ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.8 by root, Mon Nov 12 08:20:02 2007 UTC vs.
Revision 1.260 by root, Sun Jul 19 21:18:03 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
29	events (related to SIGCHLD), and event watchers dealing with the event	103	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	104	with customised rescheduling (C<ev_periodic>), synchronous signals
		105	(C<ev_signal>), process status change events (C<ev_child>), and event
		106	watchers dealing with the event loop mechanism itself (C<ev_idle>,
		107	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
		108	file watchers (C<ev_stat>) and even limited support for fork events
		109	(C<ev_fork>).
		110
		111	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	112	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	113	for example).
33		114
34	=head1 CONVENTIONS	115	=head2 CONVENTIONS
35		116
36	Libev is very configurable. In this manual the default configuration	117	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	118	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	119	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	120	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	121	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 *>)	122	name C<loop> (which is always of type C<ev_loop *>) will not have
42	will not have this argument.	123	this argument.
43		124
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	125	=head2 TIME REPRESENTATION
45		126
46	Libev represents time as a single floating point number, representing the	127	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	128	the (fractional) number of seconds since the (POSIX) epoch (somewhere
48	the beginning of 1970, details are complicated, don't ask). This type is	129	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	130	type is called C<ev_tstamp>, which is what you should use too. It usually
50	to the double type in C.	131	aliases to the C<double> type in C. When you need to do any calculations
		132	on it, you should treat it as some floating point value. Unlike the name
		133	component C<stamp> might indicate, it is also used for time differences
		134	throughout libev.
		135
		136	=head1 ERROR HANDLING
		137
		138	Libev knows three classes of errors: operating system errors, usage errors
		139	and internal errors (bugs).
		140
		141	When libev catches an operating system error it cannot handle (for example
		142	a system call indicating a condition libev cannot fix), it calls the callback
		143	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		144	abort. The default is to print a diagnostic message and to call C<abort
		145	()>.
		146
		147	When libev detects a usage error such as a negative timer interval, then
		148	it will print a diagnostic message and abort (via the C<assert> mechanism,
		149	so C<NDEBUG> will disable this checking): these are programming errors in
		150	the libev caller and need to be fixed there.
		151
		152	Libev also has a few internal error-checking C<assert>ions, and also has
		153	extensive consistency checking code. These do not trigger under normal
		154	circumstances, as they indicate either a bug in libev or worse.
		155
		156
		157	=head1 GLOBAL FUNCTIONS
		158
		159	These functions can be called anytime, even before initialising the
		160	library in any way.
51		161
52	=over 4	162	=over 4
53		163
54	=item ev_tstamp ev_time ()	164	=item ev_tstamp ev_time ()
55		165
56	Returns the current time as libev would use it.	166	Returns the current time as libev would use it. Please note that the
		167	C<ev_now> function is usually faster and also often returns the timestamp
		168	you actually want to know.
		169
		170	=item ev_sleep (ev_tstamp interval)
		171
		172	Sleep for the given interval: The current thread will be blocked until
		173	either it is interrupted or the given time interval has passed. Basically
		174	this is a sub-second-resolution C<sleep ()>.
57		175
58	=item int ev_version_major ()	176	=item int ev_version_major ()
59		177
60	=item int ev_version_minor ()	178	=item int ev_version_minor ()
61		179
62	You can find out the major and minor version numbers of the library	180	You can find out the major and minor ABI version numbers of the library
63	you linked against by calling the functions C<ev_version_major> and	181	you linked against by calling the functions C<ev_version_major> and
64	C<ev_version_minor>. If you want, you can compare against the global	182	C<ev_version_minor>. If you want, you can compare against the global
65	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	183	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
66	version of the library your program was compiled against.	184	version of the library your program was compiled against.
67		185
		186	These version numbers refer to the ABI version of the library, not the
		187	release version.
		188
68	Usually, its a good idea to terminate if the major versions mismatch,	189	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	190	as this indicates an incompatible change. Minor versions are usually
70	compatible to older versions, so a larger minor version alone is usually	191	compatible to older versions, so a larger minor version alone is usually
71	not a problem.	192	not a problem.
72		193
		194	Example: Make sure we haven't accidentally been linked against the wrong
		195	version.
		196
		197	assert (("libev version mismatch",
		198	ev_version_major () == EV_VERSION_MAJOR
		199	&& ev_version_minor () >= EV_VERSION_MINOR));
		200
		201	=item unsigned int ev_supported_backends ()
		202
		203	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		204	value) compiled into this binary of libev (independent of their
		205	availability on the system you are running on). See C<ev_default_loop> for
		206	a description of the set values.
		207
		208	Example: make sure we have the epoll method, because yeah this is cool and
		209	a must have and can we have a torrent of it please!!!11
		210
		211	assert (("sorry, no epoll, no sex",
		212	ev_supported_backends () & EVBACKEND_EPOLL));
		213
		214	=item unsigned int ev_recommended_backends ()
		215
		216	Return the set of all backends compiled into this binary of libev and also
		217	recommended for this platform. This set is often smaller than the one
		218	returned by C<ev_supported_backends>, as for example kqueue is broken on
		219	most BSDs and will not be auto-detected unless you explicitly request it
		220	(assuming you know what you are doing). This is the set of backends that
		221	libev will probe for if you specify no backends explicitly.
		222
		223	=item unsigned int ev_embeddable_backends ()
		224
		225	Returns the set of backends that are embeddable in other event loops. This
		226	is the theoretical, all-platform, value. To find which backends
		227	might be supported on the current system, you would need to look at
		228	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		229	recommended ones.
		230
		231	See the description of C<ev_embed> watchers for more info.
		232
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	233	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
74		234
75	Sets the allocation function to use (the prototype is similar to the	235	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	236	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
77	and free memory (no surprises here). If it returns zero when memory	237	used to allocate and free memory (no surprises here). If it returns zero
78	needs to be allocated, the library might abort or take some potentially	238	when memory needs to be allocated (C<size != 0>), the library might abort
79	destructive action. The default is your system realloc function.	239	or take some potentially destructive action.
