of
		925	C<release> and C<acquire> (that's their only purpose after all), no
		926	modifications done will affect the event loop, i.e. adding watchers will
		927	have no effect on the set of file descriptors being watched, or the time
		928	waited. Use an C<ev_async> watcher to wake up C<ev_loop> when you want it
		929	to take note of any changes you made.
		930
		931	In theory, threads executing C<ev_loop> will be async-cancel safe between
		932	invocations of C<release> and C<acquire>.
		933
		934	See also the locking example in the C<THREADS> section later in this
		935	document.
		936
		937	=item ev_set_userdata (loop, void *data)
		938
		939	=item ev_userdata (loop)
		940
		941	Set and retrieve a single C<void *> associated with a loop. When
		942	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		943	C<0.>
		944
		945	These two functions can be used to associate arbitrary data with a loop,
		946	and are intended solely for the C<invoke_pending_cb>, C<release> and
		947	C<acquire> callbacks described above, but of course can be (ab-)used for
		948	any other purpose as well.
		949
		950	=item ev_loop_verify (loop)
		951
		952	This function only does something when C<EV_VERIFY> support has been
		953	compiled in, which is the default for non-minimal builds. It tries to go
		954	through all internal structures and checks them for validity. If anything
		955	is found to be inconsistent, it will print an error message to standard
		956	error and call C<abort ()>.
		957
		958	This can be used to catch bugs inside libev itself: under normal
		959	circumstances, this function will never abort as of course libev keeps its
		960	data structures consistent.
329		961
330	=back	962	=back
331		963
		964
332	=head1 ANATOMY OF A WATCHER	965	=head1 ANATOMY OF A WATCHER
		966
		967	In the following description, uppercase C<TYPE> in names stands for the
		968	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		969	watchers and C<ev_io_start> for I/O watchers.
333		970
334	A watcher is a structure that you create and register to record your	971	A watcher is a structure that you create and register to record your
335	interest in some event. For instance, if you want to wait for STDIN to	972	interest in some event. For instance, if you want to wait for STDIN to
336	become readable, you would create an C<ev_io> watcher for that:	973	become readable, you would create an C<ev_io> watcher for that:
337		974
338	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	975	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
339	{	976	{
340	ev_io_stop (w);	977	ev_io_stop (w);
341	ev_unloop (loop, EVUNLOOP_ALL);	978	ev_unloop (loop, EVUNLOOP_ALL);
342	}	979	}
343		980
344	struct ev_loop *loop = ev_default_loop (0);	981	struct ev_loop *loop = ev_default_loop (0);
		982
345	struct ev_io stdin_watcher;	983	ev_io stdin_watcher;
		984
346	ev_init (&stdin_watcher, my_cb);	985	ev_init (&stdin_watcher, my_cb);
347	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	986	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
348	ev_io_start (loop, &stdin_watcher);	987	ev_io_start (loop, &stdin_watcher);
		988
349	ev_loop (loop, 0);	989	ev_loop (loop, 0);
350		990
351	As you can see, you are responsible for allocating the memory for your	991	As you can see, you are responsible for allocating the memory for your
352	watcher structures (and it is usually a bad idea to do this on the stack,	992	watcher structures (and it is I<usually> a bad idea to do this on the
353	although this can sometimes be quite valid).	993	stack).
		994
		995	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		996	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
354		997
355	Each watcher structure must be initialised by a call to C<ev_init	998	Each watcher structure must be initialised by a call to C<ev_init
356	(watcher *, callback)>, which expects a callback to be provided. This	999	(watcher *, callback)>, which expects a callback to be provided. This
357	callback gets invoked each time the event occurs (or, in the case of io	1000	callback gets invoked each time the event occurs (or, in the case of I/O
358	watchers, each time the event loop detects that the file descriptor given	1001	watchers, each time the event loop detects that the file descriptor given
359	is readable and/or writable).	1002	is readable and/or writable).
360		1003
361	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1004	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
362	with arguments specific to this watcher type. There is also a macro	1005	macro to configure it, with arguments specific to the watcher type. There
363	to combine initialisation and setting in one call: C<< ev_<type>_init	1006	is also a macro to combine initialisation and setting in one call: C<<
364	(watcher *, callback, ...) >>.	1007	ev_TYPE_init (watcher *, callback, ...) >>.
365		1008
366	To make the watcher actually watch out for events, you have to start it	1009	To make the watcher actually watch out for events, you have to start it
367	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1010	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
368	*) >>), and you can stop watching for events at any time by calling the	1011	*) >>), and you can stop watching for events at any time by calling the
369	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1012	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
370		1013
371	As long as your watcher is active (has been started but not stopped) you	1014	As long as your watcher is active (has been started but not stopped) you
372	must not touch the values stored in it. Most specifically you must never	1015	must not touch the values stored in it. Most specifically you must never
373	reinitialise it or call its set method.	1016	reinitialise it or call its C<ev_TYPE_set> macro.
374
375	You can check whether an event is active by calling the C<ev_is_active
376	(watcher *)> macro. To see whether an event is outstanding (but the
377	callback for it has not been called yet) you can use the C<ev_is_pending
378	(watcher *)> macro.
379		1017
380	Each and every callback receives the event loop pointer as first, the	1018	Each and every callback receives the event loop pointer as first, the
381	registered watcher structure as second, and a bitset of received events as	1019	registered watcher structure as second, and a bitset of received events as
382	third argument.	1020	third argument.
383		1021
…		…
407	The signal specified in the C<ev_signal> watcher has been received by a thread.	1045	The signal specified in the C<ev_signal> watcher has been received by a thread.
408		1046
409	=item C<EV_CHILD>	1047	=item C<EV_CHILD>
410		1048
411	The pid specified in the C<ev_child> watcher has received a status change.	1049	The pid specified in the C<ev_child> watcher has received a status change.
		1050
		1051	=item C<EV_STAT>
		1052
		1053	The path specified in the C<ev_stat> watcher changed its attributes somehow.
412		1054
413	=item C<EV_IDLE>	1055	=item C<EV_IDLE>
414		1056
415	The C<ev_idle> watcher has determined that you have nothing better to do.	1057	The C<ev_idle> watcher has determined that you have nothing better to do.
416		1058
…		…
424	received events. Callbacks of both watcher types can start and stop as	1066	received events. Callbacks of both watcher types can start and stop as
425	many watchers as they want, and all of them will be taken into account	1067	many watchers as they want, and all of them will be taken into account
426	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1068	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
427	C<ev_loop> from blocking).	1069	C<ev_loop> from blocking).
428		1070
		1071	=item C<EV_EMBED>
		1072
		1073	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1074
		1075	=item C<EV_FORK>
		1076
		1077	The event loop has been resumed in the child process after fork (see
		1078	C<ev_fork>).
		1079
		1080	=item C<EV_ASYNC>
		1081
		1082	The given async watcher has been asynchronously notified (see C<ev_async>).
		1083
		1084	=item C<EV_CUSTOM>
		1085
		1086	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1087	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1088
429	=item C<EV_ERROR>	1089	=item C<EV_ERROR>
430		1090
431	An unspecified error has occured, the watcher has been stopped. This might	1091	An unspecified error has occurred, the watcher has been stopped. This might
432	happen because the watcher could not be properly started because libev	1092	happen because the watcher could not be properly started because libev
433	ran out of memory, a file descriptor was found to be closed or any other	1093	ran out of memory, a file descriptor was found to be closed or any other
		1094	problem. Libev considers these application bugs.
		1095
434	problem. You best act on it by reporting the problem and somehow coping	1096	You best act on it by reporting the problem and somehow coping with the
435	with the watcher being stopped.	1097	watcher being stopped. Note that well-written programs should not receive
		1098	an error ever, so when your watcher receives it, this usually indicates a
		1099	bug in your program.
436		1100
437	Libev will usually signal a few "dummy" events together with an error,	1101	Libev will usually signal a few "dummy" events together with an error, for
438	for example it might indicate that a fd is readable or writable, and if	1102	example it might indicate that a fd is readable or writable, and if your
439	your callbacks is well-written it can just attempt the operation and cope	1103	callbacks is well-written it can just attempt the operation and cope with
440	with the error from read() or write(). This will not work in multithreaded	1104	the error from read() or write(). This will not work in multi-threaded
441	programs, though, so beware.	1105	programs, though, as the fd could already be closed and reused for another
		1106	thing, so beware.
442		1107
443	=back	1108	=back
444		1109
		1110	=head2 GENERIC WATCHER FUNCTIONS
		1111
		1112	=over 4
		1113
		1114	=item C<ev_init> (ev_TYPE *watcher, callback)
		1115
		1116	This macro initialises the generic portion of a watcher. The contents
		1117	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1118	the generic parts of the watcher are initialised, you I<need> to call
		1119	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1120	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1121	which rolls both calls into one.
		1122
		1123	You can reinitialise a watcher at any time as long as it has been stopped
		1124	(or never started) and there are no pending events outstanding.
		1125
		1126	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1127	int revents)>.
		1128
		1129	Example: Initialise an C<ev_io> watcher in two steps.
		1130
		1131	ev_io w;
		1132	ev_init (&w, my_cb);
		1133	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1134
		1135	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1136
		1137	This macro initialises the type-specific parts of a watcher. You need to
		1138	call C<ev_init> at least once before you call this macro, but you can
		1139	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1140	macro on a watcher that is active (it can be pending, however, which is a
		1141	difference to the C<ev_init> macro).
		1142
		1143	Although some watcher types do not have type-specific arguments
		1144	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1145
		1146	See C<ev_init>, above, for an example.
		1147
		1148	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1149
		1150	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1151	calls into a single call. This is the most convenient method to initialise
		1152	a watcher. The same limitations apply, of course.
		1153
		1154	Example: Initialise and set an C<ev_io> watcher in one step.
		1155
		1156	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1157
		1158	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1159
		1160	Starts (activates) the given watcher. Only active watchers will receive
		1161	events. If the watcher is already active nothing will happen.
		1162
		1163	Example: Start the C<ev_io> watcher that is being abused as example in this
		1164	whole section.
		1165
		1166	ev_io_start (EV_DEFAULT_UC, &w);
		1167
		1168	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1169
		1170	Stops the given watcher if active, and clears the pending status (whether
		1171	the watcher was active or not).
		1172
		1173	It is possible that stopped watchers are pending - for example,
		1174	non-repeating timers are being stopped when they become pending - but
		1175	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1176	pending. If you want to free or reuse the memory used by the watcher it is
		1177	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1178
		1179	=item bool ev_is_active (ev_TYPE *watcher)
		1180
		1181	Returns a true value iff the watcher is active (i.e. it has been started
		1182	and not yet been stopped). As long as a watcher is active you must not modify
		1183	it.
		1184
		1185	=item bool ev_is_pending (ev_TYPE *watcher)
		1186
		1187	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1188	events but its callback has not yet been invoked). As long as a watcher
		1189	is pending (but not active) you must not call an init function on it (but
		1190	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1191	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1192	it).
		1193
		1194	=item callback ev_cb (ev_TYPE *watcher)
		1195
		1196	Returns the callback currently set on the watcher.
		1197
		1198	=item ev_cb_set (ev_TYPE *watcher, callback)
		1199
		1200	Change the callback. You can change the callback at virtually any time
		1201	(modulo threads).
		1202
		1203	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1204
		1205	=item int ev_priority (ev_TYPE *watcher)
		1206
		1207	Set and query the priority of the watcher. The priority is a small
		1208	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1209	(default: C<-2>). Pending watchers with higher priority will be invoked
		1210	before watchers with lower priority, but priority will not keep watchers
		1211	from being executed (except for C<ev_idle> watchers).
		1212
		1213	If you need to suppress invocation when higher priority events are pending
		1214	you need to look at C<ev_idle> watchers, which provide this functionality.
		1215
		1216	You I<must not> change the priority of a watcher as long as it is active or
		1217	pending.
		1218
		1219	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1220	fine, as long as you do not mind that the priority value you query might
		1221	or might not have been clamped to the valid range.
		1222
		1223	The default priority used by watchers when no priority has been set is
		1224	always C<0>, which is supposed to not be too high and not be too low :).
		1225
		1226	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1227	priorities.
		1228
		1229	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1230
		1231	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1232	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1233	can deal with that fact, as both are simply passed through to the
		1234	callback.
		1235
		1236	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1237
		1238	If the watcher is pending, this function clears its pending status and
		1239	returns its C<revents> bitset (as if its callback was invoked). If the
		1240	watcher isn't pending it does nothing and returns C<0>.
		1241
		1242	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1243	callback to be invoked, which can be accomplished with this function.
		1244
		1245	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1246
		1247	Feeds the given event set into the event loop, as if the specified event
		1248	had happened for the specified watcher (which must be a pointer to an
		1249	initialised but not necessarily started event watcher). Obviously you must
		1250	not free the watcher as long as it has pending events.
		1251
		1252	Stopping the watcher, letting libev invoke it, or calling
		1253	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1254	not started in the first place.
		1255
		1256	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1257	functions that do not need a watcher.
		1258
		1259	=back
		1260
		1261
445	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1262	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
446		1263
447	Each watcher has, by default, a member C<void *data> that you can change	1264	Each watcher has, by default, a member C<void *data> that you can change
448	and read at any time, libev will completely ignore it. This can be used	1265	and read at any time: libev will completely ignore it. This can be used
449	to associate arbitrary data with your watcher. If you need more data and	1266	to associate arbitrary data with your watcher. If you need more data and
450	don't want to allocate memory and store a pointer to it in that data	1267	don't want to allocate memory and store a pointer to it in that data
451	member, you can also "subclass" the watcher type and provide your own	1268	member, you can also "subclass" the watcher type and provide your own
452	data:	1269	data:
453		1270
454	struct my_io	1271	struct my_io
455	{	1272	{
456	struct ev_io io;	1273	ev_io io;
457	int otherfd;	1274	int otherfd;
458	void *somedata;	1275	void *somedata;
459	struct whatever *mostinteresting;	1276	struct whatever *mostinteresting;
460	}	1277	};
		1278
		1279	...
		1280	struct my_io w;
		1281	ev_io_init (&w.io, my_cb, fd, EV_READ);
461		1282
462	And since your callback will be called with a pointer to the watcher, you	1283	And since your callback will be called with a pointer to the watcher, you
463	can cast it back to your own type:	1284	can cast it back to your own type:
464		1285
465	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1286	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
466	{	1287	{
467	struct my_io *w = (struct my_io *)w_;	1288	struct my_io *w = (struct my_io *)w_;
468	...	1289	...
469	}	1290	}
470		1291
471	More interesting and less C-conformant ways of catsing your callback type	1292	More interesting and less C-conformant ways of casting your callback type
472	have been omitted....	1293	instead have been omitted.
		1294
		1295	Another common scenario is to use some data structure with multiple
		1296	embedded watchers:
		1297
		1298	struct my_biggy
		1299	{
		1300	int some_data;
		1301	ev_timer t1;
		1302	ev_timer t2;
		1303	}
		1304
		1305	In this case getting the pointer to C<my_biggy> is a bit more
		1306	complicated: Either you store the address of your C<my_biggy> struct
		1307	in the C<data> member of the watcher (for woozies), or you need to use
		1308	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1309	programmers):
		1310
		1311	#include <stddef.h>
		1312
		1313	static void
		1314	t1_cb (EV_P_ ev_timer *w, int revents)
		1315	{
		1316	struct my_biggy big = (struct my_biggy *)
		1317	(((char *)w) - offsetof (struct my_biggy, t1));
		1318	}
		1319
		1320	static void
		1321	t2_cb (EV_P_ ev_timer *w, int revents)
		1322	{
		1323	struct my_biggy big = (struct my_biggy *)
		1324	(((char *)w) - offsetof (struct my_biggy, t2));
		1325	}
		1326
		1327	=head2 WATCHER PRIORITY MODELS
		1328
		1329	Many event loops support I<watcher priorities>, which are usually small
		1330	integers that influence the ordering of event callback invocation
		1331	between watchers in some way, all else being equal.
		1332
		1333	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1334	description for the more technical details such as the actual priority
		1335	range.
		1336
		1337	There are two common ways how these these priorities are being interpreted
		1338	by event loops:
		1339
		1340	In the more common lock-out model, higher priorities "lock out" invocation
		1341	of lower priority watchers, which means as long as higher priority
		1342	watchers receive events, lower priority watchers are not being invoked.
		1343
		1344	The less common only-for-ordering model uses priorities solely to order
		1345	callback invocation within a single event loop iteration: Higher priority
		1346	watchers are invoked before lower priority ones, but they all get invoked
		1347	before polling for new events.
		1348
		1349	Libev uses the second (only-for-ordering) model for all its watchers
		1350	except for idle watchers (which use the lock-out model).
		1351
		1352	The rationale behind this is that implementing the lock-out model for
		1353	watchers is not well supported by most kernel interfaces, and most event
		1354	libraries will just poll for the same events again and again as long as
		1355	their callbacks have not been executed, which is very inefficient in the
		1356	common case of one high-priority watcher locking out a mass of lower
		1357	priority ones.
		1358
		1359	Static (ordering) priorities are most useful when you have two or more
		1360	watchers handling the same resource: a typical usage example is having an
		1361	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1362	timeouts. Under load, data might be received while the program handles
		1363	other jobs, but since timers normally get invoked first, the timeout
		1364	handler will be executed before checking for data. In that case, giving
		1365	the timer a lower priority than the I/O watcher ensures that I/O will be
		1366	handled first even under adverse conditions (which is usually, but not
		1367	always, what you want).
		1368
		1369	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1370	will only be executed when no same or higher priority watchers have
		1371	received events, they can be used to implement the "lock-out" model when
		1372	required.
		1373
		1374	For example, to emulate how many other event libraries handle priorities,
		1375	you can associate an C<ev_idle> watcher to each such watcher, and in
		1376	the normal watcher callback, you just start the idle watcher. The real
		1377	processing is done in the idle watcher callback. This causes libev to
		1378	continously poll and process kernel event data for the watcher, but when
		1379	the lock-out case is known to be rare (which in turn is rare :), this is
		1380	workable.