		240
		241	Since some systems (at least OpenBSD and Darwin) fail to implement
		242	correct C<realloc> semantics, libev will use a wrapper around the system
		243	C<realloc> and C<free> functions by default.
80		244
81	You could override this function in high-availability programs to, say,	245	You could override this function in high-availability programs to, say,
82	free some memory if it cannot allocate memory, to use a special allocator,	246	free some memory if it cannot allocate memory, to use a special allocator,
83	or even to sleep a while and retry until some memory is available.	247	or even to sleep a while and retry until some memory is available.
84		248
		249	Example: Replace the libev allocator with one that waits a bit and then
		250	retries (example requires a standards-compliant C<realloc>).
		251
		252	static void *
		253	persistent_realloc (void *ptr, size_t size)
		254	{
		255	for (;;)
		256	{
		257	void *newptr = realloc (ptr, size);
		258
		259	if (newptr)
		260	return newptr;
		261
		262	sleep (60);
		263	}
		264	}
		265
		266	...
		267	ev_set_allocator (persistent_realloc);
		268
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	269	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
86		270
87	Set the callback function to call on a retryable syscall error (such	271	Set the callback function to call on a retryable system call error (such
88	as failed select, poll, epoll_wait). The message is a printable string	272	as failed select, poll, epoll_wait). The message is a printable string
89	indicating the system call or subsystem causing the problem. If this	273	indicating the system call or subsystem causing the problem. If this
90	callback is set, then libev will expect it to remedy the sitution, no	274	callback is set, then libev will expect it to remedy the situation, no
91	matter what, when it returns. That is, libev will generally retry the	275	matter what, when it returns. That is, libev will generally retry the
92	requested operation, or, if the condition doesn't go away, do bad stuff	276	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	277	(such as abort).
94		278
		279	Example: This is basically the same thing that libev does internally, too.
		280
		281	static void
		282	fatal_error (const char *msg)
		283	{
		284	perror (msg);
		285	abort ();
		286	}
		287
		288	...
		289	ev_set_syserr_cb (fatal_error);
		290
95	=back	291	=back
96		292
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	293	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
98		294
99	An event loop is described by a C<struct ev_loop *>. The library knows two	295	An event loop is described by a C<struct ev_loop *> (the C<struct>
100	types of such loops, the I<default> loop, which supports signals and child	296	is I<not> optional in this case, as there is also an C<ev_loop>
101	events, and dynamically created loops which do not.	297	I<function>).
102		298
103	If you use threads, a common model is to run the default event loop	299	The library knows two types of such loops, the I<default> loop, which
104	in your main thread (or in a separate thrad) and for each thread you	300	supports signals and child events, and dynamically created loops which do
105	create, you also create another event loop. Libev itself does no locking	301	not.
106	whatsoever, so if you mix calls to the same event loop in different
107	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because its hideous and inefficient).
109		302
110	=over 4	303	=over 4
111		304
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	305	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		306
114	This will initialise the default event loop if it hasn't been initialised	307	This will initialise the default event loop if it hasn't been initialised
115	yet and return it. If the default loop could not be initialised, returns	308	yet and return it. If the default loop could not be initialised, returns
116	false. If it already was initialised it simply returns it (and ignores the	309	false. If it already was initialised it simply returns it (and ignores the
117	flags).	310	flags. If that is troubling you, check C<ev_backend ()> afterwards).
118		311
119	If you don't know what event loop to use, use the one returned from this	312	If you don't know what event loop to use, use the one returned from this
120	function.	313	function.
121		314
		315	Note that this function is I<not> thread-safe, so if you want to use it
		316	from multiple threads, you have to lock (note also that this is unlikely,
		317	as loops cannot be shared easily between threads anyway).
		318
		319	The default loop is the only loop that can handle C<ev_signal> and
		320	C<ev_child> watchers, and to do this, it always registers a handler
		321	for C<SIGCHLD>. If this is a problem for your application you can either
		322	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		323	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		324	C<ev_default_init>.
		325
122	The flags argument can be used to specify special behaviour or specific	326	The flags argument can be used to specify special behaviour or specific
123	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	327	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
124		328
125	It supports the following flags:	329	The following flags are supported:
126		330
127	=over 4	331	=over 4
128		332
129	=item EVFLAG_AUTO	333	=item C<EVFLAG_AUTO>
130		334
131	The default flags value. Use this if you have no clue (its the right	335	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	336	thing, believe me).
133		337
134	=item EVFLAG_NOENV	338	=item C<EVFLAG_NOENV>
135		339
136	If this flag bit is ored into the flag value (or the program runs setuid	340	If this flag bit is or'ed into the flag value (or the program runs setuid
137	or setgid) then libev will I<not> look at the environment variable	341	or setgid) then libev will I<not> look at the environment variable
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	342	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	343	override the flags completely if it is found in the environment. This is
140	useful to try out specific backends to test their performance, or to work	344	useful to try out specific backends to test their performance, or to work
141	around bugs.	345	around bugs.
142		346
		347	=item C<EVFLAG_FORKCHECK>
		348
		349	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
		350	a fork, you can also make libev check for a fork in each iteration by
		351	enabling this flag.
		352
		353	This works by calling C<getpid ()> on every iteration of the loop,
		354	and thus this might slow down your event loop if you do a lot of loop
		355	iterations and little real work, but is usually not noticeable (on my
		356	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		357	without a system call and thus I<very> fast, but my GNU/Linux system also has
		358	C<pthread_atfork> which is even faster).
		359
		360	The big advantage of this flag is that you can forget about fork (and
		361	forget about forgetting to tell libev about forking) when you use this
		362	flag.
		363
		364	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		365	environment variable.
		366
		367	=item C<EVFLAG_NOINOTIFY>
		368
		369	When this flag is specified, then libev will not attempt to use the
		370	I<inotify> API for it's C<ev_stat> watchers. Apart from debugging and
		371	testing, this flag can be useful to conserve inotify file descriptors, as
		372	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		373
		374	=item C<EVFLAG_NOSIGNALFD>
		375
		376	When this flag is specified, then libev will not attempt to use the
		377	I<signalfd> API for it's C<ev_signal> (and C<ev_child>) watchers. This is
		378	probably only useful to work around any bugs in libev. Consequently, this
		379	flag might go away once the signalfd functionality is considered stable,
		380	so it's useful mostly in environment variables and not in program code.