		1381
		1382	Usually, however, the lock-out model implemented that way will perform
		1383	miserably under the type of load it was designed to handle. In that case,
		1384	it might be preferable to stop the real watcher before starting the
		1385	idle watcher, so the kernel will not have to process the event in case
		1386	the actual processing will be delayed for considerable time.
		1387
		1388	Here is an example of an I/O watcher that should run at a strictly lower
		1389	priority than the default, and which should only process data when no
		1390	other events are pending:
		1391
		1392	ev_idle idle; // actual processing watcher
		1393	ev_io io; // actual event watcher
		1394
		1395	static void
		1396	io_cb (EV_P_ ev_io *w, int revents)
		1397	{
		1398	// stop the I/O watcher, we received the event, but
		1399	// are not yet ready to handle it.
		1400	ev_io_stop (EV_A_ w);
		1401
		1402	// start the idle watcher to ahndle the actual event.
		1403	// it will not be executed as long as other watchers
		1404	// with the default priority are receiving events.
		1405	ev_idle_start (EV_A_ &idle);
		1406	}
		1407
		1408	static void
		1409	idle_cb (EV_P_ ev_idle *w, int revents)
		1410	{
		1411	// actual processing
		1412	read (STDIN_FILENO, ...);
		1413
		1414	// have to start the I/O watcher again, as
		1415	// we have handled the event
		1416	ev_io_start (EV_P_ &io);
		1417	}
		1418
		1419	// initialisation
		1420	ev_idle_init (&idle, idle_cb);
		1421	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1422	ev_io_start (EV_DEFAULT_ &io);
		1423
		1424	In the "real" world, it might also be beneficial to start a timer, so that
		1425	low-priority connections can not be locked out forever under load. This
		1426	enables your program to keep a lower latency for important connections
		1427	during short periods of high load, while not completely locking out less
		1428	important ones.
473		1429
474		1430
475	=head1 WATCHER TYPES	1431	=head1 WATCHER TYPES
476		1432
477	This section describes each watcher in detail, but will not repeat	1433	This section describes each watcher in detail, but will not repeat
478	information given in the last section.	1434	information given in the last section. Any initialisation/set macros,
		1435	functions and members specific to the watcher type are explained.
479		1436
		1437	Members are additionally marked with either I<[read-only]>, meaning that,
		1438	while the watcher is active, you can look at the member and expect some
		1439	sensible content, but you must not modify it (you can modify it while the
		1440	watcher is stopped to your hearts content), or I<[read-write]>, which
		1441	means you can expect it to have some sensible content while the watcher
		1442	is active, but you can also modify it. Modifying it may not do something
		1443	sensible or take immediate effect (or do anything at all), but libev will
		1444	not crash or malfunction in any way.
		1445
		1446
480	=head2 C<ev_io> - is this file descriptor readable or writable	1447	=head2 C<ev_io> - is this file descriptor readable or writable?
481		1448
482	I/O watchers check whether a file descriptor is readable or writable	1449	I/O watchers check whether a file descriptor is readable or writable
483	in each iteration of the event loop (This behaviour is called	1450	in each iteration of the event loop, or, more precisely, when reading
484	level-triggering because you keep receiving events as long as the	1451	would not block the process and writing would at least be able to write
485	condition persists. Remember you can stop the watcher if you don't want to	1452	some data. This behaviour is called level-triggering because you keep
486	act on the event and neither want to receive future events).	1453	receiving events as long as the condition persists. Remember you can stop
		1454	the watcher if you don't want to act on the event and neither want to
		1455	receive future events.
487		1456
488	In general you can register as many read and/or write event watchers per	1457	In general you can register as many read and/or write event watchers per
489	fd as you want (as long as you don't confuse yourself). Setting all file	1458	fd as you want (as long as you don't confuse yourself). Setting all file
490	descriptors to non-blocking mode is also usually a good idea (but not	1459	descriptors to non-blocking mode is also usually a good idea (but not
491	required if you know what you are doing).	1460	required if you know what you are doing).
492		1461
493	You have to be careful with dup'ed file descriptors, though. Some backends	1462	If you cannot use non-blocking mode, then force the use of a
494	(the linux epoll backend is a notable example) cannot handle dup'ed file	1463	known-to-be-good backend (at the time of this writing, this includes only
495	descriptors correctly if you register interest in two or more fds pointing	1464	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
496	to the same underlying file/socket etc. description (that is, they share	1465	descriptors for which non-blocking operation makes no sense (such as
497	the same underlying "file open").	1466	files) - libev doesn't guarentee any specific behaviour in that case.
498		1467
499	If you must do this, then force the use of a known-to-be-good backend	1468	Another thing you have to watch out for is that it is quite easy to
500	(at the time of this writing, this includes only EVMETHOD_SELECT and	1469	receive "spurious" readiness notifications, that is your callback might
501	EVMETHOD_POLL).	1470	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1471	because there is no data. Not only are some backends known to create a
		1472	lot of those (for example Solaris ports), it is very easy to get into
		1473	this situation even with a relatively standard program structure. Thus
		1474	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1475	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1476
		1477	If you cannot run the fd in non-blocking mode (for example you should
		1478	not play around with an Xlib connection), then you have to separately
		1479	re-test whether a file descriptor is really ready with a known-to-be good
		1480	interface such as poll (fortunately in our Xlib example, Xlib already
		1481	does this on its own, so its quite safe to use). Some people additionally
		1482	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1483	indefinitely.
		1484
		1485	But really, best use non-blocking mode.
		1486
		1487	=head3 The special problem of disappearing file descriptors
		1488
		1489	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1490	descriptor (either due to calling C<close> explicitly or any other means,
		1491	such as C<dup2>). The reason is that you register interest in some file
		1492	descriptor, but when it goes away, the operating system will silently drop
		1493	this interest. If another file descriptor with the same number then is
		1494	registered with libev, there is no efficient way to see that this is, in
		1495	fact, a different file descriptor.
		1496
		1497	To avoid having to explicitly tell libev about such cases, libev follows
		1498	the following policy: Each time C<ev_io_set> is being called, libev
		1499	will assume that this is potentially a new file descriptor, otherwise
		1500	it is assumed that the file descriptor stays the same. That means that
		1501	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1502	descriptor even if the file descriptor number itself did not change.
		1503
		1504	This is how one would do it normally anyway, the important point is that
		1505	the libev application should not optimise around libev but should leave
		1506	optimisations to libev.
		1507
		1508	=head3 The special problem of dup'ed file descriptors
		1509
		1510	Some backends (e.g. epoll), cannot register events for file descriptors,
		1511	but only events for the underlying file descriptions. That means when you
		1512	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1513	events for them, only one file descriptor might actually receive events.
		1514
		1515	There is no workaround possible except not registering events
		1516	for potentially C<dup ()>'ed file descriptors, or to resort to
		1517	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1518
		1519	=head3 The special problem of fork
		1520
		1521	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1522	useless behaviour. Libev fully supports fork, but needs to be told about
		1523	it in the child.
		1524
		1525	To support fork in your programs, you either have to call
		1526	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1527	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1528	C<EVBACKEND_POLL>.
		1529
		1530	=head3 The special problem of SIGPIPE
		1531
		1532	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1533	when writing to a pipe whose other end has been closed, your program gets
		1534	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1535	this is sensible behaviour, for daemons, this is usually undesirable.
		1536
		1537	So when you encounter spurious, unexplained daemon exits, make sure you
		1538	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1539	somewhere, as that would have given you a big clue).
		1540
		1541
		1542	=head3 Watcher-Specific Functions
502		1543
503	=over 4	1544	=over 4
504		1545
505	=item ev_io_init (ev_io *, callback, int fd, int events)	1546	=item ev_io_init (ev_io *, callback, int fd, int events)
506		1547
507	=item ev_io_set (ev_io *, int fd, int events)	1548	=item ev_io_set (ev_io *, int fd, int events)
508		1549
509	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1550	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
510	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1551	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
511	EV_WRITE> to receive the given events.	1552	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1553
		1554	=item int fd [read-only]
		1555
		1556	The file descriptor being watched.
		1557
		1558	=item int events [read-only]
		1559
		1560	The events being watched.
512		1561
513	=back	1562	=back
514		1563
		1564	=head3 Examples
		1565
		1566	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1567	readable, but only once. Since it is likely line-buffered, you could
		1568	attempt to read a whole line in the callback.
		1569
		1570	static void
		1571	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1572	{
		1573	ev_io_stop (loop, w);
		1574	.. read from stdin here (or from w->fd) and handle any I/O errors
		1575	}
		1576
		1577	...
		1578	struct ev_loop *loop = ev_default_init (0);
		1579	ev_io stdin_readable;
		1580	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1581	ev_io_start (loop, &stdin_readable);
		1582	ev_loop (loop, 0);
		1583
		1584
515	=head2 C<ev_timer> - relative and optionally recurring timeouts	1585	=head2 C<ev_timer> - relative and optionally repeating timeouts
516		1586
517	Timer watchers are simple relative timers that generate an event after a	1587	Timer watchers are simple relative timers that generate an event after a
518	given time, and optionally repeating in regular intervals after that.	1588	given time, and optionally repeating in regular intervals after that.
519		1589
520	The timers are based on real time, that is, if you register an event that	1590	The timers are based on real time, that is, if you register an event that
521	times out after an hour and you reset your system clock to last years	1591	times out after an hour and you reset your system clock to January last
522	time, it will still time out after (roughly) and hour. "Roughly" because	1592	year, it will still time out after (roughly) one hour. "Roughly" because
523	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1593	detecting time jumps is hard, and some inaccuracies are unavoidable (the
524	monotonic clock option helps a lot here).	1594	monotonic clock option helps a lot here).
		1595
		1596	The callback is guaranteed to be invoked only I<after> its timeout has
		1597	passed (not I<at>, so on systems with very low-resolution clocks this
		1598	might introduce a small delay). If multiple timers become ready during the
		1599	same loop iteration then the ones with earlier time-out values are invoked
		1600	before ones of the same priority with later time-out values (but this is
		1601	no longer true when a callback calls C<ev_loop> recursively).
		1602
		1603	=head3 Be smart about timeouts
		1604
		1605	Many real-world problems involve some kind of timeout, usually for error
		1606	recovery. A typical example is an HTTP request - if the other side hangs,
		1607	you want to raise some error after a while.
		1608
		1609	What follows are some ways to handle this problem, from obvious and
		1610	inefficient to smart and efficient.
		1611
		1612	In the following, a 60 second activity timeout is assumed - a timeout that
		1613	gets reset to 60 seconds each time there is activity (e.g. each time some
		1614	data or other life sign was received).
		1615
		1616	=over 4
		1617
		1618	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1619
		1620	This is the most obvious, but not the most simple way: In the beginning,
		1621	start the watcher:
		1622
		1623	ev_timer_init (timer, callback, 60., 0.);
		1624	ev_timer_start (loop, timer);
		1625
		1626	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1627	and start it again:
		1628
		1629	ev_timer_stop (loop, timer);
		1630	ev_timer_set (timer, 60., 0.);
		1631	ev_timer_start (loop, timer);
		1632
		1633	This is relatively simple to implement, but means that each time there is
		1634	some activity, libev will first have to remove the timer from its internal
		1635	data structure and then add it again. Libev tries to be fast, but it's
		1636	still not a constant-time operation.
		1637
		1638	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1639
		1640	This is the easiest way, and involves using C<ev_timer_again> instead of
		1641	C<ev_timer_start>.
		1642
		1643	To implement this, configure an C<ev_timer> with a C<repeat> value
		1644	of C<60> and then call C<ev_timer_again> at start and each time you
		1645	successfully read or write some data. If you go into an idle state where
		1646	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1647	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1648
		1649	That means you can ignore both the C<ev_timer_start> function and the
		1650	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1651	member and C<ev_timer_again>.
		1652
		1653	At start:
		1654
		1655	ev_init (timer, callback);
		1656	timer->repeat = 60.;
		1657	ev_timer_again (loop, timer);
		1658
		1659	Each time there is some activity:
		1660
		1661	ev_timer_again (loop, timer);
		1662
		1663	It is even possible to change the time-out on the fly, regardless of
		1664	whether the watcher is active or not:
		1665
		1666	timer->repeat = 30.;
		1667	ev_timer_again (loop, timer);
		1668
		1669	This is slightly more efficient then stopping/starting the timer each time
		1670	you want to modify its timeout value, as libev does not have to completely
		1671	remove and re-insert the timer from/into its internal data structure.
		1672
		1673	It is, however, even simpler than the "obvious" way to do it.
		1674
		1675	=item 3. Let the timer time out, but then re-arm it as required.
		1676
		1677	This method is more tricky, but usually most efficient: Most timeouts are
		1678	relatively long compared to the intervals between other activity - in
		1679	our example, within 60 seconds, there are usually many I/O events with
		1680	associated activity resets.
		1681
		1682	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1683	but remember the time of last activity, and check for a real timeout only
		1684	within the callback:
		1685
		1686	ev_tstamp last_activity; // time of last activity
		1687
		1688	static void
		1689	callback (EV_P_ ev_timer *w, int revents)
		1690	{
		1691	ev_tstamp now = ev_now (EV_A);
		1692	ev_tstamp timeout = last_activity + 60.;
		1693
		1694	// if last_activity + 60. is older than now, we did time out
		1695	if (timeout < now)
		1696	{
		1697	// timeout occured, take action
		1698	}
		1699	else
		1700	{
		1701	// callback was invoked, but there was some activity, re-arm
		1702	// the watcher to fire in last_activity + 60, which is
		1703	// guaranteed to be in the future, so "again" is positive:
		1704	w->repeat = timeout - now;
		1705	ev_timer_again (EV_A_ w);
		1706	}
		1707	}
		1708
		1709	To summarise the callback: first calculate the real timeout (defined
		1710	as "60 seconds after the last activity"), then check if that time has
		1711	been reached, which means something I<did>, in fact, time out. Otherwise
		1712	the callback was invoked too early (C<timeout> is in the future), so
		1713	re-schedule the timer to fire at that future time, to see if maybe we have
		1714	a timeout then.
		1715
		1716	Note how C<ev_timer_again> is used, taking advantage of the
		1717	C<ev_timer_again> optimisation when the timer is already running.
		1718
		1719	This scheme causes more callback invocations (about one every 60 seconds
		1720	minus half the average time between activity), but virtually no calls to
		1721	libev to change the timeout.
		1722
		1723	To start the timer, simply initialise the watcher and set C<last_activity>
		1724	to the current time (meaning we just have some activity :), then call the
		1725	callback, which will "do the right thing" and start the timer:
		1726
		1727	ev_init (timer, callback);
		1728	last_activity = ev_now (loop);
		1729	callback (loop, timer, EV_TIMEOUT);
		1730
		1731	And when there is some activity, simply store the current time in
		1732	C<last_activity>, no libev calls at all:
		1733
		1734	last_actiivty = ev_now (loop);
		1735
		1736	This technique is slightly more complex, but in most cases where the
		1737	time-out is unlikely to be triggered, much more efficient.
		1738
		1739	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1740	callback :) - just change the timeout and invoke the callback, which will
		1741	fix things for you.
		1742
		1743	=item 4. Wee, just use a double-linked list for your timeouts.
		1744
		1745	If there is not one request, but many thousands (millions...), all
		1746	employing some kind of timeout with the same timeout value, then one can
		1747	do even better:
		1748
		1749	When starting the timeout, calculate the timeout value and put the timeout
		1750	at the I<end> of the list.
		1751
		1752	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1753	the list is expected to fire (for example, using the technique #3).
		1754
		1755	When there is some activity, remove the timer from the list, recalculate
		1756	the timeout, append it to the end of the list again, and make sure to
		1757	update the C<ev_timer> if it was taken from the beginning of the list.
		1758
		1759	This way, one can manage an unlimited number of timeouts in O(1) time for
		1760	starting, stopping and updating the timers, at the expense of a major
		1761	complication, and having to use a constant timeout. The constant timeout
		1762	ensures that the list stays sorted.
		1763
		1764	=back
		1765
		1766	So which method the best?
		1767
		1768	Method #2 is a simple no-brain-required solution that is adequate in most
		1769	situations. Method #3 requires a bit more thinking, but handles many cases
		1770	better, and isn't very complicated either. In most case, choosing either
		1771	one is fine, with #3 being better in typical situations.
		1772
		1773	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1774	rather complicated, but extremely efficient, something that really pays
		1775	off after the first million or so of active timers, i.e. it's usually
		1776	overkill :)
		1777
		1778	=head3 The special problem of time updates
		1779
		1780	Establishing the current time is a costly operation (it usually takes at
		1781	least two system calls): EV therefore updates its idea of the current
		1782	time only before and after C<ev_loop> collects new events, which causes a
		1783	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1784	lots of events in one iteration.
525		1785
526	The relative timeouts are calculated relative to the C<ev_now ()>	1786	The relative timeouts are calculated relative to the C<ev_now ()>
527	time. This is usually the right thing as this timestamp refers to the time	1787	time. This is usually the right thing as this timestamp refers to the time
528	of the event triggering whatever timeout you are modifying/starting. If	1788	of the event triggering whatever timeout you are modifying/starting. If
529	you suspect event processing to be delayed and you I<need> to base the timeout	1789	you suspect event processing to be delayed and you I<need> to base the
530	on the current time, use something like this to adjust for this:	1790	timeout on the current time, use something like this to adjust for this:
531		1791
532	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1792	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
533		1793
534	The callback is guarenteed to be invoked only when its timeout has passed,	1794	If the event loop is suspended for a long time, you can also force an
535	but if multiple timers become ready during the same loop iteration then	1795	update of the time returned by C<ev_now ()> by calling C<ev_now_update
536	order of execution is undefined.	1796	()>.
		1797
		1798	=head3 The special problems of suspended animation
		1799
		1800	When you leave the server world it is quite customary to hit machines that
		1801	can suspend/hibernate - what happens to the clocks during such a suspend?