		381
143	=item EVMETHOD_SELECT portable select backend	382	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		383
145	=item EVMETHOD_POLL poll backend (everywhere except windows)	384	This is your standard select(2) backend. Not I<completely> standard, as
		385	libev tries to roll its own fd_set with no limits on the number of fds,
		386	but if that fails, expect a fairly low limit on the number of fds when
		387	using this backend. It doesn't scale too well (O(highest_fd)), but its
		388	usually the fastest backend for a low number of (low-numbered :) fds.
146		389
147	=item EVMETHOD_EPOLL linux only	390	To get good performance out of this backend you need a high amount of
		391	parallelism (most of the file descriptors should be busy). If you are
		392	writing a server, you should C<accept ()> in a loop to accept as many
		393	connections as possible during one iteration. You might also want to have
		394	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		395	readiness notifications you get per iteration.
148		396
149	=item EVMETHOD_KQUEUE some bsds only	397	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		398	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		399	C<exceptfds> set on that platform).
150		400
151	=item EVMETHOD_DEVPOLL solaris 8 only	401	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
152		402
153	=item EVMETHOD_PORT solaris 10 only	403	And this is your standard poll(2) backend. It's more complicated
		404	than select, but handles sparse fds better and has no artificial
		405	limit on the number of fds you can use (except it will slow down
		406	considerably with a lot of inactive fds). It scales similarly to select,
		407	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		408	performance tips.
154		409
155	If one or more of these are ored into the flags value, then only these	410	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
156	backends will be tried (in the reverse order as given here). If one are	411	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157	specified, any backend will do.	412
		413	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		414
		415	For few fds, this backend is a bit little slower than poll and select,
		416	but it scales phenomenally better. While poll and select usually scale
		417	like O(total_fds) where n is the total number of fds (or the highest fd),
		418	epoll scales either O(1) or O(active_fds).
		419
		420	The epoll mechanism deserves honorable mention as the most misdesigned
		421	of the more advanced event mechanisms: mere annoyances include silently
		422	dropping file descriptors, requiring a system call per change per file
		423	descriptor (and unnecessary guessing of parameters), problems with dup and
		424	so on. The biggest issue is fork races, however - if a program forks then
		425	I<both> parent and child process have to recreate the epoll set, which can
		426	take considerable time (one syscall per file descriptor) and is of course
		427	hard to detect.
		428
		429	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		430	of course I<doesn't>, and epoll just loves to report events for totally
		431	I<different> file descriptors (even already closed ones, so one cannot
		432	even remove them from the set) than registered in the set (especially
		433	on SMP systems). Libev tries to counter these spurious notifications by
		434	employing an additional generation counter and comparing that against the
		435	events to filter out spurious ones, recreating the set when required.
		436
		437	While stopping, setting and starting an I/O watcher in the same iteration
		438	will result in some caching, there is still a system call per such
		439	incident (because the same I<file descriptor> could point to a different
		440	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		441	file descriptors might not work very well if you register events for both
		442	file descriptors.
		443
		444	Best performance from this backend is achieved by not unregistering all
		445	watchers for a file descriptor until it has been closed, if possible,
		446	i.e. keep at least one watcher active per fd at all times. Stopping and
		447	starting a watcher (without re-setting it) also usually doesn't cause
		448	extra overhead. A fork can both result in spurious notifications as well
		449	as in libev having to destroy and recreate the epoll object, which can
		450	take considerable time and thus should be avoided.
		451
		452	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		453	faster than epoll for maybe up to a hundred file descriptors, depending on
		454	the usage. So sad.
		455
		456	While nominally embeddable in other event loops, this feature is broken in
		457	all kernel versions tested so far.
		458
		459	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		460	C<EVBACKEND_POLL>.
		461
		462	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		463
		464	Kqueue deserves special mention, as at the time of this writing, it
		465	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		466	with anything but sockets and pipes, except on Darwin, where of course
		467	it's completely useless). Unlike epoll, however, whose brokenness
		468	is by design, these kqueue bugs can (and eventually will) be fixed
		469	without API changes to existing programs. For this reason it's not being
		470	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		471	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		472	system like NetBSD.
		473
		474	You still can embed kqueue into a normal poll or select backend and use it
		475	only for sockets (after having made sure that sockets work with kqueue on
		476	the target platform). See C<ev_embed> watchers for more info.
		477
		478	It scales in the same way as the epoll backend, but the interface to the
		479	kernel is more efficient (which says nothing about its actual speed, of
		480	course). While stopping, setting and starting an I/O watcher does never
		481	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		482	two event changes per incident. Support for C<fork ()> is very bad (but
		483	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		484	cases
		485
		486	This backend usually performs well under most conditions.
		487
		488	While nominally embeddable in other event loops, this doesn't work
		489	everywhere, so you might need to test for this. And since it is broken
		490	almost everywhere, you should only use it when you have a lot of sockets
		491	(for which it usually works), by embedding it into another event loop
		492	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		493	also broken on OS X)) and, did I mention it, using it only for sockets.
		494
		495	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		496	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		497	C<NOTE_EOF>.
		498
		499	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		500
		501	This is not implemented yet (and might never be, unless you send me an
		502	implementation). According to reports, C</dev/poll> only supports sockets
		503	and is not embeddable, which would limit the usefulness of this backend
		504	immensely.
		505
		506	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		507
		508	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		509	it's really slow, but it still scales very well (O(active_fds)).
		510
		511	Please note that Solaris event ports can deliver a lot of spurious
		512	notifications, so you need to use non-blocking I/O or other means to avoid
		513	blocking when no data (or space) is available.
		514
		515	While this backend scales well, it requires one system call per active
		516	file descriptor per loop iteration. For small and medium numbers of file
		517	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		518	might perform better.
		519
		520	On the positive side, with the exception of the spurious readiness
		521	notifications, this backend actually performed fully to specification
		522	in all tests and is fully embeddable, which is a rare feat among the
		523	OS-specific backends (I vastly prefer correctness over speed hacks).
		524
		525	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		526	C<EVBACKEND_POLL>.