		1802
		1803	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1804	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1805	to run until the system is suspended, but they will not advance while the
		1806	system is suspended. That means, on resume, it will be as if the program
		1807	was frozen for a few seconds, but the suspend time will not be counted
		1808	towards C<ev_timer> when a monotonic clock source is used. The real time
		1809	clock advanced as expected, but if it is used as sole clocksource, then a
		1810	long suspend would be detected as a time jump by libev, and timers would
		1811	be adjusted accordingly.
		1812
		1813	I would not be surprised to see different behaviour in different between
		1814	operating systems, OS versions or even different hardware.
		1815
		1816	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1817	time jump in the monotonic clocks and the realtime clock. If the program
		1818	is suspended for a very long time, and monotonic clock sources are in use,
		1819	then you can expect C<ev_timer>s to expire as the full suspension time
		1820	will be counted towards the timers. When no monotonic clock source is in
		1821	use, then libev will again assume a timejump and adjust accordingly.
		1822
		1823	It might be beneficial for this latter case to call C<ev_suspend>
		1824	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1825	deterministic behaviour in this case (you can do nothing against
		1826	C<SIGSTOP>).
		1827
		1828	=head3 Watcher-Specific Functions and Data Members
537		1829
538	=over 4	1830	=over 4
539		1831
540	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1832	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
541		1833
542	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1834	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
543		1835
544	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1836	Configure the timer to trigger after C<after> seconds. If C<repeat>
545	C<0.>, then it will automatically be stopped. If it is positive, then the	1837	is C<0.>, then it will automatically be stopped once the timeout is
546	timer will automatically be configured to trigger again C<repeat> seconds	1838	reached. If it is positive, then the timer will automatically be
547	later, again, and again, until stopped manually.	1839	configured to trigger again C<repeat> seconds later, again, and again,
		1840	until stopped manually.
548		1841
549	The timer itself will do a best-effort at avoiding drift, that is, if you	1842	The timer itself will do a best-effort at avoiding drift, that is, if
550	configure a timer to trigger every 10 seconds, then it will trigger at	1843	you configure a timer to trigger every 10 seconds, then it will normally
551	exactly 10 second intervals. If, however, your program cannot keep up with	1844	trigger at exactly 10 second intervals. If, however, your program cannot
552	the timer (because it takes longer than those 10 seconds to do stuff) the	1845	keep up with the timer (because it takes longer than those 10 seconds to
553	timer will not fire more than once per event loop iteration.	1846	do stuff) the timer will not fire more than once per event loop iteration.
554		1847
555	=item ev_timer_again (loop)	1848	=item ev_timer_again (loop, ev_timer *)
556		1849
557	This will act as if the timer timed out and restart it again if it is	1850	This will act as if the timer timed out and restart it again if it is
558	repeating. The exact semantics are:	1851	repeating. The exact semantics are:
559		1852
		1853	If the timer is pending, its pending status is cleared.
		1854
560	If the timer is started but nonrepeating, stop it.	1855	If the timer is started but non-repeating, stop it (as if it timed out).
561		1856
562	If the timer is repeating, either start it if necessary (with the repeat	1857	If the timer is repeating, either start it if necessary (with the
563	value), or reset the running timer to the repeat value.	1858	C<repeat> value), or reset the running timer to the C<repeat> value.
564		1859
565	This sounds a bit complicated, but here is a useful and typical	1860	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
566	example: Imagine you have a tcp connection and you want a so-called idle	1861	usage example.
567	timeout, that is, you want to be called when there have been, say, 60	1862
568	seconds of inactivity on the socket. The easiest way to do this is to	1863	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
569	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1864
570	time you successfully read or write some data. If you go into an idle	1865	Returns the remaining time until a timer fires. If the timer is active,
571	state where you do not expect data to travel on the socket, you can stop	1866	then this time is relative to the current event loop time, otherwise it's
572	the timer, and again will automatically restart it if need be.	1867	the timeout value currently configured.
		1868
		1869	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1870	C<5>. When the timer is started and one second passes, C<ev_timer_remain>
		1871	will return C<4>. When the timer expires and is restarted, it will return
		1872	roughly C<7> (likely slightly less as callback invocation takes some time,
		1873	too), and so on.
		1874
		1875	=item ev_tstamp repeat [read-write]
		1876
		1877	The current C<repeat> value. Will be used each time the watcher times out
		1878	or C<ev_timer_again> is called, and determines the next timeout (if any),
		1879	which is also when any modifications are taken into account.
573		1880
574	=back	1881	=back
575		1882
		1883	=head3 Examples
		1884
		1885	Example: Create a timer that fires after 60 seconds.
		1886
		1887	static void
		1888	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1889	{
		1890	.. one minute over, w is actually stopped right here
		1891	}
		1892
		1893	ev_timer mytimer;
		1894	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1895	ev_timer_start (loop, &mytimer);
		1896
		1897	Example: Create a timeout timer that times out after 10 seconds of
		1898	inactivity.
		1899
		1900	static void
		1901	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1902	{
		1903	.. ten seconds without any activity
		1904	}
		1905
		1906	ev_timer mytimer;
		1907	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1908	ev_timer_again (&mytimer); /* start timer */
		1909	ev_loop (loop, 0);
		1910
		1911	// and in some piece of code that gets executed on any "activity":
		1912	// reset the timeout to start ticking again at 10 seconds
		1913	ev_timer_again (&mytimer);
		1914
		1915
576	=head2 C<ev_periodic> - to cron or not to cron	1916	=head2 C<ev_periodic> - to cron or not to cron?
577		1917
578	Periodic watchers are also timers of a kind, but they are very versatile	1918	Periodic watchers are also timers of a kind, but they are very versatile
579	(and unfortunately a bit complex).	1919	(and unfortunately a bit complex).
580		1920
581	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1921	Unlike C<ev_timer>, periodic watchers are not based on real time (or
582	but on wallclock time (absolute time). You can tell a periodic watcher	1922	relative time, the physical time that passes) but on wall clock time
583	to trigger "at" some specific point in time. For example, if you tell a	1923	(absolute time, the thing you can read on your calender or clock). The
584	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1924	difference is that wall clock time can run faster or slower than real
585	+ 10.>) and then reset your system clock to the last year, then it will	1925	time, and time jumps are not uncommon (e.g. when you adjust your
586	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1926	wrist-watch).
587	roughly 10 seconds later and of course not if you reset your system time
588	again).
589		1927
590	They can also be used to implement vastly more complex timers, such as	1928	You can tell a periodic watcher to trigger after some specific point
		1929	in time: for example, if you tell a periodic watcher to trigger "in 10
		1930	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1931	not a delay) and then reset your system clock to January of the previous
		1932	year, then it will take a year or more to trigger the event (unlike an
		1933	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1934	it, as it uses a relative timeout).
		1935
		1936	C<ev_periodic> watchers can also be used to implement vastly more complex
591	triggering an event on eahc midnight, local time.	1937	timers, such as triggering an event on each "midnight, local time", or
		1938	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1939	those cannot react to time jumps.
592		1940
593	As with timers, the callback is guarenteed to be invoked only when the	1941	As with timers, the callback is guaranteed to be invoked only when the
594	time (C<at>) has been passed, but if multiple periodic timers become ready	1942	point in time where it is supposed to trigger has passed. If multiple
595	during the same loop iteration then order of execution is undefined.	1943	timers become ready during the same loop iteration then the ones with
		1944	earlier time-out values are invoked before ones with later time-out values
		1945	(but this is no longer true when a callback calls C<ev_loop> recursively).
		1946
		1947	=head3 Watcher-Specific Functions and Data Members
596		1948
597	=over 4	1949	=over 4
598		1950
599	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1951	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
600		1952
601	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1953	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
602		1954
603	Lots of arguments, lets sort it out... There are basically three modes of	1955	Lots of arguments, let's sort it out... There are basically three modes of
604	operation, and we will explain them from simplest to complex:	1956	operation, and we will explain them from simplest to most complex:
605		1957
606	=over 4	1958	=over 4
607		1959
608	=item * absolute timer (interval = reschedule_cb = 0)	1960	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
609		1961
610	In this configuration the watcher triggers an event at the wallclock time	1962	In this configuration the watcher triggers an event after the wall clock
611	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1963	time C<offset> has passed. It will not repeat and will not adjust when a
612	that is, if it is to be run at January 1st 2011 then it will run when the	1964	time jump occurs, that is, if it is to be run at January 1st 2011 then it
613	system time reaches or surpasses this time.	1965	will be stopped and invoked when the system clock reaches or surpasses
		1966	this point in time.
614		1967
615	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1968	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
616		1969
617	In this mode the watcher will always be scheduled to time out at the next	1970	In this mode the watcher will always be scheduled to time out at the next
618	C<at + N * interval> time (for some integer N) and then repeat, regardless	1971	C<offset + N * interval> time (for some integer N, which can also be
619	of any time jumps.	1972	negative) and then repeat, regardless of any time jumps. The C<offset>
		1973	argument is merely an offset into the C<interval> periods.
620		1974
621	This can be used to create timers that do not drift with respect to system	1975	This can be used to create timers that do not drift with respect to the
622	time:	1976	system clock, for example, here is an C<ev_periodic> that triggers each
		1977	hour, on the hour (with respect to UTC):
623		1978
624	ev_periodic_set (&periodic, 0., 3600., 0);	1979	ev_periodic_set (&periodic, 0., 3600., 0);
625		1980
626	This doesn't mean there will always be 3600 seconds in between triggers,	1981	This doesn't mean there will always be 3600 seconds in between triggers,
627	but only that the the callback will be called when the system time shows a	1982	but only that the callback will be called when the system time shows a
628	full hour (UTC), or more correctly, when the system time is evenly divisible	1983	full hour (UTC), or more correctly, when the system time is evenly divisible
629	by 3600.	1984	by 3600.
630		1985
631	Another way to think about it (for the mathematically inclined) is that	1986	Another way to think about it (for the mathematically inclined) is that
632	C<ev_periodic> will try to run the callback in this mode at the next possible	1987	C<ev_periodic> will try to run the callback in this mode at the next possible
633	time where C<time = at (mod interval)>, regardless of any time jumps.	1988	time where C<time = offset (mod interval)>, regardless of any time jumps.
634		1989
		1990	For numerical stability it is preferable that the C<offset> value is near
		1991	C<ev_now ()> (the current time), but there is no range requirement for
		1992	this value, and in fact is often specified as zero.
		1993
		1994	Note also that there is an upper limit to how often a timer can fire (CPU
		1995	speed for example), so if C<interval> is very small then timing stability
		1996	will of course deteriorate. Libev itself tries to be exact to be about one
		1997	millisecond (if the OS supports it and the machine is fast enough).
		1998
635	=item * manual reschedule mode (reschedule_cb = callback)	1999	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
636		2000
637	In this mode the values for C<interval> and C<at> are both being	2001	In this mode the values for C<interval> and C<offset> are both being
638	ignored. Instead, each time the periodic watcher gets scheduled, the	2002	ignored. Instead, each time the periodic watcher gets scheduled, the
639	reschedule callback will be called with the watcher as first, and the	2003	reschedule callback will be called with the watcher as first, and the
640	current time as second argument.	2004	current time as second argument.
641		2005
642	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2006	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
643	ever, or make any event loop modifications>. If you need to stop it,	2007	or make ANY other event loop modifications whatsoever, unless explicitly
644	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2008	allowed by documentation here>.
645	starting a prepare watcher).
646		2009
		2010	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2011	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2012	only event loop modification you are allowed to do).
		2013
647	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2014	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
648	ev_tstamp now)>, e.g.:	2015	*w, ev_tstamp now)>, e.g.:
649		2016
		2017	static ev_tstamp
650	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2018	my_rescheduler (ev_periodic *w, ev_tstamp now)
651	{	2019	{
652	return now + 60.;	2020	return now + 60.;
653	}	2021	}
654		2022
655	It must return the next time to trigger, based on the passed time value	2023	It must return the next time to trigger, based on the passed time value
656	(that is, the lowest time value larger than to the second argument). It	2024	(that is, the lowest time value larger than to the second argument). It
657	will usually be called just before the callback will be triggered, but	2025	will usually be called just before the callback will be triggered, but
658	might be called at other times, too.	2026	might be called at other times, too.
659		2027
660	NOTE: I<< This callback must always return a time that is later than the	2028	NOTE: I<< This callback must always return a time that is higher than or
661	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2029	equal to the passed C<now> value >>.
662		2030
663	This can be used to create very complex timers, such as a timer that	2031	This can be used to create very complex timers, such as a timer that
664	triggers on each midnight, local time. To do this, you would calculate the	2032	triggers on "next midnight, local time". To do this, you would calculate the
665	next midnight after C<now> and return the timestamp value for this. How	2033	next midnight after C<now> and return the timestamp value for this. How
666	you do this is, again, up to you (but it is not trivial, which is the main	2034	you do this is, again, up to you (but it is not trivial, which is the main
667	reason I omitted it as an example).	2035	reason I omitted it as an example).
668		2036
669	=back	2037	=back
…		…
673	Simply stops and restarts the periodic watcher again. This is only useful	2041	Simply stops and restarts the periodic watcher again. This is only useful
674	when you changed some parameters or the reschedule callback would return	2042	when you changed some parameters or the reschedule callback would return
675	a different time than the last time it was called (e.g. in a crond like	2043	a different time than the last time it was called (e.g. in a crond like
676	program when the crontabs have changed).	2044	program when the crontabs have changed).
677		2045
		2046	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2047
		2048	When active, returns the absolute time that the watcher is supposed
		2049	to trigger next. This is not the same as the C<offset> argument to
		2050	C<ev_periodic_set>, but indeed works even in interval and manual
		2051	rescheduling modes.
		2052
		2053	=item ev_tstamp offset [read-write]
		2054
		2055	When repeating, this contains the offset value, otherwise this is the
		2056	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2057	although libev might modify this value for better numerical stability).
		2058
		2059	Can be modified any time, but changes only take effect when the periodic
		2060	timer fires or C<ev_periodic_again> is being called.
		2061
		2062	=item ev_tstamp interval [read-write]
		2063
		2064	The current interval value. Can be modified any time, but changes only
		2065	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2066	called.
		2067
		2068	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2069
		2070	The current reschedule callback, or C<0>, if this functionality is
		2071	switched off. Can be changed any time, but changes only take effect when
		2072	the periodic timer fires or C<ev_periodic_again> is being called.
		2073
678	=back	2074	=back
679		2075
		2076	=head3 Examples
		2077
		2078	Example: Call a callback every hour, or, more precisely, whenever the
		2079	system time is divisible by 3600. The callback invocation times have
		2080	potentially a lot of jitter, but good long-term stability.
		2081
		2082	static void
		2083	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		2084	{
		2085	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2086	}
		2087
		2088	ev_periodic hourly_tick;
		2089	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2090	ev_periodic_start (loop, &hourly_tick);
		2091
		2092	Example: The same as above, but use a reschedule callback to do it:
		2093
		2094	#include <math.h>
		2095
		2096	static ev_tstamp
		2097	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2098	{
		2099	return now + (3600. - fmod (now, 3600.));
		2100	}
		2101
		2102	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2103
		2104	Example: Call a callback every hour, starting now:
		2105
		2106	ev_periodic hourly_tick;
		2107	ev_periodic_init (&hourly_tick, clock_cb,
		2108	fmod (ev_now (loop), 3600.), 3600., 0);
		2109	ev_periodic_start (loop, &hourly_tick);
		2110
		2111
680	=head2 C<ev_signal> - signal me when a signal gets signalled	2112	=head2 C<ev_signal> - signal me when a signal gets signalled!
681		2113
682	Signal watchers will trigger an event when the process receives a specific	2114	Signal watchers will trigger an event when the process receives a specific
683	signal one or more times. Even though signals are very asynchronous, libev	2115	signal one or more times. Even though signals are very asynchronous, libev
684	will try it's best to deliver signals synchronously, i.e. as part of the	2116	will try it's best to deliver signals synchronously, i.e. as part of the
685	normal event processing, like any other event.	2117	normal event processing, like any other event.
686		2118
		2119	If you want signals to be delivered truly asynchronously, just use
		2120	C<sigaction> as you would do without libev and forget about sharing
		2121	the signal. You can even use C<ev_async> from a signal handler to
		2122	synchronously wake up an event loop.
		2123
687	You can configure as many watchers as you like per signal. Only when the	2124	You can configure as many watchers as you like for the same signal, but
		2125	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2126	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2127	C<SIGINT> in both the default loop and another loop at the same time. At
		2128	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2129
688	first watcher gets started will libev actually register a signal watcher	2130	When the first watcher gets started will libev actually register something
689	with the kernel (thus it coexists with your own signal handlers as long	2131	with the kernel (thus it coexists with your own signal handlers as long as
690	as you don't register any with libev). Similarly, when the last signal	2132	you don't register any with libev for the same signal).
691	watcher for a signal is stopped libev will reset the signal handler to	2133
692	SIG_DFL (regardless of what it was set to before).	2134	If possible and supported, libev will install its handlers with
		2135	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2136	not be unduly interrupted. If you have a problem with system calls getting
		2137	interrupted by signals you can block all signals in an C<ev_check> watcher
		2138	and unblock them in an C<ev_prepare> watcher.
		2139
		2140	=head3 The special problem of inheritance over fork/execve/pthread_create
		2141
		2142	Both the signal mask (C<sigprocmask>) and the signal disposition
		2143	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2144	stopping it again), that is, libev might or might not block the signal,
		2145	and might or might not set or restore the installed signal handler.
		2146
		2147	While this does not matter for the signal disposition (libev never
		2148	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2149	C<execve>), this matters for the signal mask: many programs do not expect
		2150	certain signals to be blocked.
		2151
		2152	This means that before calling C<exec> (from the child) you should reset
		2153	the signal mask to whatever "default" you expect (all clear is a good
		2154	choice usually).