		527
		528	=item C<EVBACKEND_ALL>
		529
		530	Try all backends (even potentially broken ones that wouldn't be tried
		531	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		532	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		533
		534	It is definitely not recommended to use this flag.
158		535
159	=back	536	=back
		537
		538	If one or more of the backend flags are or'ed into the flags value,
		539	then only these backends will be tried (in the reverse order as listed
		540	here). If none are specified, all backends in C<ev_recommended_backends
		541	()> will be tried.
		542
		543	Example: This is the most typical usage.
		544
		545	if (!ev_default_loop (0))
		546	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		547
		548	Example: Restrict libev to the select and poll backends, and do not allow
		549	environment settings to be taken into account:
		550
		551	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		552
		553	Example: Use whatever libev has to offer, but make sure that kqueue is
		554	used if available (warning, breaks stuff, best use only with your own
		555	private event loop and only if you know the OS supports your types of
		556	fds):
		557
		558	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
160		559
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	560	=item struct ev_loop *ev_loop_new (unsigned int flags)
162		561
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	562	Similar to C<ev_default_loop>, but always creates a new event loop that is
164	always distinct from the default loop. Unlike the default loop, it cannot	563	always distinct from the default loop. Unlike the default loop, it cannot
165	handle signal and child watchers, and attempts to do so will be greeted by	564	handle signal and child watchers, and attempts to do so will be greeted by
166	undefined behaviour (or a failed assertion if assertions are enabled).	565	undefined behaviour (or a failed assertion if assertions are enabled).
167		566
		567	Note that this function I<is> thread-safe, and the recommended way to use
		568	libev with threads is indeed to create one loop per thread, and using the
		569	default loop in the "main" or "initial" thread.
		570
		571	Example: Try to create a event loop that uses epoll and nothing else.
		572
		573	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		574	if (!epoller)
		575	fatal ("no epoll found here, maybe it hides under your chair");
		576
168	=item ev_default_destroy ()	577	=item ev_default_destroy ()
169		578
170	Destroys the default loop again (frees all memory and kernel state	579	Destroys the default loop again (frees all memory and kernel state
171	etc.). This stops all registered event watchers (by not touching them in	580	etc.). None of the active event watchers will be stopped in the normal
172	any way whatsoever, although you cnanot rely on this :).	581	sense, so e.g. C<ev_is_active> might still return true. It is your
		582	responsibility to either stop all watchers cleanly yourself I<before>
		583	calling this function, or cope with the fact afterwards (which is usually
		584	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		585	for example).
		586
		587	Note that certain global state, such as signal state (and installed signal
		588	handlers), will not be freed by this function, and related watchers (such
		589	as signal and child watchers) would need to be stopped manually.
		590
		591	In general it is not advisable to call this function except in the
		592	rare occasion where you really need to free e.g. the signal handling
		593	pipe fds. If you need dynamically allocated loops it is better to use
		594	C<ev_loop_new> and C<ev_loop_destroy>).
173		595
174	=item ev_loop_destroy (loop)	596	=item ev_loop_destroy (loop)
175		597
176	Like C<ev_default_destroy>, but destroys an event loop created by an	598	Like C<ev_default_destroy>, but destroys an event loop created by an
177	earlier call to C<ev_loop_new>.	599	earlier call to C<ev_loop_new>.
178		600
179	=item ev_default_fork ()	601	=item ev_default_fork ()
180		602
		603	This function sets a flag that causes subsequent C<ev_loop> iterations
181	This function reinitialises the kernel state for backends that have	604	to reinitialise the kernel state for backends that have one. Despite the
182	one. Despite the name, you can call it anytime, but it makes most sense	605	name, you can call it anytime, but it makes most sense after forking, in
183	after forking, in either the parent or child process (or both, but that	606	the child process (or both child and parent, but that again makes little
184	again makes little sense).	607	sense). You I<must> call it in the child before using any of the libev
		608	functions, and it will only take effect at the next C<ev_loop> iteration.
185		609
186	You I<must> call this function after forking if and only if you want to	610	On the other hand, you only need to call this function in the child
187	use the event library in both processes. If you just fork+exec, you don't	611	process if and only if you want to use the event library in the child. If
188	have to call it.	612	you just fork+exec, you don't have to call it at all.
189		613
190	The function itself is quite fast and its usually not a problem to call	614	The function itself is quite fast and it's usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in	615	it just in case after a fork. To make this easy, the function will fit in
192	quite nicely into a call to C<pthread_atfork>:	616	quite nicely into a call to C<pthread_atfork>:
193		617
194	pthread_atfork (0, 0, ev_default_fork);	618	pthread_atfork (0, 0, ev_default_fork);
195		619
196	=item ev_loop_fork (loop)	620	=item ev_loop_fork (loop)
197		621
198	Like C<ev_default_fork>, but acts on an event loop created by	622	Like C<ev_default_fork>, but acts on an event loop created by
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	623	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
200	after fork, and how you do this is entirely your own problem.	624	after fork that you want to re-use in the child, and how you do this is
		625	entirely your own problem.
201		626
		627	=item int ev_is_default_loop (loop)
		628
		629	Returns true when the given loop is, in fact, the default loop, and false
		630	otherwise.
		631
		632	=item unsigned int ev_loop_count (loop)
		633
		634	Returns the count of loop iterations for the loop, which is identical to
		635	the number of times libev did poll for new events. It starts at C<0> and
		636	happily wraps around with enough iterations.
		637
		638	This value can sometimes be useful as a generation counter of sorts (it
		639	"ticks" the number of loop iterations), as it roughly corresponds with
		640	C<ev_prepare> and C<ev_check> calls.
		641
		642	=item unsigned int ev_loop_depth (loop)
		643
		644	Returns the number of times C<ev_loop> was entered minus the number of
		645	times C<ev_loop> was exited, in other words, the recursion depth.
		646
		647	Outside C<ev_loop>, this number is zero. In a callback, this number is
		648	C<1>, unless C<ev_loop> was invoked recursively (or from another thread),
		649	in which case it is higher.
		650
		651	Leaving C<ev_loop> abnormally (setjmp/longjmp, cancelling the thread
		652	etc.), doesn't count as exit.
		653
202	=item unsigned int ev_method (loop)	654	=item unsigned int ev_backend (loop)
203		655
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	656	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	657	use.
206		658
207	=item ev_tstamp = ev_now (loop)	659	=item ev_tstamp ev_now (loop)
208		660
209	Returns the current "event loop time", which is the time the event loop	661	Returns the current "event loop time", which is the time the event loop
210	got events and started processing them. This timestamp does not change	662	received events and started processing them. This timestamp does not
211	as long as callbacks are being processed, and this is also the base time	663	change as long as callbacks are being processed, and this is also the base
212	used for relative timers. You can treat it as the timestamp of the event	664	time used for relative timers. You can treat it as the timestamp of the
213	occuring (or more correctly, the mainloop finding out about it).	665	event occurring (or more correctly, libev finding out about it).
		666
		667	=item ev_now_update (loop)
		668
		669	Establishes the current time by querying the kernel, updating the time
		670	returned by C<ev_now ()> in the progress. This is a costly operation and
		671	is usually done automatically within C<ev_loop ()>.
		672
		673	This function is rarely useful, but when some event callback runs for a
		674	very long time without entering the event loop, updating libev's idea of
		675	the current time is a good idea.
		676
		677	See also L<The special problem of time updates> in the C<ev_timer> section.
		678
		679	=item ev_suspend (loop)
		680
		681	=item ev_resume (loop)
		682
		683	These two functions suspend and resume a loop, for use when the loop is
		684	not used for a while and timeouts should not be processed.
		685
		686	A typical use case would be an interactive program such as a game: When
		687	the user presses C<^Z> to suspend the game and resumes it an hour later it
		688	would be best to handle timeouts as if no time had actually passed while
		689	the program was suspended. This can be achieved by calling C<ev_suspend>
		690	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		691	C<ev_resume> directly afterwards to resume timer processing.
		692
		693	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		694	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		695	will be rescheduled (that is, they will lose any events that would have
		696	occured while suspended).
		697
		698	After calling C<ev_suspend> you B<must not> call I<any> function on the
		699	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		700	without a previous call to C<ev_suspend>.
		701
		702	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		703	event loop time (see C<ev_now_update>).
214		704
215	=item ev_loop (loop, int flags)	705	=item ev_loop (loop, int flags)
216		706
217	Finally, this is it, the event handler. This function usually is called	707	Finally, this is it, the event handler. This function usually is called
218	after you initialised all your watchers and you want to start handling	708	after you initialised all your watchers and you want to start handling
219	events.	709	events.
220		710
221	If the flags argument is specified as 0, it will not return until either	711	If the flags argument is specified as C<0>, it will not return until
222	no event watchers are active anymore or C<ev_unloop> was called.	712	either no event watchers are active anymore or C<ev_unloop> was called.
		713
		714	Please note that an explicit C<ev_unloop> is usually better than
		715	relying on all watchers to be stopped when deciding when a program has
		716	finished (especially in interactive programs), but having a program
		717	that automatically loops as long as it has to and no longer by virtue
		718	of relying on its watchers stopping correctly, that is truly a thing of
		719	beauty.
223		720
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	721	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
225	those events and any outstanding ones, but will not block your process in	722	those events and any already outstanding ones, but will not block your
226	case there are no events.	723	process in case there are no events and will return after one iteration of
		724	the loop.
227		725
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	726	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
229	neccessary) and will handle those and any outstanding ones. It will block	727	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives.	728	will block your process until at least one new event arrives (which could
		729	be an event internal to libev itself, so there is no guarantee that a
		730	user-registered callback will be called), and will return after one
		731	iteration of the loop.
231		732
232	This flags value could be used to implement alternative looping	733	This is useful if you are waiting for some external event in conjunction
233	constructs, but the C<prepare> and C<check> watchers provide a better and	734	with something not expressible using other libev watchers (i.e. "roll your
234	more generic mechanism.	735	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		736	usually a better approach for this kind of thing.
		737
		738	Here are the gory details of what C<ev_loop> does:
		739
		740	- Before the first iteration, call any pending watchers.
		741	* If EVFLAG_FORKCHECK was used, check for a fork.
		742	- If a fork was detected (by any means), queue and call all fork watchers.
		743	- Queue and call all prepare watchers.
		744	- If we have been forked, detach and recreate the kernel state
		745	as to not disturb the other process.
		746	- Update the kernel state with all outstanding changes.
		747	- Update the "event loop time" (ev_now ()).
		748	- Calculate for how long to sleep or block, if at all
		749	(active idle watchers, EVLOOP_NONBLOCK or not having
		750	any active watchers at all will result in not sleeping).
		751	- Sleep if the I/O and timer collect interval say so.
		752	- Block the process, waiting for any events.
		753	- Queue all outstanding I/O (fd) events.
		754	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		755	- Queue all expired timers.
		756	- Queue all expired periodics.
		757	- Unless any events are pending now, queue all idle watchers.
		758	- Queue all check watchers.
		759	- Call all queued watchers in reverse order (i.e. check watchers first).
		760	Signals and child watchers are implemented as I/O watchers, and will
		761	be handled here by queueing them when their watcher gets executed.
		762	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		763	were used, or there are no active watchers, return, otherwise
		764	continue with step *.
		765
		766	Example: Queue some jobs and then loop until no events are outstanding
		767	anymore.
		768
		769	... queue jobs here, make sure they register event watchers as long
		770	... as they still have work to do (even an idle watcher will do..)
		771	ev_loop (my_loop, 0);
		772	... jobs done or somebody called unloop. yeah!
235		773
236	=item ev_unloop (loop, how)	774	=item ev_unloop (loop, how)
237		775
238	Can be used to make a call to C<ev_loop> return early. The C<how> argument	776	Can be used to make a call to C<ev_loop> return early (but only after it
		777	has processed all outstanding events). The C<how> argument must be either
239	must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop>	778	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
240	call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop>	779	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
241	calls return.	780
		781	This "unloop state" will be cleared when entering C<ev_loop> again.
		782
		783	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
242		784
243	=item ev_ref (loop)	785	=item ev_ref (loop)
244		786
245	=item ev_unref (loop)	787	=item ev_unref (loop)
246		788
247	Ref/unref can be used to add or remove a refcount on the event loop: Every	789	Ref/unref can be used to add or remove a reference count on the event
248	watcher keeps one reference. If you have a long-runing watcher you never	790	loop: Every watcher keeps one reference, and as long as the reference
249	unregister that should not keep ev_loop from running, ev_unref() after	791	count is nonzero, C<ev_loop> will not return on its own.