		2155
		2156	The simplest way to ensure that the signal mask is reset in the child is
		2157	to install a fork handler with C<pthread_atfork> that resets it. That will
		2158	catch fork calls done by libraries (such as the libc) as well.
		2159
		2160	In current versions of libev, the signal will not be blocked indefinitely
		2161	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2162	the window of opportunity for problems, it will not go away, as libev
		2163	I<has> to modify the signal mask, at least temporarily.
		2164
		2165	So I can't stress this enough I<if you do not reset your signal mask
		2166	when you expect it to be empty, you have a race condition in your
		2167	program>. This is not a libev-specific thing, this is true for most event
		2168	libraries.
		2169
		2170	=head3 Watcher-Specific Functions and Data Members
693		2171
694	=over 4	2172	=over 4
695		2173
696	=item ev_signal_init (ev_signal *, callback, int signum)	2174	=item ev_signal_init (ev_signal *, callback, int signum)
697		2175
698	=item ev_signal_set (ev_signal *, int signum)	2176	=item ev_signal_set (ev_signal *, int signum)
699		2177
700	Configures the watcher to trigger on the given signal number (usually one	2178	Configures the watcher to trigger on the given signal number (usually one
701	of the C<SIGxxx> constants).	2179	of the C<SIGxxx> constants).
702		2180
		2181	=item int signum [read-only]
		2182
		2183	The signal the watcher watches out for.
		2184
703	=back	2185	=back
704		2186
		2187	=head3 Examples
		2188
		2189	Example: Try to exit cleanly on SIGINT.
		2190
		2191	static void
		2192	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2193	{
		2194	ev_unloop (loop, EVUNLOOP_ALL);
		2195	}
		2196
		2197	ev_signal signal_watcher;
		2198	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2199	ev_signal_start (loop, &signal_watcher);
		2200
		2201
705	=head2 C<ev_child> - wait for pid status changes	2202	=head2 C<ev_child> - watch out for process status changes
706		2203
707	Child watchers trigger when your process receives a SIGCHLD in response to	2204	Child watchers trigger when your process receives a SIGCHLD in response to
708	some child status changes (most typically when a child of yours dies).	2205	some child status changes (most typically when a child of yours dies or
		2206	exits). It is permissible to install a child watcher I<after> the child
		2207	has been forked (which implies it might have already exited), as long
		2208	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2209	forking and then immediately registering a watcher for the child is fine,
		2210	but forking and registering a watcher a few event loop iterations later or
		2211	in the next callback invocation is not.
		2212
		2213	Only the default event loop is capable of handling signals, and therefore
		2214	you can only register child watchers in the default event loop.
		2215
		2216	Due to some design glitches inside libev, child watchers will always be
		2217	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2218	libev)
		2219
		2220	=head3 Process Interaction
		2221
		2222	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2223	initialised. This is necessary to guarantee proper behaviour even if the
		2224	first child watcher is started after the child exits. The occurrence
		2225	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2226	synchronously as part of the event loop processing. Libev always reaps all
		2227	children, even ones not watched.
		2228
		2229	=head3 Overriding the Built-In Processing
		2230
		2231	Libev offers no special support for overriding the built-in child
		2232	processing, but if your application collides with libev's default child
		2233	handler, you can override it easily by installing your own handler for
		2234	C<SIGCHLD> after initialising the default loop, and making sure the
		2235	default loop never gets destroyed. You are encouraged, however, to use an
		2236	event-based approach to child reaping and thus use libev's support for
		2237	that, so other libev users can use C<ev_child> watchers freely.
		2238
		2239	=head3 Stopping the Child Watcher
		2240
		2241	Currently, the child watcher never gets stopped, even when the
		2242	child terminates, so normally one needs to stop the watcher in the
		2243	callback. Future versions of libev might stop the watcher automatically
		2244	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2245	problem).
		2246
		2247	=head3 Watcher-Specific Functions and Data Members
709		2248
710	=over 4	2249	=over 4
711		2250
712	=item ev_child_init (ev_child *, callback, int pid)	2251	=item ev_child_init (ev_child *, callback, int pid, int trace)
713		2252
714	=item ev_child_set (ev_child *, int pid)	2253	=item ev_child_set (ev_child *, int pid, int trace)
715		2254
716	Configures the watcher to wait for status changes of process C<pid> (or	2255	Configures the watcher to wait for status changes of process C<pid> (or
717	I<any> process if C<pid> is specified as C<0>). The callback can look	2256	I<any> process if C<pid> is specified as C<0>). The callback can look
718	at the C<rstatus> member of the C<ev_child> watcher structure to see	2257	at the C<rstatus> member of the C<ev_child> watcher structure to see
719	the status word (use the macros from C<sys/wait.h> and see your systems	2258	the status word (use the macros from C<sys/wait.h> and see your systems
720	C<waitpid> documentation). The C<rpid> member contains the pid of the	2259	C<waitpid> documentation). The C<rpid> member contains the pid of the
721	process causing the status change.	2260	process causing the status change. C<trace> must be either C<0> (only
		2261	activate the watcher when the process terminates) or C<1> (additionally
		2262	activate the watcher when the process is stopped or continued).
		2263
		2264	=item int pid [read-only]
		2265
		2266	The process id this watcher watches out for, or C<0>, meaning any process id.
		2267
		2268	=item int rpid [read-write]
		2269
		2270	The process id that detected a status change.
		2271
		2272	=item int rstatus [read-write]
		2273
		2274	The process exit/trace status caused by C<rpid> (see your systems
		2275	C<waitpid> and C<sys/wait.h> documentation for details).
722		2276
723	=back	2277	=back
724		2278
		2279	=head3 Examples
		2280
		2281	Example: C<fork()> a new process and install a child handler to wait for
		2282	its completion.
		2283
		2284	ev_child cw;
		2285
		2286	static void
		2287	child_cb (EV_P_ ev_child *w, int revents)
		2288	{
		2289	ev_child_stop (EV_A_ w);
		2290	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2291	}
		2292
		2293	pid_t pid = fork ();
		2294
		2295	if (pid < 0)
		2296	// error
		2297	else if (pid == 0)
		2298	{
		2299	// the forked child executes here
		2300	exit (1);
		2301	}
		2302	else
		2303	{
		2304	ev_child_init (&cw, child_cb, pid, 0);
		2305	ev_child_start (EV_DEFAULT_ &cw);
		2306	}
		2307
		2308
		2309	=head2 C<ev_stat> - did the file attributes just change?
		2310
		2311	This watches a file system path for attribute changes. That is, it calls
		2312	C<stat> on that path in regular intervals (or when the OS says it changed)
		2313	and sees if it changed compared to the last time, invoking the callback if
		2314	it did.
		2315
		2316	The path does not need to exist: changing from "path exists" to "path does
		2317	not exist" is a status change like any other. The condition "path does not
		2318	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2319	C<st_nlink> field being zero (which is otherwise always forced to be at
		2320	least one) and all the other fields of the stat buffer having unspecified
		2321	contents.
		2322
		2323	The path I<must not> end in a slash or contain special components such as
		2324	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2325	your working directory changes, then the behaviour is undefined.
		2326
		2327	Since there is no portable change notification interface available, the
		2328	portable implementation simply calls C<stat(2)> regularly on the path
		2329	to see if it changed somehow. You can specify a recommended polling
		2330	interval for this case. If you specify a polling interval of C<0> (highly
		2331	recommended!) then a I<suitable, unspecified default> value will be used
		2332	(which you can expect to be around five seconds, although this might
		2333	change dynamically). Libev will also impose a minimum interval which is
		2334	currently around C<0.1>, but that's usually overkill.
		2335
		2336	This watcher type is not meant for massive numbers of stat watchers,
		2337	as even with OS-supported change notifications, this can be
		2338	resource-intensive.
		2339
		2340	At the time of this writing, the only OS-specific interface implemented
		2341	is the Linux inotify interface (implementing kqueue support is left as an
		2342	exercise for the reader. Note, however, that the author sees no way of
		2343	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2344
		2345	=head3 ABI Issues (Largefile Support)
		2346
		2347	Libev by default (unless the user overrides this) uses the default
		2348	compilation environment, which means that on systems with large file
		2349	support disabled by default, you get the 32 bit version of the stat
		2350	structure. When using the library from programs that change the ABI to
		2351	use 64 bit file offsets the programs will fail. In that case you have to
		2352	compile libev with the same flags to get binary compatibility. This is
		2353	obviously the case with any flags that change the ABI, but the problem is
		2354	most noticeably displayed with ev_stat and large file support.
		2355
		2356	The solution for this is to lobby your distribution maker to make large
		2357	file interfaces available by default (as e.g. FreeBSD does) and not
		2358	optional. Libev cannot simply switch on large file support because it has
		2359	to exchange stat structures with application programs compiled using the
		2360	default compilation environment.
		2361
		2362	=head3 Inotify and Kqueue
		2363
		2364	When C<inotify (7)> support has been compiled into libev and present at
		2365	runtime, it will be used to speed up change detection where possible. The
		2366	inotify descriptor will be created lazily when the first C<ev_stat>
		2367	watcher is being started.
		2368
		2369	Inotify presence does not change the semantics of C<ev_stat> watchers
		2370	except that changes might be detected earlier, and in some cases, to avoid
		2371	making regular C<stat> calls. Even in the presence of inotify support
		2372	there are many cases where libev has to resort to regular C<stat> polling,
		2373	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2374	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2375	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2376	xfs are fully working) libev usually gets away without polling.
		2377
		2378	There is no support for kqueue, as apparently it cannot be used to
		2379	implement this functionality, due to the requirement of having a file
		2380	descriptor open on the object at all times, and detecting renames, unlinks
		2381	etc. is difficult.
		2382
		2383	=head3 C<stat ()> is a synchronous operation
		2384
		2385	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2386	the process. The exception are C<ev_stat> watchers - those call C<stat
		2387	()>, which is a synchronous operation.
		2388
		2389	For local paths, this usually doesn't matter: unless the system is very
		2390	busy or the intervals between stat's are large, a stat call will be fast,
		2391	as the path data is usually in memory already (except when starting the
		2392	watcher).
		2393
		2394	For networked file systems, calling C<stat ()> can block an indefinite
		2395	time due to network issues, and even under good conditions, a stat call
		2396	often takes multiple milliseconds.
		2397
		2398	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2399	paths, although this is fully supported by libev.
		2400
		2401	=head3 The special problem of stat time resolution
		2402
		2403	The C<stat ()> system call only supports full-second resolution portably,
		2404	and even on systems where the resolution is higher, most file systems
		2405	still only support whole seconds.
		2406
		2407	That means that, if the time is the only thing that changes, you can
		2408	easily miss updates: on the first update, C<ev_stat> detects a change and
		2409	calls your callback, which does something. When there is another update
		2410	within the same second, C<ev_stat> will be unable to detect unless the
		2411	stat data does change in other ways (e.g. file size).
		2412
		2413	The solution to this is to delay acting on a change for slightly more
		2414	than a second (or till slightly after the next full second boundary), using
		2415	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2416	ev_timer_again (loop, w)>).
		2417
		2418	The C<.02> offset is added to work around small timing inconsistencies
		2419	of some operating systems (where the second counter of the current time
		2420	might be be delayed. One such system is the Linux kernel, where a call to
		2421	C<gettimeofday> might return a timestamp with a full second later than
		2422	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2423	update file times then there will be a small window where the kernel uses
		2424	the previous second to update file times but libev might already execute
		2425	the timer callback).
		2426
		2427	=head3 Watcher-Specific Functions and Data Members
		2428
		2429	=over 4
		2430
		2431	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2432
		2433	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2434
		2435	Configures the watcher to wait for status changes of the given
		2436	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2437	be detected and should normally be specified as C<0> to let libev choose
		2438	a suitable value. The memory pointed to by C<path> must point to the same
		2439	path for as long as the watcher is active.
		2440
		2441	The callback will receive an C<EV_STAT> event when a change was detected,
		2442	relative to the attributes at the time the watcher was started (or the
		2443	last change was detected).
		2444
		2445	=item ev_stat_stat (loop, ev_stat *)
		2446
		2447	Updates the stat buffer immediately with new values. If you change the
		2448	watched path in your callback, you could call this function to avoid
		2449	detecting this change (while introducing a race condition if you are not
		2450	the only one changing the path). Can also be useful simply to find out the
		2451	new values.
		2452
		2453	=item ev_statdata attr [read-only]
		2454
		2455	The most-recently detected attributes of the file. Although the type is
		2456	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2457	suitable for your system, but you can only rely on the POSIX-standardised
		2458	members to be present. If the C<st_nlink> member is C<0>, then there was
		2459	some error while C<stat>ing the file.
		2460
		2461	=item ev_statdata prev [read-only]
		2462
		2463	The previous attributes of the file. The callback gets invoked whenever
		2464	C<prev> != C<attr>, or, more precisely, one or more of these members
		2465	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2466	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2467
		2468	=item ev_tstamp interval [read-only]
		2469
		2470	The specified interval.
		2471
		2472	=item const char *path [read-only]
		2473
		2474	The file system path that is being watched.
		2475
		2476	=back
		2477
		2478	=head3 Examples
		2479
		2480	Example: Watch C</etc/passwd> for attribute changes.
		2481
		2482	static void
		2483	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2484	{
		2485	/* /etc/passwd changed in some way */
		2486	if (w->attr.st_nlink)
		2487	{
		2488	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2489	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2490	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2491	}
		2492	else
		2493	/* you shalt not abuse printf for puts */
		2494	puts ("wow, /etc/passwd is not there, expect problems. "
		2495	"if this is windows, they already arrived\n");
		2496	}
		2497
		2498	...
		2499	ev_stat passwd;
		2500
		2501	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2502	ev_stat_start (loop, &passwd);
		2503
		2504	Example: Like above, but additionally use a one-second delay so we do not
		2505	miss updates (however, frequent updates will delay processing, too, so
		2506	one might do the work both on C<ev_stat> callback invocation I<and> on
		2507	C<ev_timer> callback invocation).
		2508
		2509	static ev_stat passwd;
		2510	static ev_timer timer;
		2511
		2512	static void
		2513	timer_cb (EV_P_ ev_timer *w, int revents)
		2514	{
		2515	ev_timer_stop (EV_A_ w);
		2516
		2517	/* now it's one second after the most recent passwd change */
		2518	}
		2519
		2520	static void
		2521	stat_cb (EV_P_ ev_stat *w, int revents)
		2522	{
		2523	/* reset the one-second timer */
		2524	ev_timer_again (EV_A_ &timer);
		2525	}
		2526
		2527	...
		2528	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2529	ev_stat_start (loop, &passwd);
		2530	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2531
		2532
725	=head2 C<ev_idle> - when you've got nothing better to do	2533	=head2 C<ev_idle> - when you've got nothing better to do...
726		2534
727	Idle watchers trigger events when there are no other events are pending	2535	Idle watchers trigger events when no other events of the same or higher
728	(prepare, check and other idle watchers do not count). That is, as long	2536	priority are pending (prepare, check and other idle watchers do not count
729	as your process is busy handling sockets or timeouts (or even signals,	2537	as receiving "events").
730	imagine) it will not be triggered. But when your process is idle all idle	2538
731	watchers are being called again and again, once per event loop iteration -	2539	That is, as long as your process is busy handling sockets or timeouts
		2540	(or even signals, imagine) of the same or higher priority it will not be
		2541	triggered. But when your process is idle (or only lower-priority watchers
		2542	are pending), the idle watchers are being called once per event loop
732	until stopped, that is, or your process receives more events and becomes	2543	iteration - until stopped, that is, or your process receives more events
733	busy.	2544	and becomes busy again with higher priority stuff.
734		2545
735	The most noteworthy effect is that as long as any idle watchers are	2546	The most noteworthy effect is that as long as any idle watchers are
736	active, the process will not block when waiting for new events.	2547	active, the process will not block when waiting for new events.
737		2548
738	Apart from keeping your process non-blocking (which is a useful	2549	Apart from keeping your process non-blocking (which is a useful
739	effect on its own sometimes), idle watchers are a good place to do	2550	effect on its own sometimes), idle watchers are a good place to do
740	"pseudo-background processing", or delay processing stuff to after the	2551	"pseudo-background processing", or delay processing stuff to after the
741	event loop has handled all outstanding events.	2552	event loop has handled all outstanding events.
742		2553
		2554	=head3 Watcher-Specific Functions and Data Members
		2555
743	=over 4	2556	=over 4
744		2557
745	=item ev_idle_init (ev_signal *, callback)	2558	=item ev_idle_init (ev_idle *, callback)
746		2559
747	Initialises and configures the idle watcher - it has no parameters of any	2560	Initialises and configures the idle watcher - it has no parameters of any
748	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2561	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
749	believe me.	2562	believe me.
750		2563
751	=back	2564	=back
752		2565
		2566	=head3 Examples
		2567
		2568	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2569	callback, free it. Also, use no error checking, as usual.
		2570
		2571	static void
		2572	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2573	{
		2574	free (w);
		2575	// now do something you wanted to do when the program has
		2576	// no longer anything immediate to do.
		2577	}
		2578
		2579	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2580	ev_idle_init (idle_watcher, idle_cb);
		2581	ev_idle_start (loop, idle_watcher);
		2582
		2583
753	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2584	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
754		2585
755	Prepare and check watchers are usually (but not always) used in tandem:	2586	Prepare and check watchers are usually (but not always) used in pairs:
756	prepare watchers get invoked before the process blocks and check watchers	2587	prepare watchers get invoked before the process blocks and check watchers
757	afterwards.	2588	afterwards.
758		2589
		2590	You I<must not> call C<ev_loop> or similar functions that enter
		2591	the current event loop from either C<ev_prepare> or C<ev_check>
		2592	watchers. Other loops than the current one are fine, however. The
		2593	rationale behind this is that you do not need to check for recursion in
		2594	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2595	C<ev_check> so if you have one watcher of each kind they will always be
		2596	called in pairs bracketing the blocking call.