250	starting, and ev_ref() before stopping it. Libev itself uses this for	792
251	example for its internal signal pipe: It is not visible to you as a user	793	If you have a watcher you never unregister that should not keep C<ev_loop>
252	and should not keep C<ev_loop> from exiting if the work is done. It is	794	from returning, call ev_unref() after starting, and ev_ref() before
		795	stopping it.
		796
		797	As an example, libev itself uses this for its internal signal pipe: It
		798	is not visible to the libev user and should not keep C<ev_loop> from
		799	exiting if no event watchers registered by it are active. It is also an
253	also an excellent way to do this for generic recurring timers or from	800	excellent way to do this for generic recurring timers or from within
254	within third-party libraries. Just remember to unref after start and ref	801	third-party libraries. Just remember to I<unref after start> and I<ref
255	before stop.	802	before stop> (but only if the watcher wasn't active before, or was active
		803	before, respectively. Note also that libev might stop watchers itself
		804	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		805	in the callback).
		806
		807	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		808	running when nothing else is active.
		809
		810	ev_signal exitsig;
		811	ev_signal_init (&exitsig, sig_cb, SIGINT);
		812	ev_signal_start (loop, &exitsig);
		813	evf_unref (loop);
		814
		815	Example: For some weird reason, unregister the above signal handler again.
		816
		817	ev_ref (loop);
		818	ev_signal_stop (loop, &exitsig);
		819
		820	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		821
		822	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		823
		824	These advanced functions influence the time that libev will spend waiting
		825	for events. Both time intervals are by default C<0>, meaning that libev
		826	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		827	latency.
		828
		829	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		830	allows libev to delay invocation of I/O and timer/periodic callbacks
		831	to increase efficiency of loop iterations (or to increase power-saving
		832	opportunities).
		833
		834	The idea is that sometimes your program runs just fast enough to handle
		835	one (or very few) event(s) per loop iteration. While this makes the
		836	program responsive, it also wastes a lot of CPU time to poll for new
		837	events, especially with backends like C<select ()> which have a high
		838	overhead for the actual polling but can deliver many events at once.
		839
		840	By setting a higher I<io collect interval> you allow libev to spend more
		841	time collecting I/O events, so you can handle more events per iteration,
		842	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		843	C<ev_timer>) will be not affected. Setting this to a non-null value will
		844	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		845	sleep time ensures that libev will not poll for I/O events more often then
		846	once per this interval, on average.
		847
		848	Likewise, by setting a higher I<timeout collect interval> you allow libev
		849	to spend more time collecting timeouts, at the expense of increased
		850	latency/jitter/inexactness (the watcher callback will be called
		851	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		852	value will not introduce any overhead in libev.
		853
		854	Many (busy) programs can usually benefit by setting the I/O collect
		855	interval to a value near C<0.1> or so, which is often enough for
		856	interactive servers (of course not for games), likewise for timeouts. It
		857	usually doesn't make much sense to set it to a lower value than C<0.01>,
		858	as this approaches the timing granularity of most systems. Note that if
		859	you do transactions with the outside world and you can't increase the
		860	parallelity, then this setting will limit your transaction rate (if you
		861	need to poll once per transaction and the I/O collect interval is 0.01,
		862	then you can't do more than 100 transations per second).
		863
		864	Setting the I<timeout collect interval> can improve the opportunity for
		865	saving power, as the program will "bundle" timer callback invocations that
		866	are "near" in time together, by delaying some, thus reducing the number of
		867	times the process sleeps and wakes up again. Another useful technique to
		868	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		869	they fire on, say, one-second boundaries only.
		870
		871	Example: we only need 0.1s timeout granularity, and we wish not to poll
		872	more often than 100 times per second:
		873
		874	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		875	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		876
		877	=item ev_invoke_pending (loop)
		878
		879	This call will simply invoke all pending watchers while resetting their
		880	pending state. Normally, C<ev_loop> does this automatically when required,
		881	but when overriding the invoke callback this call comes handy.
		882
		883	=item int ev_pending_count (loop)
		884
		885	Returns the number of pending watchers - zero indicates that no watchers
		886	are pending.
		887
		888	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		889
		890	This overrides the invoke pending functionality of the loop: Instead of
		891	invoking all pending watchers when there are any, C<ev_loop> will call
		892	this callback instead. This is useful, for example, when you want to
		893	invoke the actual watchers inside another context (another thread etc.).
		894
		895	If you want to reset the callback, use C<ev_invoke_pending> as new
		896	callback.
		897
		898	=item ev_set_loop_release_cb (loop, void (*release)(EV_P), void (*acquire)(EV_P))
		899
		900	Sometimes you want to share the same loop between multiple threads. This
		901	can be done relatively simply by putting mutex_lock/unlock calls around
		902	each call to a libev function.
		903
		904	However, C<ev_loop> can run an indefinite time, so it is not feasible to
		905	wait for it to return. One way around this is to wake up the loop via
		906	C<ev_unloop> and C<av_async_send>, another way is to set these I<release>
		907	and I<acquire> callbacks on the loop.
		908
		909	When set, then C<release> will be called just before the thread is
		910	suspended waiting for new events, and C<acquire> is called just
		911	afterwards.
		912
		913	Ideally, C<release> will just call your mutex_unlock function, and
		914	C<acquire> will just call the mutex_lock function again.
		915
		916	While event loop modifications are allowed between invocations of
		917	C<release> and C<acquire> (that's their only purpose after all), no
		918	modifications done will affect the event loop, i.e. adding watchers will
		919	have no effect on the set of file descriptors being watched, or the time
		920	waited. USe an C<ev_async> watcher to wake up C<ev_loop> when you want it
		921	to take note of any changes you made.