		2597
759	Their main purpose is to integrate other event mechanisms into libev. This	2598	Their main purpose is to integrate other event mechanisms into libev and
760	could be used, for example, to track variable changes, implement your own	2599	their use is somewhat advanced. They could be used, for example, to track
761	watchers, integrate net-snmp or a coroutine library and lots more.	2600	variable changes, implement your own watchers, integrate net-snmp or a
		2601	coroutine library and lots more. They are also occasionally useful if
		2602	you cache some data and want to flush it before blocking (for example,
		2603	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2604	watcher).
762		2605
763	This is done by examining in each prepare call which file descriptors need	2606	This is done by examining in each prepare call which file descriptors
764	to be watched by the other library, registering C<ev_io> watchers for	2607	need to be watched by the other library, registering C<ev_io> watchers
765	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2608	for them and starting an C<ev_timer> watcher for any timeouts (many
766	provide just this functionality). Then, in the check watcher you check for	2609	libraries provide exactly this functionality). Then, in the check watcher,
767	any events that occured (by checking the pending status of all watchers	2610	you check for any events that occurred (by checking the pending status
768	and stopping them) and call back into the library. The I/O and timer	2611	of all watchers and stopping them) and call back into the library. The
769	callbacks will never actually be called (but must be valid nevertheless,	2612	I/O and timer callbacks will never actually be called (but must be valid
770	because you never know, you know?).	2613	nevertheless, because you never know, you know?).
771		2614
772	As another example, the Perl Coro module uses these hooks to integrate	2615	As another example, the Perl Coro module uses these hooks to integrate
773	coroutines into libev programs, by yielding to other active coroutines	2616	coroutines into libev programs, by yielding to other active coroutines
774	during each prepare and only letting the process block if no coroutines	2617	during each prepare and only letting the process block if no coroutines
775	are ready to run (it's actually more complicated: it only runs coroutines	2618	are ready to run (it's actually more complicated: it only runs coroutines
776	with priority higher than or equal to the event loop and one coroutine	2619	with priority higher than or equal to the event loop and one coroutine
777	of lower priority, but only once, using idle watchers to keep the event	2620	of lower priority, but only once, using idle watchers to keep the event
778	loop from blocking if lower-priority coroutines are active, thus mapping	2621	loop from blocking if lower-priority coroutines are active, thus mapping
779	low-priority coroutines to idle/background tasks).	2622	low-priority coroutines to idle/background tasks).
780		2623
		2624	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2625	priority, to ensure that they are being run before any other watchers
		2626	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2627
		2628	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2629	activate ("feed") events into libev. While libev fully supports this, they
		2630	might get executed before other C<ev_check> watchers did their job. As
		2631	C<ev_check> watchers are often used to embed other (non-libev) event
		2632	loops those other event loops might be in an unusable state until their
		2633	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2634	others).
		2635
		2636	=head3 Watcher-Specific Functions and Data Members
		2637
781	=over 4	2638	=over 4
782		2639
783	=item ev_prepare_init (ev_prepare *, callback)	2640	=item ev_prepare_init (ev_prepare *, callback)
784		2641
785	=item ev_check_init (ev_check *, callback)	2642	=item ev_check_init (ev_check *, callback)
786		2643
787	Initialises and configures the prepare or check watcher - they have no	2644	Initialises and configures the prepare or check watcher - they have no
788	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2645	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
789	macros, but using them is utterly, utterly and completely pointless.	2646	macros, but using them is utterly, utterly, utterly and completely
		2647	pointless.
790		2648
791	=back	2649	=back
792		2650
		2651	=head3 Examples
		2652
		2653	There are a number of principal ways to embed other event loops or modules
		2654	into libev. Here are some ideas on how to include libadns into libev
		2655	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2656	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2657	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2658	Glib event loop).
		2659
		2660	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2661	and in a check watcher, destroy them and call into libadns. What follows
		2662	is pseudo-code only of course. This requires you to either use a low
		2663	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2664	the callbacks for the IO/timeout watchers might not have been called yet.
		2665
		2666	static ev_io iow [nfd];
		2667	static ev_timer tw;
		2668
		2669	static void
		2670	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2671	{
		2672	}
		2673
		2674	// create io watchers for each fd and a timer before blocking
		2675	static void
		2676	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2677	{
		2678	int timeout = 3600000;
		2679	struct pollfd fds [nfd];
		2680	// actual code will need to loop here and realloc etc.
		2681	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2682
		2683	/* the callback is illegal, but won't be called as we stop during check */
		2684	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2685	ev_timer_start (loop, &tw);
		2686
		2687	// create one ev_io per pollfd
		2688	for (int i = 0; i < nfd; ++i)
		2689	{
		2690	ev_io_init (iow + i, io_cb, fds [i].fd,
		2691	((fds [i].events & POLLIN ? EV_READ : 0)
		2692	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2693
		2694	fds [i].revents = 0;
		2695	ev_io_start (loop, iow + i);
		2696	}
		2697	}
		2698
		2699	// stop all watchers after blocking
		2700	static void
		2701	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2702	{
		2703	ev_timer_stop (loop, &tw);
		2704
		2705	for (int i = 0; i < nfd; ++i)
		2706	{
		2707	// set the relevant poll flags
		2708	// could also call adns_processreadable etc. here
		2709	struct pollfd *fd = fds + i;
		2710	int revents = ev_clear_pending (iow + i);
		2711	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2712	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2713
		2714	// now stop the watcher
		2715	ev_io_stop (loop, iow + i);
		2716	}
		2717
		2718	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2719	}
		2720
		2721	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2722	in the prepare watcher and would dispose of the check watcher.
		2723
		2724	Method 3: If the module to be embedded supports explicit event
		2725	notification (libadns does), you can also make use of the actual watcher
		2726	callbacks, and only destroy/create the watchers in the prepare watcher.
		2727
		2728	static void
		2729	timer_cb (EV_P_ ev_timer *w, int revents)
		2730	{
		2731	adns_state ads = (adns_state)w->data;
		2732	update_now (EV_A);
		2733
		2734	adns_processtimeouts (ads, &tv_now);
		2735	}
		2736
		2737	static void
		2738	io_cb (EV_P_ ev_io *w, int revents)
		2739	{
		2740	adns_state ads = (adns_state)w->data;
		2741	update_now (EV_A);
		2742
		2743	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2744	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2745	}
		2746
		2747	// do not ever call adns_afterpoll
		2748
		2749	Method 4: Do not use a prepare or check watcher because the module you
		2750	want to embed is not flexible enough to support it. Instead, you can
		2751	override their poll function. The drawback with this solution is that the
		2752	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2753	this approach, effectively embedding EV as a client into the horrible
		2754	libglib event loop.
		2755
		2756	static gint
		2757	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2758	{
		2759	int got_events = 0;
		2760
		2761	for (n = 0; n < nfds; ++n)
		2762	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2763
		2764	if (timeout >= 0)
		2765	// create/start timer
		2766
		2767	// poll
		2768	ev_loop (EV_A_ 0);
		2769
		2770	// stop timer again
		2771	if (timeout >= 0)
		2772	ev_timer_stop (EV_A_ &to);
		2773
		2774	// stop io watchers again - their callbacks should have set
		2775	for (n = 0; n < nfds; ++n)
		2776	ev_io_stop (EV_A_ iow [n]);
		2777
		2778	return got_events;
		2779	}
		2780
		2781
		2782	=head2 C<ev_embed> - when one backend isn't enough...
		2783
		2784	This is a rather advanced watcher type that lets you embed one event loop
		2785	into another (currently only C<ev_io> events are supported in the embedded
		2786	loop, other types of watchers might be handled in a delayed or incorrect
		2787	fashion and must not be used).
		2788
		2789	There are primarily two reasons you would want that: work around bugs and
		2790	prioritise I/O.
		2791
		2792	As an example for a bug workaround, the kqueue backend might only support
		2793	sockets on some platform, so it is unusable as generic backend, but you
		2794	still want to make use of it because you have many sockets and it scales
		2795	so nicely. In this case, you would create a kqueue-based loop and embed
		2796	it into your default loop (which might use e.g. poll). Overall operation
		2797	will be a bit slower because first libev has to call C<poll> and then
		2798	C<kevent>, but at least you can use both mechanisms for what they are
		2799	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2800
		2801	As for prioritising I/O: under rare circumstances you have the case where
		2802	some fds have to be watched and handled very quickly (with low latency),
		2803	and even priorities and idle watchers might have too much overhead. In
		2804	this case you would put all the high priority stuff in one loop and all
		2805	the rest in a second one, and embed the second one in the first.
		2806
		2807	As long as the watcher is active, the callback will be invoked every
		2808	time there might be events pending in the embedded loop. The callback
		2809	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2810	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2811	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2812	to give the embedded loop strictly lower priority for example).
		2813
		2814	You can also set the callback to C<0>, in which case the embed watcher
		2815	will automatically execute the embedded loop sweep whenever necessary.
		2816
		2817	Fork detection will be handled transparently while the C<ev_embed> watcher
		2818	is active, i.e., the embedded loop will automatically be forked when the
		2819	embedding loop forks. In other cases, the user is responsible for calling
		2820	C<ev_loop_fork> on the embedded loop.
		2821
		2822	Unfortunately, not all backends are embeddable: only the ones returned by
		2823	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2824	portable one.
		2825
		2826	So when you want to use this feature you will always have to be prepared
		2827	that you cannot get an embeddable loop. The recommended way to get around
		2828	this is to have a separate variables for your embeddable loop, try to
		2829	create it, and if that fails, use the normal loop for everything.
		2830
		2831	=head3 C<ev_embed> and fork
		2832
		2833	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2834	automatically be applied to the embedded loop as well, so no special
		2835	fork handling is required in that case. When the watcher is not running,
		2836	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2837	as applicable.
		2838
		2839	=head3 Watcher-Specific Functions and Data Members
		2840
		2841	=over 4
		2842
		2843	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2844
		2845	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2846
		2847	Configures the watcher to embed the given loop, which must be
		2848	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2849	invoked automatically, otherwise it is the responsibility of the callback
		2850	to invoke it (it will continue to be called until the sweep has been done,
		2851	if you do not want that, you need to temporarily stop the embed watcher).
		2852
		2853	=item ev_embed_sweep (loop, ev_embed *)
		2854
		2855	Make a single, non-blocking sweep over the embedded loop. This works
		2856	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2857	appropriate way for embedded loops.
		2858
		2859	=item struct ev_loop *other [read-only]
		2860
		2861	The embedded event loop.
		2862
		2863	=back
		2864
		2865	=head3 Examples
		2866
		2867	Example: Try to get an embeddable event loop and embed it into the default
		2868	event loop. If that is not possible, use the default loop. The default
		2869	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2870	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2871	used).
		2872
		2873	struct ev_loop *loop_hi = ev_default_init (0);
		2874	struct ev_loop *loop_lo = 0;
		2875	ev_embed embed;
		2876
		2877	// see if there is a chance of getting one that works
		2878	// (remember that a flags value of 0 means autodetection)
		2879	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2880	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2881	: 0;
		2882
		2883	// if we got one, then embed it, otherwise default to loop_hi
		2884	if (loop_lo)
		2885	{
		2886	ev_embed_init (&embed, 0, loop_lo);
		2887	ev_embed_start (loop_hi, &embed);
		2888	}
		2889	else
		2890	loop_lo = loop_hi;
		2891
		2892	Example: Check if kqueue is available but not recommended and create
		2893	a kqueue backend for use with sockets (which usually work with any
		2894	kqueue implementation). Store the kqueue/socket-only event loop in
		2895	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2896
		2897	struct ev_loop *loop = ev_default_init (0);
		2898	struct ev_loop *loop_socket = 0;
		2899	ev_embed embed;
		2900
		2901	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2902	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2903	{
		2904	ev_embed_init (&embed, 0, loop_socket);
		2905	ev_embed_start (loop, &embed);
		2906	}
		2907
		2908	if (!loop_socket)
		2909	loop_socket = loop;
		2910
		2911	// now use loop_socket for all sockets, and loop for everything else
		2912
		2913
		2914	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2915
		2916	Fork watchers are called when a C<fork ()> was detected (usually because
		2917	whoever is a good citizen cared to tell libev about it by calling
		2918	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2919	event loop blocks next and before C<ev_check> watchers are being called,
		2920	and only in the child after the fork. If whoever good citizen calling
		2921	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2922	handlers will be invoked, too, of course.
		2923
		2924	=head3 The special problem of life after fork - how is it possible?
		2925
		2926	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2927	up/change the process environment, followed by a call to C<exec()>. This
		2928	sequence should be handled by libev without any problems.
		2929
		2930	This changes when the application actually wants to do event handling
		2931	in the child, or both parent in child, in effect "continuing" after the
		2932	fork.
		2933
		2934	The default mode of operation (for libev, with application help to detect
		2935	forks) is to duplicate all the state in the child, as would be expected
		2936	when I<either> the parent I<or> the child process continues.
		2937
		2938	When both processes want to continue using libev, then this is usually the
		2939	wrong result. In that case, usually one process (typically the parent) is
		2940	supposed to continue with all watchers in place as before, while the other
		2941	process typically wants to start fresh, i.e. without any active watchers.
		2942
		2943	The cleanest and most efficient way to achieve that with libev is to
		2944	simply create a new event loop, which of course will be "empty", and
		2945	use that for new watchers. This has the advantage of not touching more
		2946	memory than necessary, and thus avoiding the copy-on-write, and the
		2947	disadvantage of having to use multiple event loops (which do not support
		2948	signal watchers).
		2949
		2950	When this is not possible, or you want to use the default loop for
		2951	other reasons, then in the process that wants to start "fresh", call
		2952	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2953	the default loop will "orphan" (not stop) all registered watchers, so you
		2954	have to be careful not to execute code that modifies those watchers. Note
		2955	also that in that case, you have to re-register any signal watchers.
		2956
		2957	=head3 Watcher-Specific Functions and Data Members
		2958
		2959	=over 4
		2960
		2961	=item ev_fork_init (ev_signal *, callback)
		2962
		2963	Initialises and configures the fork watcher - it has no parameters of any
		2964	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2965	believe me.
		2966
		2967	=back
		2968
		2969
		2970	=head2 C<ev_async> - how to wake up another event loop
		2971
		2972	In general, you cannot use an C<ev_loop> from multiple threads or other
		2973	asynchronous sources such as signal handlers (as opposed to multiple event
		2974	loops - those are of course safe to use in different threads).
		2975
		2976	Sometimes, however, you need to wake up another event loop you do not
		2977	control, for example because it belongs to another thread. This is what
		2978	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2979	can signal it by calling C<ev_async_send>, which is thread- and signal
		2980	safe.
		2981
		2982	This functionality is very similar to C<ev_signal> watchers, as signals,
		2983	too, are asynchronous in nature, and signals, too, will be compressed
		2984	(i.e. the number of callback invocations may be less than the number of
		2985	C<ev_async_sent> calls).
		2986
		2987	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2988	just the default loop.
		2989
		2990	=head3 Queueing
		2991
		2992	C<ev_async> does not support queueing of data in any way. The reason
		2993	is that the author does not know of a simple (or any) algorithm for a
		2994	multiple-writer-single-reader queue that works in all cases and doesn't
		2995	need elaborate support such as pthreads or unportable memory access
		2996	semantics.
		2997
		2998	That means that if you want to queue data, you have to provide your own
		2999	queue. But at least I can tell you how to implement locking around your
		3000	queue:
		3001
		3002	=over 4
		3003
		3004	=item queueing from a signal handler context
		3005
		3006	To implement race-free queueing, you simply add to the queue in the signal
		3007	handler but you block the signal handler in the watcher callback. Here is
		3008	an example that does that for some fictitious SIGUSR1 handler:
		3009
		3010	static ev_async mysig;
		3011
		3012	static void
		3013	sigusr1_handler (void)
		3014	{
		3015	sometype data;
		3016
		3017	// no locking etc.
		3018	queue_put (data);
		3019	ev_async_send (EV_DEFAULT_ &mysig);
		3020	}
		3021
		3022	static void
		3023	mysig_cb (EV_P_ ev_async *w, int revents)
		3024	{
		3025	sometype data;
		3026	sigset_t block, prev;
		3027
		3028	sigemptyset (&block);
		3029	sigaddset (&block, SIGUSR1);
		3030	sigprocmask (SIG_BLOCK, &block, &prev);
		3031
		3032	while (queue_get (&data))
		3033	process (data);
		3034
		3035	if (sigismember (&prev, SIGUSR1)
		3036	sigprocmask (SIG_UNBLOCK, &block, 0);
		3037	}
		3038
		3039	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3040	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3041	either...).
		3042
		3043	=item queueing from a thread context
		3044
		3045	The strategy for threads is different, as you cannot (easily) block
		3046	threads but you can easily preempt them, so to queue safely you need to
		3047	employ a traditional mutex lock, such as in this pthread example:
		3048
		3049	static ev_async mysig;
		3050	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3051
		3052	static void
		3053	otherthread (void)
		3054	{
		3055	// only need to lock the actual queueing operation
		3056	pthread_mutex_lock (&mymutex);
		3057	queue_put (data);
		3058	pthread_mutex_unlock (&mymutex);
		3059
		3060	ev_async_send (EV_DEFAULT_ &mysig);
		3061	}
		3062
		3063	static void
		3064	mysig_cb (EV_P_ ev_async *w, int revents)
		3065	{
		3066	pthread_mutex_lock (&mymutex);
		3067
		3068	while (queue_get (&data))
		3069	process (data);
		3070
		3071	pthread_mutex_unlock (&mymutex);
		3072	}
		3073
		3074	=back
		3075
		3076
		3077	=head3 Watcher-Specific Functions and Data Members
		3078
		3079	=over 4
		3080
		3081	=item ev_async_init (ev_async *, callback)
		3082
		3083	Initialises and configures the async watcher - it has no parameters of any
		3084	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3085	trust me.