		922
		923	In theory, threads executing C<ev_loop> will be async-cancel safe between
		924	invocations of C<release> and C<acquire>.
		925
		926	See also the locking example in the C<THREADS> section later in this
		927	document.
		928
		929	=item ev_set_userdata (loop, void *data)
		930
		931	=item ev_userdata (loop)
		932
		933	Set and retrieve a single C<void *> associated with a loop. When
		934	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		935	C<0.>
		936
		937	These two functions can be used to associate arbitrary data with a loop,
		938	and are intended solely for the C<invoke_pending_cb>, C<release> and
		939	C<acquire> callbacks described above, but of course can be (ab-)used for
		940	any other purpose as well.
		941
		942	=item ev_loop_verify (loop)
		943
		944	This function only does something when C<EV_VERIFY> support has been
		945	compiled in, which is the default for non-minimal builds. It tries to go
		946	through all internal structures and checks them for validity. If anything
		947	is found to be inconsistent, it will print an error message to standard
		948	error and call C<abort ()>.
		949
		950	This can be used to catch bugs inside libev itself: under normal
		951	circumstances, this function will never abort as of course libev keeps its
		952	data structures consistent.
256		953
257	=back	954	=back
258		955
		956
259	=head1 ANATOMY OF A WATCHER	957	=head1 ANATOMY OF A WATCHER
		958
		959	In the following description, uppercase C<TYPE> in names stands for the
		960	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		961	watchers and C<ev_io_start> for I/O watchers.
260		962
261	A watcher is a structure that you create and register to record your	963	A watcher is a structure that you create and register to record your
262	interest in some event. For instance, if you want to wait for STDIN to	964	interest in some event. For instance, if you want to wait for STDIN to
263	become readable, you would create an ev_io watcher for that:	965	become readable, you would create an C<ev_io> watcher for that:
264		966
265	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	967	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
266	{	968	{
267	ev_io_stop (w);	969	ev_io_stop (w);
268	ev_unloop (loop, EVUNLOOP_ALL);	970	ev_unloop (loop, EVUNLOOP_ALL);
269	}	971	}
270		972
271	struct ev_loop *loop = ev_default_loop (0);	973	struct ev_loop *loop = ev_default_loop (0);
		974
272	struct ev_io stdin_watcher;	975	ev_io stdin_watcher;
		976
273	ev_init (&stdin_watcher, my_cb);	977	ev_init (&stdin_watcher, my_cb);
274	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	978	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
275	ev_io_start (loop, &stdin_watcher);	979	ev_io_start (loop, &stdin_watcher);
		980
276	ev_loop (loop, 0);	981	ev_loop (loop, 0);
277		982
278	As you can see, you are responsible for allocating the memory for your	983	As you can see, you are responsible for allocating the memory for your
279	watcher structures (and it is usually a bad idea to do this on the stack,	984	watcher structures (and it is I<usually> a bad idea to do this on the
280	although this can sometimes be quite valid).	985	stack).
		986
		987	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		988	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
281		989
282	Each watcher structure must be initialised by a call to C<ev_init	990	Each watcher structure must be initialised by a call to C<ev_init
283	(watcher *, callback)>, which expects a callback to be provided. This	991	(watcher *, callback)>, which expects a callback to be provided. This
284	callback gets invoked each time the event occurs (or, in the case of io	992	callback gets invoked each time the event occurs (or, in the case of I/O
285	watchers, each time the event loop detects that the file descriptor given	993	watchers, each time the event loop detects that the file descriptor given
286	is readable and/or writable).	994	is readable and/or writable).
287		995
288	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	996	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
289	with arguments specific to this watcher type. There is also a macro	997	macro to configure it, with arguments specific to the watcher type. There
290	to combine initialisation and setting in one call: C<< ev_<type>_init	998	is also a macro to combine initialisation and setting in one call: C<<
291	(watcher *, callback, ...) >>.	999	ev_TYPE_init (watcher *, callback, ...) >>.
292		1000
293	To make the watcher actually watch out for events, you have to start it	1001	To make the watcher actually watch out for events, you have to start it
294	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1002	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
295	*) >>), and you can stop watching for events at any time by calling the	1003	*) >>), and you can stop watching for events at any time by calling the
296	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1004	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
297		1005
298	As long as your watcher is active (has been started but not stopped) you	1006	As long as your watcher is active (has been started but not stopped) you
299	must not touch the values stored in it. Most specifically you must never	1007	must not touch the values stored in it. Most specifically you must never
300	reinitialise it or call its set method.	1008	reinitialise it or call its C<ev_TYPE_set> macro.
301
302	You cna check whether an event is active by calling the C<ev_is_active
303	(watcher *)> macro. To see whether an event is outstanding (but the
304	callback for it has not been called yet) you cna use the C<ev_is_pending
305	(watcher *)> macro.
306		1009
307	Each and every callback receives the event loop pointer as first, the	1010	Each and every callback receives the event loop pointer as first, the
308	registered watcher structure as second, and a bitset of received events as	1011	registered watcher structure as second, and a bitset of received events as
309	third argument.	1012	third argument.
310		1013
311	The rceeived events usually include a single bit per event type received	1014	The received events usually include a single bit per event type received
312	(you can receive multiple events at the same time). The possible bit masks	1015	(you can receive multiple events at the same time). The possible bit masks
313	are:	1016	are:
314		1017
315	=over 4	1018	=over 4
316		1019
317	=item EV_READ	1020	=item C<EV_READ>
318		1021
319	=item EV_WRITE	1022	=item C<EV_WRITE>
320		1023
321	The file descriptor in the ev_io watcher has become readable and/or	1024	The file descriptor in the C<ev_io> watcher has become readable and/or
322	writable.	1025	writable.
323		1026
324	=item EV_TIMEOUT	1027	=item C<EV_TIMEOUT>
325		1028
326	The ev_timer watcher has timed out.	1029	The C<ev_timer> watcher has timed out.
327		1030
328	=item EV_PERIODIC	1031	=item C<EV_PERIODIC>
329		1032
330	The ev_periodic watcher has timed out.	1033	The C<ev_periodic> watcher has timed out.
331		1034
332	=item EV_SIGNAL	1035	=item C<EV_SIGNAL>
333		1036
334	The signal specified in the ev_signal watcher has been received by a thread.	1037	The signal specified in the C<ev_signal> watcher has been received by a thread.