		3086
		3087	=item ev_async_send (loop, ev_async *)
		3088
		3089	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3090	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3091	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3092	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3093	section below on what exactly this means).
		3094
		3095	Note that, as with other watchers in libev, multiple events might get
		3096	compressed into a single callback invocation (another way to look at this
		3097	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3098	reset when the event loop detects that).
		3099
		3100	This call incurs the overhead of a system call only once per event loop
		3101	iteration, so while the overhead might be noticeable, it doesn't apply to
		3102	repeated calls to C<ev_async_send> for the same event loop.
		3103
		3104	=item bool = ev_async_pending (ev_async *)
		3105
		3106	Returns a non-zero value when C<ev_async_send> has been called on the
		3107	watcher but the event has not yet been processed (or even noted) by the
		3108	event loop.
		3109
		3110	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3111	the loop iterates next and checks for the watcher to have become active,
		3112	it will reset the flag again. C<ev_async_pending> can be used to very
		3113	quickly check whether invoking the loop might be a good idea.
		3114
		3115	Not that this does I<not> check whether the watcher itself is pending,
		3116	only whether it has been requested to make this watcher pending: there
		3117	is a time window between the event loop checking and resetting the async
		3118	notification, and the callback being invoked.
		3119
		3120	=back
		3121
		3122
793	=head1 OTHER FUNCTIONS	3123	=head1 OTHER FUNCTIONS
794		3124
795	There are some other functions of possible interest. Described. Here. Now.	3125	There are some other functions of possible interest. Described. Here. Now.
796		3126
797	=over 4	3127	=over 4
798		3128
799	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3129	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
800		3130
801	This function combines a simple timer and an I/O watcher, calls your	3131	This function combines a simple timer and an I/O watcher, calls your
802	callback on whichever event happens first and automatically stop both	3132	callback on whichever event happens first and automatically stops both
803	watchers. This is useful if you want to wait for a single event on an fd	3133	watchers. This is useful if you want to wait for a single event on an fd
804	or timeout without having to allocate/configure/start/stop/free one or	3134	or timeout without having to allocate/configure/start/stop/free one or
805	more watchers yourself.	3135	more watchers yourself.
806		3136
807	If C<fd> is less than 0, then no I/O watcher will be started and events	3137	If C<fd> is less than 0, then no I/O watcher will be started and the
808	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3138	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
809	C<events> set will be craeted and started.	3139	the given C<fd> and C<events> set will be created and started.
810		3140
811	If C<timeout> is less than 0, then no timeout watcher will be	3141	If C<timeout> is less than 0, then no timeout watcher will be
812	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3142	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
813	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3143	repeat = 0) will be started. C<0> is a valid timeout.
814	dubious value.
815		3144
816	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3145	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
817	passed an C<revents> set like normal event callbacks (a combination of	3146	passed an C<revents> set like normal event callbacks (a combination of
818	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3147	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
819	value passed to C<ev_once>:	3148	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3149	a timeout and an io event at the same time - you probably should give io
		3150	events precedence.
820		3151
		3152	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3153
821	static void stdin_ready (int revents, void *arg)	3154	static void stdin_ready (int revents, void *arg)
822	{	3155	{
823	if (revents & EV_TIMEOUT)
824	/* doh, nothing entered */;
825	else if (revents & EV_READ)	3156	if (revents & EV_READ)
826	/* stdin might have data for us, joy! */;	3157	/* stdin might have data for us, joy! */;
		3158	else if (revents & EV_TIMEOUT)
		3159	/* doh, nothing entered */;
827	}	3160	}
828		3161
829	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3162	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
830
831	=item ev_feed_event (loop, watcher, int events)
832
833	Feeds the given event set into the event loop, as if the specified event
834	had happened for the specified watcher (which must be a pointer to an
835	initialised but not necessarily started event watcher).
836		3163
837	=item ev_feed_fd_event (loop, int fd, int revents)	3164	=item ev_feed_fd_event (loop, int fd, int revents)
838		3165
839	Feed an event on the given fd, as if a file descriptor backend detected	3166	Feed an event on the given fd, as if a file descriptor backend detected
840	the given events it.	3167	the given events it.
841		3168
842	=item ev_feed_signal_event (loop, int signum)	3169	=item ev_feed_signal_event (loop, int signum)
843		3170
844	Feed an event as if the given signal occured (loop must be the default loop!).	3171	Feed an event as if the given signal occurred (C<loop> must be the default
		3172	loop!).
845		3173
846	=back	3174	=back
		3175
847		3176
848	=head1 LIBEVENT EMULATION	3177	=head1 LIBEVENT EMULATION
849		3178
850	Libev offers a compatibility emulation layer for libevent. It cannot	3179	Libev offers a compatibility emulation layer for libevent. It cannot
851	emulate the internals of libevent, so here are some usage hints:	3180	emulate the internals of libevent, so here are some usage hints:
…		…
863		3192
864	=item * Priorities are not currently supported. Initialising priorities	3193	=item * Priorities are not currently supported. Initialising priorities
865	will fail and all watchers will have the same priority, even though there	3194	will fail and all watchers will have the same priority, even though there
866	is an ev_pri field.	3195	is an ev_pri field.
867		3196
		3197	=item * In libevent, the last base created gets the signals, in libev, the
		3198	first base created (== the default loop) gets the signals.
		3199
868	=item * Other members are not supported.	3200	=item * Other members are not supported.
869		3201
870	=item * The libev emulation is I<not> ABI compatible to libevent, you need	3202	=item * The libev emulation is I<not> ABI compatible to libevent, you need
871	to use the libev header file and library.	3203	to use the libev header file and library.
872		3204
873	=back	3205	=back
874		3206
875	=head1 C++ SUPPORT	3207	=head1 C++ SUPPORT
876		3208
877	TBD.	3209	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3210	you to use some convenience methods to start/stop watchers and also change
		3211	the callback model to a model using method callbacks on objects.
		3212
		3213	To use it,
		3214
		3215	#include <ev++.h>
		3216
		3217	This automatically includes F<ev.h> and puts all of its definitions (many
		3218	of them macros) into the global namespace. All C++ specific things are
		3219	put into the C<ev> namespace. It should support all the same embedding
		3220	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3221
		3222	Care has been taken to keep the overhead low. The only data member the C++
		3223	classes add (compared to plain C-style watchers) is the event loop pointer
		3224	that the watcher is associated with (or no additional members at all if
		3225	you disable C<EV_MULTIPLICITY> when embedding libev).
		3226
		3227	Currently, functions, and static and non-static member functions can be
		3228	used as callbacks. Other types should be easy to add as long as they only
		3229	need one additional pointer for context. If you need support for other
		3230	types of functors please contact the author (preferably after implementing
		3231	it).
		3232
		3233	Here is a list of things available in the C<ev> namespace:
		3234
		3235	=over 4
		3236
		3237	=item C<ev::READ>, C<ev::WRITE> etc.
		3238
		3239	These are just enum values with the same values as the C<EV_READ> etc.
		3240	macros from F<ev.h>.
		3241
		3242	=item C<ev::tstamp>, C<ev::now>
		3243
		3244	Aliases to the same types/functions as with the C<ev_> prefix.
		3245
		3246	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3247
		3248	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3249	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3250	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3251	defines by many implementations.
		3252
		3253	All of those classes have these methods:
		3254
		3255	=over 4
		3256
		3257	=item ev::TYPE::TYPE ()
		3258
		3259	=item ev::TYPE::TYPE (loop)
		3260
		3261	=item ev::TYPE::~TYPE
		3262
		3263	The constructor (optionally) takes an event loop to associate the watcher
		3264	with. If it is omitted, it will use C<EV_DEFAULT>.
		3265
		3266	The constructor calls C<ev_init> for you, which means you have to call the
		3267	C<set> method before starting it.
		3268
		3269	It will not set a callback, however: You have to call the templated C<set>
		3270	method to set a callback before you can start the watcher.
		3271
		3272	(The reason why you have to use a method is a limitation in C++ which does
		3273	not allow explicit template arguments for constructors).
		3274
		3275	The destructor automatically stops the watcher if it is active.
		3276
		3277	=item w->set<class, &class::method> (object *)
		3278
		3279	This method sets the callback method to call. The method has to have a
		3280	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3281	first argument and the C<revents> as second. The object must be given as
		3282	parameter and is stored in the C<data> member of the watcher.
		3283
		3284	This method synthesizes efficient thunking code to call your method from
		3285	the C callback that libev requires. If your compiler can inline your
		3286	callback (i.e. it is visible to it at the place of the C<set> call and
		3287	your compiler is good :), then the method will be fully inlined into the
		3288	thunking function, making it as fast as a direct C callback.
		3289
		3290	Example: simple class declaration and watcher initialisation
		3291
		3292	struct myclass
		3293	{
		3294	void io_cb (ev::io &w, int revents) { }
		3295	}
		3296
		3297	myclass obj;
		3298	ev::io iow;
		3299	iow.set <myclass, &myclass::io_cb> (&obj);
		3300
		3301	=item w->set (object *)
		3302
		3303	This is an B<experimental> feature that might go away in a future version.
		3304
		3305	This is a variation of a method callback - leaving out the method to call
		3306	will default the method to C<operator ()>, which makes it possible to use
		3307	functor objects without having to manually specify the C<operator ()> all
		3308	the time. Incidentally, you can then also leave out the template argument
		3309	list.
		3310
		3311	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3312	int revents)>.
		3313
		3314	See the method-C<set> above for more details.
		3315
		3316	Example: use a functor object as callback.
		3317
		3318	struct myfunctor
		3319	{
		3320	void operator() (ev::io &w, int revents)
		3321	{
		3322	...
		3323	}
		3324	}
		3325
		3326	myfunctor f;
		3327
		3328	ev::io w;
		3329	w.set (&f);
		3330
		3331	=item w->set<function> (void *data = 0)
		3332
		3333	Also sets a callback, but uses a static method or plain function as
		3334	callback. The optional C<data> argument will be stored in the watcher's
		3335	C<data> member and is free for you to use.
		3336
		3337	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3338
		3339	See the method-C<set> above for more details.
		3340
		3341	Example: Use a plain function as callback.
		3342
		3343	static void io_cb (ev::io &w, int revents) { }
		3344	iow.set <io_cb> ();
		3345
		3346	=item w->set (loop)
		3347
		3348	Associates a different C<struct ev_loop> with this watcher. You can only
		3349	do this when the watcher is inactive (and not pending either).
		3350
		3351	=item w->set ([arguments])
		3352
		3353	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		3354	called at least once. Unlike the C counterpart, an active watcher gets
		3355	automatically stopped and restarted when reconfiguring it with this
		3356	method.
		3357
		3358	=item w->start ()
		3359
		3360	Starts the watcher. Note that there is no C<loop> argument, as the
		3361	constructor already stores the event loop.
		3362
		3363	=item w->stop ()
		3364
		3365	Stops the watcher if it is active. Again, no C<loop> argument.
		3366
		3367	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3368
		3369	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3370	C<ev_TYPE_again> function.
		3371
		3372	=item w->sweep () (C<ev::embed> only)
		3373
		3374	Invokes C<ev_embed_sweep>.
		3375
		3376	=item w->update () (C<ev::stat> only)
		3377
		3378	Invokes C<ev_stat_stat>.
		3379
		3380	=back
		3381
		3382	=back
		3383
		3384	Example: Define a class with an IO and idle watcher, start one of them in
		3385	the constructor.
		3386
		3387	class myclass
		3388	{
		3389	ev::io io ; void io_cb (ev::io &w, int revents);
		3390	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3391
		3392	myclass (int fd)
		3393	{
		3394	io .set <myclass, &myclass::io_cb > (this);
		3395	idle.set <myclass, &myclass::idle_cb> (this);
		3396
		3397	io.start (fd, ev::READ);
		3398	}
		3399	};
		3400
		3401
		3402	=head1 OTHER LANGUAGE BINDINGS
		3403
		3404	Libev does not offer other language bindings itself, but bindings for a
		3405	number of languages exist in the form of third-party packages. If you know
		3406	any interesting language binding in addition to the ones listed here, drop
		3407	me a note.
		3408
		3409	=over 4
		3410
		3411	=item Perl
		3412
		3413	The EV module implements the full libev API and is actually used to test
		3414	libev. EV is developed together with libev. Apart from the EV core module,
		3415	there are additional modules that implement libev-compatible interfaces
		3416	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3417	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3418	and C<EV::Glib>).
		3419
		3420	It can be found and installed via CPAN, its homepage is at
		3421	L<http://software.schmorp.de/pkg/EV>.
		3422
		3423	=item Python
		3424
		3425	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3426	seems to be quite complete and well-documented.
		3427
		3428	=item Ruby
		3429
		3430	Tony Arcieri has written a ruby extension that offers access to a subset
		3431	of the libev API and adds file handle abstractions, asynchronous DNS and
		3432	more on top of it. It can be found via gem servers. Its homepage is at
		3433	L<http://rev.rubyforge.org/>.
		3434
		3435	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3436	makes rev work even on mingw.
		3437
		3438	=item Haskell
		3439
		3440	A haskell binding to libev is available at
		3441	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3442
		3443	=item D
		3444
		3445	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3446	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3447
		3448	=item Ocaml
		3449
		3450	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3451	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3452
		3453	=item Lua
		3454
		3455	Brian Maher has written a partial interface to libev
		3456	for lua (only C<ev_io> and C<ev_timer>), to be found at
		3457	L<http://github.com/brimworks/lua-ev>.
		3458
		3459	=back
		3460
		3461
		3462	=head1 MACRO MAGIC
		3463
		3464	Libev can be compiled with a variety of options, the most fundamental
		3465	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3466	functions and callbacks have an initial C<struct ev_loop *> argument.
		3467
		3468	To make it easier to write programs that cope with either variant, the
		3469	following macros are defined:
		3470
		3471	=over 4
		3472
		3473	=item C<EV_A>, C<EV_A_>
		3474
		3475	This provides the loop I<argument> for functions, if one is required ("ev
		3476	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3477	C<EV_A_> is used when other arguments are following. Example:
		3478
		3479	ev_unref (EV_A);
		3480	ev_timer_add (EV_A_ watcher);
		3481	ev_loop (EV_A_ 0);
		3482
		3483	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3484	which is often provided by the following macro.
		3485
		3486	=item C<EV_P>, C<EV_P_>
		3487
		3488	This provides the loop I<parameter> for functions, if one is required ("ev
		3489	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3490	C<EV_P_> is used when other parameters are following. Example:
		3491
		3492	// this is how ev_unref is being declared
		3493	static void ev_unref (EV_P);
		3494
		3495	// this is how you can declare your typical callback
		3496	static void cb (EV_P_ ev_timer *w, int revents)
		3497
		3498	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3499	suitable for use with C<EV_A>.
		3500
		3501	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3502
		3503	Similar to the other two macros, this gives you the value of the default
		3504	loop, if multiple loops are supported ("ev loop default").
		3505
		3506	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3507
		3508	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3509	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3510	is undefined when the default loop has not been initialised by a previous
		3511	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3512
		3513	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3514	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3515
		3516	=back
		3517
		3518	Example: Declare and initialise a check watcher, utilising the above
		3519	macros so it will work regardless of whether multiple loops are supported
		3520	or not.
		3521
		3522	static void
		3523	check_cb (EV_P_ ev_timer *w, int revents)
		3524	{
		3525	ev_check_stop (EV_A_ w);
		3526	}
		3527
		3528	ev_check check;
		3529	ev_check_init (&check, check_cb);
		3530	ev_check_start (EV_DEFAULT_ &check);
		3531	ev_loop (EV_DEFAULT_ 0);
		3532
		3533	=head1 EMBEDDING
		3534
		3535	Libev can (and often is) directly embedded into host
		3536	applications. Examples of applications that embed it include the Deliantra
		3537	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3538	and rxvt-unicode.
		3539
		3540	The goal is to enable you to just copy the necessary files into your
		3541	source directory without having to change even a single line in them, so
		3542	you can easily upgrade by simply copying (or having a checked-out copy of
		3543	libev somewhere in your source tree).
		3544
		3545	=head2 FILESETS
		3546
		3547	Depending on what features you need you need to include one or more sets of files
		3548	in your application.
		3549
		3550	=head3 CORE EVENT LOOP
		3551
		3552	To include only the libev core (all the C<ev_*> functions), with manual
		3553	configuration (no autoconf):
		3554
		3555	#define EV_STANDALONE 1
		3556	#include "ev.c"
		3557
		3558	This will automatically include F<ev.h>, too, and should be done in a
		3559	single C source file only to provide the function implementations. To use
		3560	it, do the same for F<ev.h> in all files wishing to use this API (best
		3561	done by writing a wrapper around F<ev.h> that you can include instead and
		3562	where you can put other configuration options):
		3563
		3564	#define EV_STANDALONE 1
		3565	#include "ev.h"
		3566
		3567	Both header files and implementation files can be compiled with a C++
		3568	compiler (at least, that's a stated goal, and breakage will be treated
		3569	as a bug).
		3570
		3571	You need the following files in your source tree, or in a directory
		3572	in your include path (e.g. in libev/ when using -Ilibev):
		3573
		3574	ev.h
		3575	ev.c
		3576	ev_vars.h
		3577	ev_wrap.h
		3578
		3579	ev_win32.c required on win32 platforms only
		3580
		3581	ev_select.c only when select backend is enabled (which is enabled by default)
		3582	ev_poll.c only when poll backend is enabled (disabled by default)
		3583	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3584	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3585	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3586
		3587	F<ev.c> includes the backend files directly when enabled, so you only need
		3588	to compile this single file.
		3589
		3590	=head3 LIBEVENT COMPATIBILITY API
		3591
		3592	To include the libevent compatibility API, also include:
		3593
		3594	#include "event.c"
		3595
		3596	in the file including F<ev.c>, and:
		3597
		3598	#include "event.h"
		3599
		3600	in the files that want to use the libevent API. This also includes F<ev.h>.
		3601
		3602	You need the following additional files for this:
		3603
		3604	event.h
		3605	event.c
		3606
		3607	=head3 AUTOCONF SUPPORT
		3608
		3609	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3610	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3611	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3612	include F<config.h> and configure itself accordingly.
		3613
		3614	For this of course you need the m4 file:
		3615
		3616	libev.m4
		3617
		3618	=head2 PREPROCESSOR SYMBOLS/MACROS
		3619
		3620	Libev can be configured via a variety of preprocessor symbols you have to
		3621	define before including any of its files. The default in the absence of
		3622	autoconf is documented for every option.
		3623
		3624	=over 4
		3625
		3626	=item EV_STANDALONE
		3627
		3628	Must always be C<1> if you do not use autoconf configuration, which
		3629	keeps libev from including F<config.h>, and it also defines dummy
		3630	implementations for some libevent functions (such as logging, which is not
		3631	supported). It will also not define any of the structs usually found in
		3632	F<event.h> that are not directly supported by the libev core alone.
		3633
		3634	In standalone mode, libev will still try to automatically deduce the
		3635	configuration, but has to be more conservative.
		3636
		3637	=item EV_USE_MONOTONIC
		3638
		3639	If defined to be C<1>, libev will try to detect the availability of the
		3640	monotonic clock option at both compile time and runtime. Otherwise no
		3641	use of the monotonic clock option will be attempted. If you enable this,
		3642	you usually have to link against librt or something similar. Enabling it
		3643	when the functionality isn't available is safe, though, although you have
		3644	to make sure you link against any libraries where the C<clock_gettime>
		3645	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3646
		3647	=item EV_USE_REALTIME
		3648
		3649	If defined to be C<1>, libev will try to detect the availability of the
		3650	real-time clock option at compile time (and assume its availability
		3651	at runtime if successful). Otherwise no use of the real-time clock
		3652	option will be attempted. This effectively replaces C<gettimeofday>
		3653	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3654	correctness. See the note about libraries in the description of
		3655	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3656	C<EV_USE_CLOCK_SYSCALL>.
		3657
		3658	=item EV_USE_CLOCK_SYSCALL
		3659
		3660	If defined to be C<1>, libev will try to use a direct syscall instead
		3661	of calling the system-provided C<clock_gettime> function. This option
		3662	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3663	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3664	programs needlessly. Using a direct syscall is slightly slower (in
		3665	theory), because no optimised vdso implementation can be used, but avoids
		3666	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3667	higher, as it simplifies linking (no need for C<-lrt>).
		3668
		3669	=item EV_USE_NANOSLEEP
		3670
		3671	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3672	and will use it for delays. Otherwise it will use C<select ()>.
		3673
		3674	=item EV_USE_EVENTFD
		3675
		3676	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3677	available and will probe for kernel support at runtime. This will improve
		3678	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3679	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3680	2.7 or newer, otherwise disabled.
		3681
		3682	=item EV_USE_SELECT
		3683
		3684	If undefined or defined to be C<1>, libev will compile in support for the
		3685	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3686	other method takes over, select will be it. Otherwise the select backend
		3687	will not be compiled in.
		3688
		3689	=item EV_SELECT_USE_FD_SET
		3690
		3691	If defined to C<1>, then the select backend will use the system C<fd_set>
		3692	structure. This is useful if libev doesn't compile due to a missing
		3693	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3694	on exotic systems. This usually limits the range of file descriptors to
		3695	some low limit such as 1024 or might have other limitations (winsocket
		3696	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3697	configures the maximum size of the C<fd_set>.
		3698
		3699	=item EV_SELECT_IS_WINSOCKET
		3700
		3701	When defined to C<1>, the select backend will assume that
		3702	select/socket/connect etc. don't understand file descriptors but
		3703	wants osf handles on win32 (this is the case when the select to
		3704	be used is the winsock select). This means that it will call
		3705	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3706	it is assumed that all these functions actually work on fds, even
		3707	on win32. Should not be defined on non-win32 platforms.
		3708
		3709	=item EV_FD_TO_WIN32_HANDLE(fd)
		3710
		3711	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3712	file descriptors to socket handles. When not defining this symbol (the
		3713	default), then libev will call C<_get_osfhandle>, which is usually
		3714	correct. In some cases, programs use their own file descriptor management,
		3715	in which case they can provide this function to map fds to socket handles.
		3716
		3717	=item EV_WIN32_HANDLE_TO_FD(handle)
		3718
		3719	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		3720	using the standard C<_open_osfhandle> function. For programs implementing
		3721	their own fd to handle mapping, overwriting this function makes it easier
		3722	to do so. This can be done by defining this macro to an appropriate value.
		3723
		3724	=item EV_WIN32_CLOSE_FD(fd)
		3725
		3726	If programs implement their own fd to handle mapping on win32, then this
		3727	macro can be used to override the C<close> function, useful to unregister
		3728	file descriptors again. Note that the replacement function has to close
		3729	the underlying OS handle.
		3730
		3731	=item EV_USE_POLL
		3732
		3733	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3734	backend. Otherwise it will be enabled on non-win32 platforms. It
		3735	takes precedence over select.
		3736
		3737	=item EV_USE_EPOLL
		3738
		3739	If defined to be C<1>, libev will compile in support for the Linux
		3740	C<epoll>(7) backend. Its availability will be detected at runtime,
		3741	otherwise another method will be used as fallback. This is the preferred
		3742	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3743	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3744
		3745	=item EV_USE_KQUEUE
		3746
		3747	If defined to be C<1>, libev will compile in support for the BSD style
		3748	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3749	otherwise another method will be used as fallback. This is the preferred
		3750	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3751	supports some types of fds correctly (the only platform we found that
		3752	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3753	not be used unless explicitly requested. The best way to use it is to find
		3754	out whether kqueue supports your type of fd properly and use an embedded
		3755	kqueue loop.
		3756
		3757	=item EV_USE_PORT
		3758
		3759	If defined to be C<1>, libev will compile in support for the Solaris
		3760	10 port style backend. Its availability will be detected at runtime,
		3761	otherwise another method will be used as fallback. This is the preferred
		3762	backend for Solaris 10 systems.
		3763
		3764	=item EV_USE_DEVPOLL
		3765
		3766	Reserved for future expansion, works like the USE symbols above.
		3767
		3768	=item EV_USE_INOTIFY
		3769
		3770	If defined to be C<1>, libev will compile in support for the Linux inotify
		3771	interface to speed up C<ev_stat> watchers. Its actual availability will
		3772	be detected at runtime. If undefined, it will be enabled if the headers
		3773	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3774
		3775	=item EV_ATOMIC_T
		3776
		3777	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3778	access is atomic with respect to other threads or signal contexts. No such
		3779	type is easily found in the C language, so you can provide your own type
		3780	that you know is safe for your purposes. It is used both for signal handler "locking"
		3781	as well as for signal and thread safety in C<ev_async> watchers.
		3782
		3783	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3784	(from F<signal.h>), which is usually good enough on most platforms.
		3785
		3786	=item EV_H
		3787
		3788	The name of the F<ev.h> header file used to include it. The default if
		3789	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3790	used to virtually rename the F<ev.h> header file in case of conflicts.
		3791
		3792	=item EV_CONFIG_H
		3793
		3794	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3795	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3796	C<EV_H>, above.
		3797
		3798	=item EV_EVENT_H
		3799
		3800	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3801	of how the F<event.h> header can be found, the default is C<"event.h">.
		3802
		3803	=item EV_PROTOTYPES
		3804
		3805	If defined to be C<0>, then F<ev.h> will not define any function
		3806	prototypes, but still define all the structs and other symbols. This is
		3807	occasionally useful if you want to provide your own wrapper functions
		3808	around libev functions.
		3809
		3810	=item EV_MULTIPLICITY
		3811
		3812	If undefined or defined to C<1>, then all event-loop-specific functions
		3813	will have the C<struct ev_loop *> as first argument, and you can create
		3814	additional independent event loops. Otherwise there will be no support
		3815	for multiple event loops and there is no first event loop pointer
		3816	argument. Instead, all functions act on the single default loop.
		3817
		3818	=item EV_MINPRI
		3819
		3820	=item EV_MAXPRI
		3821
		3822	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3823	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3824	provide for more priorities by overriding those symbols (usually defined
		3825	to be C<-2> and C<2>, respectively).
		3826
		3827	When doing priority-based operations, libev usually has to linearly search
		3828	all the priorities, so having many of them (hundreds) uses a lot of space
		3829	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3830	fine.
		3831
		3832	If your embedding application does not need any priorities, defining these
		3833	both to C<0> will save some memory and CPU.
		3834
		3835	=item EV_PERIODIC_ENABLE
		3836
		3837	If undefined or defined to be C<1>, then periodic timers are supported. If
		3838	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3839	code.
		3840
		3841	=item EV_IDLE_ENABLE
		3842
		3843	If undefined or defined to be C<1>, then idle watchers are supported. If
		3844	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3845	code.
		3846
		3847	=item EV_EMBED_ENABLE
		3848
		3849	If undefined or defined to be C<1>, then embed watchers are supported. If
		3850	defined to be C<0>, then they are not. Embed watchers rely on most other
		3851	watcher types, which therefore must not be disabled.
		3852
		3853	=item EV_STAT_ENABLE
		3854
		3855	If undefined or defined to be C<1>, then stat watchers are supported. If
		3856	defined to be C<0>, then they are not.
		3857
		3858	=item EV_FORK_ENABLE
		3859
		3860	If undefined or defined to be C<1>, then fork watchers are supported. If
		3861	defined to be C<0>, then they are not.
		3862
		3863	=item EV_ASYNC_ENABLE
		3864
		3865	If undefined or defined to be C<1>, then async watchers are supported. If
		3866	defined to be C<0>, then they are not.
		3867
		3868	=item EV_MINIMAL
		3869
		3870	If you need to shave off some kilobytes of code at the expense of some
		3871	speed (but with the full API), define this symbol to C<1>. Currently this
		3872	is used to override some inlining decisions, saves roughly 30% code size
		3873	on amd64. It also selects a much smaller 2-heap for timer management over
		3874	the default 4-heap.
		3875
		3876	You can save even more by disabling watcher types you do not need
		3877	and setting C<EV_MAXPRI> == C<EV_MINPRI>. Also, disabling C<assert>
		3878	(C<-DNDEBUG>) will usually reduce code size a lot.
		3879
		3880	Defining C<EV_MINIMAL> to C<2> will additionally reduce the core API to
		3881	provide a bare-bones event library. See C<ev.h> for details on what parts
		3882	of the API are still available, and do not complain if this subset changes
		3883	over time.
		3884
		3885	=item EV_NSIG
		3886
		3887	The highest supported signal number, +1 (or, the number of
		3888	signals): Normally, libev tries to deduce the maximum number of signals
		3889	automatically, but sometimes this fails, in which case it can be
		3890	specified. Also, using a lower number than detected (C<32> should be
		3891	good for about any system in existance) can save some memory, as libev
		3892	statically allocates some 12-24 bytes per signal number.
		3893
		3894	=item EV_PID_HASHSIZE
		3895
		3896	C<ev_child> watchers use a small hash table to distribute workload by
		3897	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3898	than enough. If you need to manage thousands of children you might want to
		3899	increase this value (I<must> be a power of two).
		3900
		3901	=item EV_INOTIFY_HASHSIZE
		3902
		3903	C<ev_stat> watchers use a small hash table to distribute workload by
		3904	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3905	usually more than enough. If you need to manage thousands of C<ev_stat>
		3906	watchers you might want to increase this value (I<must> be a power of
		3907	two).
		3908
		3909	=item EV_USE_4HEAP
		3910
		3911	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3912	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3913	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3914	faster performance with many (thousands) of watchers.
		3915
		3916	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3917	(disabled).
		3918
		3919	=item EV_HEAP_CACHE_AT
		3920
		3921	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3922	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3923	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3924	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3925	but avoids random read accesses on heap changes. This improves performance
		3926	noticeably with many (hundreds) of watchers.
		3927
		3928	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3929	(disabled).
		3930
		3931	=item EV_VERIFY
		3932
		3933	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3934	be done: If set to C<0>, no internal verification code will be compiled
		3935	in. If set to C<1>, then verification code will be compiled in, but not
		3936	called. If set to C<2>, then the internal verification code will be
		3937	called once per loop, which can slow down libev. If set to C<3>, then the
		3938	verification code will be called very frequently, which will slow down
		3939	libev considerably.
		3940
		3941	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3942	C<0>.
		3943
		3944	=item EV_COMMON
		3945
		3946	By default, all watchers have a C<void *data> member. By redefining
		3947	this macro to a something else you can include more and other types of
		3948	members. You have to define it each time you include one of the files,
		3949	though, and it must be identical each time.
		3950
		3951	For example, the perl EV module uses something like this:
		3952
		3953	#define EV_COMMON \
		3954	SV *self; /* contains this struct */ \
		3955	SV *cb_sv, *fh /* note no trailing ";" */
		3956
		3957	=item EV_CB_DECLARE (type)
		3958
		3959	=item EV_CB_INVOKE (watcher, revents)
		3960
		3961	=item ev_set_cb (ev, cb)
		3962
		3963	Can be used to change the callback member declaration in each watcher,
		3964	and the way callbacks are invoked and set. Must expand to a struct member
		3965	definition and a statement, respectively. See the F<ev.h> header file for
		3966	their default definitions. One possible use for overriding these is to
		3967	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3968	method calls instead of plain function calls in C++.
		3969
		3970	=back
		3971
		3972	=head2 EXPORTED API SYMBOLS
		3973
		3974	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3975	exported symbols, you can use the provided F<Symbol.*> files which list
		3976	all public symbols, one per line:
		3977
		3978	Symbols.ev for libev proper
		3979	Symbols.event for the libevent emulation
		3980
		3981	This can also be used to rename all public symbols to avoid clashes with
		3982	multiple versions of libev linked together (which is obviously bad in
		3983	itself, but sometimes it is inconvenient to avoid this).
		3984
		3985	A sed command like this will create wrapper C<#define>'s that you need to
		3986	include before including F<ev.h>:
		3987
		3988	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3989
		3990	This would create a file F<wrap.h> which essentially looks like this:
		3991
		3992	#define ev_backend myprefix_ev_backend
		3993	#define ev_check_start myprefix_ev_check_start
		3994	#define ev_check_stop myprefix_ev_check_stop
		3995	...
		3996
		3997	=head2 EXAMPLES
		3998
		3999	For a real-world example of a program the includes libev
		4000	verbatim, you can have a look at the EV perl module
		4001	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		4002	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		4003	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		4004	will be compiled. It is pretty complex because it provides its own header
		4005	file.
		4006
		4007	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		4008	that everybody includes and which overrides some configure choices:
		4009
		4010	#define EV_MINIMAL 1
		4011	#define EV_USE_POLL 0
		4012	#define EV_MULTIPLICITY 0
		4013	#define EV_PERIODIC_ENABLE 0
		4014	#define EV_STAT_ENABLE 0
		4015	#define EV_FORK_ENABLE 0
		4016	#define EV_CONFIG_H <config.h>
		4017	#define EV_MINPRI 0
		4018	#define EV_MAXPRI 0
		4019
		4020	#include "ev++.h"
		4021
		4022	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		4023
		4024	#include "ev_cpp.h"
		4025	#include "ev.c"
		4026
		4027	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		4028
		4029	=head2 THREADS AND COROUTINES
		4030
		4031	=head3 THREADS
		4032
		4033	All libev functions are reentrant and thread-safe unless explicitly
		4034	documented otherwise, but libev implements no locking itself. This means
		4035	that you can use as many loops as you want in parallel, as long as there
		4036	are no concurrent calls into any libev function with the same loop
		4037	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		4038	of course): libev guarantees that different event loops share no data
		4039	structures that need any locking.
		4040
		4041	Or to put it differently: calls with different loop parameters can be done
		4042	concurrently from multiple threads, calls with the same loop parameter
		4043	must be done serially (but can be done from different threads, as long as
		4044	only one thread ever is inside a call at any point in time, e.g. by using
		4045	a mutex per loop).
		4046
		4047	Specifically to support threads (and signal handlers), libev implements
		4048	so-called C<ev_async> watchers, which allow some limited form of
		4049	concurrency on the same event loop, namely waking it up "from the
		4050	outside".
		4051
		4052	If you want to know which design (one loop, locking, or multiple loops
		4053	without or something else still) is best for your problem, then I cannot
		4054	help you, but here is some generic advice:
		4055
		4056	=over 4
		4057
		4058	=item * most applications have a main thread: use the default libev loop
		4059	in that thread, or create a separate thread running only the default loop.
		4060
		4061	This helps integrating other libraries or software modules that use libev
		4062	themselves and don't care/know about threading.
		4063
		4064	=item * one loop per thread is usually a good model.
		4065
		4066	Doing this is almost never wrong, sometimes a better-performance model
		4067	exists, but it is always a good start.
		4068
		4069	=item * other models exist, such as the leader/follower pattern, where one
		4070	loop is handed through multiple threads in a kind of round-robin fashion.
		4071
		4072	Choosing a model is hard - look around, learn, know that usually you can do
		4073	better than you currently do :-)
		4074
		4075	=item * often you need to talk to some other thread which blocks in the
		4076	event loop.
		4077
		4078	C<ev_async> watchers can be used to wake them up from other threads safely
		4079	(or from signal contexts...).
		4080
		4081	An example use would be to communicate signals or other events that only
		4082	work in the default loop by registering the signal watcher with the
		4083	default loop and triggering an C<ev_async> watcher from the default loop
		4084	watcher callback into the event loop interested in the signal.
		4085
		4086	=back
		4087
		4088	=head4 THREAD LOCKING EXAMPLE
		4089
		4090	Here is a fictitious example of how to run an event loop in a different
		4091	thread than where callbacks are being invoked and watchers are
		4092	created/added/removed.
		4093
		4094	For a real-world example, see the C<EV::Loop::Async> perl module,
		4095	which uses exactly this technique (which is suited for many high-level
		4096	languages).
		4097
		4098	The example uses a pthread mutex to protect the loop data, a condition
		4099	variable to wait for callback invocations, an async watcher to notify the
		4100	event loop thread and an unspecified mechanism to wake up the main thread.
		4101
		4102	First, you need to associate some data with the event loop:
		4103
		4104	typedef struct {
		4105	mutex_t lock; /* global loop lock */
		4106	ev_async async_w;
		4107	thread_t tid;
		4108	cond_t invoke_cv;
		4109	} userdata;
		4110
		4111	void prepare_loop (EV_P)
		4112	{
		4113	// for simplicity, we use a static userdata struct.
		4114	static userdata u;
		4115
		4116	ev_async_init (&u->async_w, async_cb);
		4117	ev_async_start (EV_A_ &u->async_w);
		4118
		4119	pthread_mutex_init (&u->lock, 0);
		4120	pthread_cond_init (&u->invoke_cv, 0);
		4121
		4122	// now associate this with the loop
		4123	ev_set_userdata (EV_A_ u);
		4124	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4125	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4126
		4127	// then create the thread running ev_loop
		4128	pthread_create (&u->tid, 0, l_run, EV_A);
		4129	}
		4130
		4131	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4132	solely to wake up the event loop so it takes notice of any new watchers
		4133	that might have been added:
		4134
		4135	static void
		4136	async_cb (EV_P_ ev_async *w, int revents)
		4137	{
		4138	// just used for the side effects
		4139	}
		4140
		4141	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4142	protecting the loop data, respectively.
		4143
		4144	static void
		4145	l_release (EV_P)
		4146	{
		4147	userdata *u = ev_userdata (EV_A);
		4148	pthread_mutex_unlock (&u->lock);
		4149	}
		4150
		4151	static void
		4152	l_acquire (EV_P)
		4153	{
		4154	userdata *u = ev_userdata (EV_A);
		4155	pthread_mutex_lock (&u->lock);
		4156	}
		4157
		4158	The event loop thread first acquires the mutex, and then jumps straight
		4159	into C<ev_loop>:
		4160
		4161	void *
		4162	l_run (void *thr_arg)
		4163	{
		4164	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4165
		4166	l_acquire (EV_A);
		4167	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4168	ev_loop (EV_A_ 0);
		4169	l_release (EV_A);
		4170
		4171	return 0;
		4172	}
		4173
		4174	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4175	signal the main thread via some unspecified mechanism (signals? pipe
		4176	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4177	have been called (in a while loop because a) spurious wakeups are possible
		4178	and b) skipping inter-thread-communication when there are no pending
		4179	watchers is very beneficial):
		4180
		4181	static void
		4182	l_invoke (EV_P)
		4183	{
		4184	userdata *u = ev_userdata (EV_A);
		4185
		4186	while (ev_pending_count (EV_A))
		4187	{
		4188	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4189	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4190	}
		4191	}
		4192
		4193	Now, whenever the main thread gets told to invoke pending watchers, it
		4194	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4195	thread to continue:
		4196
		4197	static void
		4198	real_invoke_pending (EV_P)
		4199	{
		4200	userdata *u = ev_userdata (EV_A);
		4201
		4202	pthread_mutex_lock (&u->lock);
		4203	ev_invoke_pending (EV_A);
		4204	pthread_cond_signal (&u->invoke_cv);
		4205	pthread_mutex_unlock (&u->lock);
		4206	}
		4207
		4208	Whenever you want to start/stop a watcher or do other modifications to an
		4209	event loop, you will now have to lock:
		4210
		4211	ev_timer timeout_watcher;
		4212	userdata *u = ev_userdata (EV_A);
		4213
		4214	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4215
		4216	pthread_mutex_lock (&u->lock);
		4217	ev_timer_start (EV_A_ &timeout_watcher);
		4218	ev_async_send (EV_A_ &u->async_w);
		4219	pthread_mutex_unlock (&u->lock);
		4220
		4221	Note that sending the C<ev_async> watcher is required because otherwise
		4222	an event loop currently blocking in the kernel will have no knowledge
		4223	about the newly added timer. By waking up the loop it will pick up any new
		4224	watchers in the next event loop iteration.
		4225
		4226	=head3 COROUTINES
		4227
		4228	Libev is very accommodating to coroutines ("cooperative threads"):
		4229	libev fully supports nesting calls to its functions from different
		4230	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		4231	different coroutines, and switch freely between both coroutines running
		4232	the loop, as long as you don't confuse yourself). The only exception is
		4233	that you must not do this from C<ev_periodic> reschedule callbacks.
		4234
		4235	Care has been taken to ensure that libev does not keep local state inside
		4236	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		4237	they do not call any callbacks.
		4238
		4239	=head2 COMPILER WARNINGS
		4240
		4241	Depending on your compiler and compiler settings, you might get no or a
		4242	lot of warnings when compiling libev code. Some people are apparently
		4243	scared by this.
		4244
		4245	However, these are unavoidable for many reasons. For one, each compiler
		4246	has different warnings, and each user has different tastes regarding
		4247	warning options. "Warn-free" code therefore cannot be a goal except when
		4248	targeting a specific compiler and compiler-version.
		4249
		4250	Another reason is that some compiler warnings require elaborate
		4251	workarounds, or other changes to the code that make it less clear and less
		4252	maintainable.
		4253
		4254	And of course, some compiler warnings are just plain stupid, or simply
		4255	wrong (because they don't actually warn about the condition their message
		4256	seems to warn about). For example, certain older gcc versions had some
		4257	warnings that resulted an extreme number of false positives. These have
		4258	been fixed, but some people still insist on making code warn-free with
		4259	such buggy versions.
		4260
		4261	While libev is written to generate as few warnings as possible,
		4262	"warn-free" code is not a goal, and it is recommended not to build libev
		4263	with any compiler warnings enabled unless you are prepared to cope with
		4264	them (e.g. by ignoring them). Remember that warnings are just that:
		4265	warnings, not errors, or proof of bugs.
		4266
		4267
		4268	=head2 VALGRIND
		4269
		4270	Valgrind has a special section here because it is a popular tool that is
		4271	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4272
		4273	If you think you found a bug (memory leak, uninitialised data access etc.)
		4274	in libev, then check twice: If valgrind reports something like:
		4275
		4276	==2274== definitely lost: 0 bytes in 0 blocks.
		4277	==2274== possibly lost: 0 bytes in 0 blocks.
		4278	==2274== still reachable: 256 bytes in 1 blocks.
		4279
		4280	Then there is no memory leak, just as memory accounted to global variables
		4281	is not a memleak - the memory is still being referenced, and didn't leak.
		4282
		4283	Similarly, under some circumstances, valgrind might report kernel bugs
		4284	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4285	although an acceptable workaround has been found here), or it might be
		4286	confused.
		4287
		4288	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4289	make it into some kind of religion.
		4290
		4291	If you are unsure about something, feel free to contact the mailing list
		4292	with the full valgrind report and an explanation on why you think this
		4293	is a bug in libev (best check the archives, too :). However, don't be
		4294	annoyed when you get a brisk "this is no bug" answer and take the chance
		4295	of learning how to interpret valgrind properly.
		4296
		4297	If you need, for some reason, empty reports from valgrind for your project
		4298	I suggest using suppression lists.
		4299
		4300
		4301	=head1 PORTABILITY NOTES
		4302
		4303	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4304
		4305	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4306	requires, and its I/O model is fundamentally incompatible with the POSIX
		4307	model. Libev still offers limited functionality on this platform in
		4308	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4309	descriptors. This only applies when using Win32 natively, not when using
		4310	e.g. cygwin.
		4311
		4312	Lifting these limitations would basically require the full
		4313	re-implementation of the I/O system. If you are into these kinds of
		4314	things, then note that glib does exactly that for you in a very portable
		4315	way (note also that glib is the slowest event library known to man).
		4316
		4317	There is no supported compilation method available on windows except
		4318	embedding it into other applications.
		4319
		4320	Sensible signal handling is officially unsupported by Microsoft - libev
		4321	tries its best, but under most conditions, signals will simply not work.
		4322
		4323	Not a libev limitation but worth mentioning: windows apparently doesn't
		4324	accept large writes: instead of resulting in a partial write, windows will
		4325	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4326	so make sure you only write small amounts into your sockets (less than a
		4327	megabyte seems safe, but this apparently depends on the amount of memory
		4328	available).
		4329
		4330	Due to the many, low, and arbitrary limits on the win32 platform and
		4331	the abysmal performance of winsockets, using a large number of sockets
		4332	is not recommended (and not reasonable). If your program needs to use
		4333	more than a hundred or so sockets, then likely it needs to use a totally
		4334	different implementation for windows, as libev offers the POSIX readiness
		4335	notification model, which cannot be implemented efficiently on windows
		4336	(due to Microsoft monopoly games).
		4337
		4338	A typical way to use libev under windows is to embed it (see the embedding
		4339	section for details) and use the following F<evwrap.h> header file instead
		4340	of F<ev.h>:
		4341
		4342	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4343	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4344
		4345	#include "ev.h"
		4346
		4347	And compile the following F<evwrap.c> file into your project (make sure
		4348	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4349
		4350	#include "evwrap.h"
		4351	#include "ev.c"
		4352
		4353	=over 4
		4354
		4355	=item The winsocket select function
		4356
		4357	The winsocket C<select> function doesn't follow POSIX in that it
		4358	requires socket I<handles> and not socket I<file descriptors> (it is
		4359	also extremely buggy). This makes select very inefficient, and also
		4360	requires a mapping from file descriptors to socket handles (the Microsoft
		4361	C runtime provides the function C<_open_osfhandle> for this). See the
		4362	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4363	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4364
		4365	The configuration for a "naked" win32 using the Microsoft runtime
		4366	libraries and raw winsocket select is:
		4367
		4368	#define EV_USE_SELECT 1
		4369	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4370
		4371	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4372	complexity in the O(n²) range when using win32.
		4373
		4374	=item Limited number of file descriptors
		4375
		4376	Windows has numerous arbitrary (and low) limits on things.
		4377
		4378	Early versions of winsocket's select only supported waiting for a maximum
		4379	of C<64> handles (probably owning to the fact that all windows kernels
		4380	can only wait for C<64> things at the same time internally; Microsoft
		4381	recommends spawning a chain of threads and wait for 63 handles and the
		4382	previous thread in each. Sounds great!).
		4383
		4384	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4385	to some high number (e.g. C<2048>) before compiling the winsocket select
		4386	call (which might be in libev or elsewhere, for example, perl and many
		4387	other interpreters do their own select emulation on windows).
		4388
		4389	Another limit is the number of file descriptors in the Microsoft runtime
		4390	libraries, which by default is C<64> (there must be a hidden I<64>
		4391	fetish or something like this inside Microsoft). You can increase this
		4392	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4393	(another arbitrary limit), but is broken in many versions of the Microsoft
		4394	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4395	(depending on windows version and/or the phase of the moon). To get more,
		4396	you need to wrap all I/O functions and provide your own fd management, but
		4397	the cost of calling select (O(n²)) will likely make this unworkable.
		4398
		4399	=back
		4400
		4401	=head2 PORTABILITY REQUIREMENTS
		4402
		4403	In addition to a working ISO-C implementation and of course the
		4404	backend-specific APIs, libev relies on a few additional extensions:
		4405
		4406	=over 4
		4407
		4408	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4409	calling conventions regardless of C<ev_watcher_type *>.
		4410
		4411	Libev assumes not only that all watcher pointers have the same internal
		4412	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4413	assumes that the same (machine) code can be used to call any watcher
		4414	callback: The watcher callbacks have different type signatures, but libev
		4415	calls them using an C<ev_watcher *> internally.
		4416
		4417	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4418
		4419	The type C<sig_atomic_t volatile> (or whatever is defined as
		4420	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4421	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4422	believed to be sufficiently portable.
		4423
		4424	=item C<sigprocmask> must work in a threaded environment
		4425
		4426	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4427	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4428	pthread implementations will either allow C<sigprocmask> in the "main
		4429	thread" or will block signals process-wide, both behaviours would
		4430	be compatible with libev. Interaction between C<sigprocmask> and
		4431	C<pthread_sigmask> could complicate things, however.
		4432
		4433	The most portable way to handle signals is to block signals in all threads
		4434	except the initial one, and run the default loop in the initial thread as
		4435	well.
		4436
		4437	=item C<long> must be large enough for common memory allocation sizes
		4438
		4439	To improve portability and simplify its API, libev uses C<long> internally
		4440	instead of C<size_t> when allocating its data structures. On non-POSIX
		4441	systems (Microsoft...) this might be unexpectedly low, but is still at
		4442	least 31 bits everywhere, which is enough for hundreds of millions of
		4443	watchers.
		4444
		4445	=item C<double> must hold a time value in seconds with enough accuracy
		4446
		4447	The type C<double> is used to represent timestamps. It is required to
		4448	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		4449	enough for at least into the year 4000. This requirement is fulfilled by
		4450	implementations implementing IEEE 754, which is basically all existing
		4451	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4452	2200.
		4453
		4454	=back
		4455
		4456	If you know of other additional requirements drop me a note.
		4457
		4458
		4459	=head1 ALGORITHMIC COMPLEXITIES
		4460
		4461	In this section the complexities of (many of) the algorithms used inside
		4462	libev will be documented. For complexity discussions about backends see
		4463	the documentation for C<ev_default_init>.
		4464
		4465	All of the following are about amortised time: If an array needs to be
		4466	extended, libev needs to realloc and move the whole array, but this
		4467	happens asymptotically rarer with higher number of elements, so O(1) might
		4468	mean that libev does a lengthy realloc operation in rare cases, but on
		4469	average it is much faster and asymptotically approaches constant time.
		4470
		4471	=over 4
		4472
		4473	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4474
		4475	This means that, when you have a watcher that triggers in one hour and
		4476	there are 100 watchers that would trigger before that, then inserting will
		4477	have to skip roughly seven (C<ld 100>) of these watchers.
		4478
		4479	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4480
		4481	That means that changing a timer costs less than removing/adding them,
		4482	as only the relative motion in the event queue has to be paid for.
		4483
		4484	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4485
		4486	These just add the watcher into an array or at the head of a list.
		4487
		4488	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4489
		4490	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4491
		4492	These watchers are stored in lists, so they need to be walked to find the
		4493	correct watcher to remove. The lists are usually short (you don't usually
		4494	have many watchers waiting for the same fd or signal: one is typical, two
		4495	is rare).
		4496
		4497	=item Finding the next timer in each loop iteration: O(1)
		4498
		4499	By virtue of using a binary or 4-heap, the next timer is always found at a
		4500	fixed position in the storage array.
		4501
		4502	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4503
		4504	A change means an I/O watcher gets started or stopped, which requires
		4505	libev to recalculate its status (and possibly tell the kernel, depending
		4506	on backend and whether C<ev_io_set> was used).
		4507
		4508	=item Activating one watcher (putting it into the pending state): O(1)
		4509
		4510	=item Priority handling: O(number_of_priorities)
		4511
		4512	Priorities are implemented by allocating some space for each
		4513	priority. When doing priority-based operations, libev usually has to
		4514	linearly search all the priorities, but starting/stopping and activating
		4515	watchers becomes O(1) with respect to priority handling.
		4516
		4517	=item Sending an ev_async: O(1)
		4518
		4519	=item Processing ev_async_send: O(number_of_async_watchers)
		4520
		4521	=item Processing signals: O(max_signal_number)
		4522
		4523	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4524	calls in the current loop iteration. Checking for async and signal events
		4525	involves iterating over all running async watchers or all signal numbers.
		4526
		4527	=back
		4528
		4529
		4530	=head1 GLOSSARY
		4531
		4532	=over 4
		4533
		4534	=item active
		4535
		4536	A watcher is active as long as it has been started (has been attached to
		4537	an event loop) but not yet stopped (disassociated from the event loop).
		4538
		4539	=item application
		4540
		4541	In this document, an application is whatever is using libev.
		4542
		4543	=item callback
		4544
		4545	The address of a function that is called when some event has been
		4546	detected. Callbacks are being passed the event loop, the watcher that
		4547	received the event, and the actual event bitset.
		4548
		4549	=item callback invocation
		4550
		4551	The act of calling the callback associated with a watcher.
		4552
		4553	=item event
		4554
		4555	A change of state of some external event, such as data now being available
		4556	for reading on a file descriptor, time having passed or simply not having
		4557	any other events happening anymore.
		4558
		4559	In libev, events are represented as single bits (such as C<EV_READ> or
		4560	C<EV_TIMEOUT>).
		4561
		4562	=item event library
		4563
		4564	A software package implementing an event model and loop.
		4565
		4566	=item event loop
		4567
		4568	An entity that handles and processes external events and converts them
		4569	into callback invocations.
		4570
		4571	=item event model
		4572
		4573	The model used to describe how an event loop handles and processes
		4574	watchers and events.
		4575
		4576	=item pending
		4577
		4578	A watcher is pending as soon as the corresponding event has been detected,
		4579	and stops being pending as soon as the watcher will be invoked or its
		4580	pending status is explicitly cleared by the application.
		4581
		4582	A watcher can be pending, but not active. Stopping a watcher also clears
		4583	its pending status.
		4584
		4585	=item real time
		4586
		4587	The physical time that is observed. It is apparently strictly monotonic :)
		4588
		4589	=item wall-clock time
		4590
		4591	The time and date as shown on clocks. Unlike real time, it can actually
		4592	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4593	clock.
		4594
		4595	=item watcher
		4596
		4597	A data structure that describes interest in certain events. Watchers need
		4598	to be started (attached to an event loop) before they can receive events.
		4599
		4600	=item watcher invocation
		4601
		4602	The act of calling the callback associated with a watcher.
		4603
		4604	=back
878		4605
879	=head1 AUTHOR	4606	=head1 AUTHOR
880		4607
881	Marc Lehmann <libev@schmorp.de>.	4608	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
882		4609

Diff Legend

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