335		1038
336	=item EV_CHILD	1039	=item C<EV_CHILD>
337		1040
338	The pid specified in the ev_child watcher has received a status change.	1041	The pid specified in the C<ev_child> watcher has received a status change.
339		1042
		1043	=item C<EV_STAT>
		1044
		1045	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1046
340	=item EV_IDLE	1047	=item C<EV_IDLE>
341		1048
342	The ev_idle watcher has determined that you have nothing better to do.	1049	The C<ev_idle> watcher has determined that you have nothing better to do.
343		1050
344	=item EV_PREPARE	1051	=item C<EV_PREPARE>
345		1052
346	=item EV_CHECK	1053	=item C<EV_CHECK>
347		1054
348	All ev_prepare watchers are invoked just I<before> C<ev_loop> starts	1055	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
349	to gather new events, and all ev_check watchers are invoked just after	1056	to gather new events, and all C<ev_check> watchers are invoked just after
350	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1057	C<ev_loop> has gathered them, but before it invokes any callbacks for any
351	received events. Callbacks of both watcher types can start and stop as	1058	received events. Callbacks of both watcher types can start and stop as
352	many watchers as they want, and all of them will be taken into account	1059	many watchers as they want, and all of them will be taken into account
353	(for example, a ev_prepare watcher might start an idle watcher to keep	1060	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
354	C<ev_loop> from blocking).	1061	C<ev_loop> from blocking).
355		1062
		1063	=item C<EV_EMBED>
		1064
		1065	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1066
		1067	=item C<EV_FORK>
		1068
		1069	The event loop has been resumed in the child process after fork (see
		1070	C<ev_fork>).
		1071
		1072	=item C<EV_ASYNC>
		1073
		1074	The given async watcher has been asynchronously notified (see C<ev_async>).
		1075
		1076	=item C<EV_CUSTOM>
		1077
		1078	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1079	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		1080
356	=item EV_ERROR	1081	=item C<EV_ERROR>
357		1082
358	An unspecified error has occured, the watcher has been stopped. This might	1083	An unspecified error has occurred, the watcher has been stopped. This might
359	happen because the watcher could not be properly started because libev	1084	happen because the watcher could not be properly started because libev
360	ran out of memory, a file descriptor was found to be closed or any other	1085	ran out of memory, a file descriptor was found to be closed or any other
		1086	problem. Libev considers these application bugs.
		1087
361	problem. You best act on it by reporting the problem and somehow coping	1088	You best act on it by reporting the problem and somehow coping with the
362	with the watcher being stopped.	1089	watcher being stopped. Note that well-written programs should not receive
		1090	an error ever, so when your watcher receives it, this usually indicates a
		1091	bug in your program.
363		1092
364	Libev will usually signal a few "dummy" events together with an error,	1093	Libev will usually signal a few "dummy" events together with an error, for
365	for example it might indicate that a fd is readable or writable, and if	1094	example it might indicate that a fd is readable or writable, and if your
366	your callbacks is well-written it can just attempt the operation and cope	1095	callbacks is well-written it can just attempt the operation and cope with
367	with the error from read() or write(). This will not work in multithreaded	1096	the error from read() or write(). This will not work in multi-threaded
368	programs, though, so beware.	1097	programs, though, as the fd could already be closed and reused for another
		1098	thing, so beware.
369		1099
370	=back	1100	=back
371		1101
		1102	=head2 GENERIC WATCHER FUNCTIONS
		1103
		1104	=over 4
		1105
		1106	=item C<ev_init> (ev_TYPE *watcher, callback)
		1107
		1108	This macro initialises the generic portion of a watcher. The contents
		1109	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1110	the generic parts of the watcher are initialised, you I<need> to call
		1111	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1112	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1113	which rolls both calls into one.
		1114
		1115	You can reinitialise a watcher at any time as long as it has been stopped
		1116	(or never started) and there are no pending events outstanding.
		1117
		1118	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1119	int revents)>.
		1120
		1121	Example: Initialise an C<ev_io> watcher in two steps.
		1122
		1123	ev_io w;
		1124	ev_init (&w, my_cb);
		1125	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1126
		1127	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		1128
		1129	This macro initialises the type-specific parts of a watcher. You need to
		1130	call C<ev_init> at least once before you call this macro, but you can
		1131	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1132	macro on a watcher that is active (it can be pending, however, which is a
		1133	difference to the C<ev_init> macro).
		1134
		1135	Although some watcher types do not have type-specific arguments
		1136	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1137
		1138	See C<ev_init>, above, for an example.
		1139
		1140	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1141
		1142	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1143	calls into a single call. This is the most convenient method to initialise
		1144	a watcher. The same limitations apply, of course.
		1145
		1146	Example: Initialise and set an C<ev_io> watcher in one step.
		1147
		1148	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1149
		1150	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		1151
		1152	Starts (activates) the given watcher. Only active watchers will receive
		1153	events. If the watcher is already active nothing will happen.
		1154
		1155	Example: Start the C<ev_io> watcher that is being abused as example in this
		1156	whole section.
		1157
		1158	ev_io_start (EV_DEFAULT_UC, &w);
		1159
		1160	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		1161
		1162	Stops the given watcher if active, and clears the pending status (whether
		1163	the watcher was active or not).
		1164
		1165	It is possible that stopped watchers are pending - for example,
		1166	non-repeating timers are being stopped when they become pending - but
		1167	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1168	pending. If you want to free or reuse the memory used by the watcher it is
		1169	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1170
		1171	=item bool ev_is_active (ev_TYPE *watcher)
		1172
		1173	Returns a true value iff the watcher is active (i.e. it has been started
		1174	and not yet been stopped). As long as a watcher is active you must not modify
		1175	it.
		1176
		1177	=item bool ev_is_pending (ev_TYPE *watcher)
		1178
		1179	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1180	events but its callback has not yet been invoked). As long as a watcher
		1181	is pending (but not active) you must not call an init function on it (but
		1182	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1183	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1184	it).
		1185
		1186	=item callback ev_cb (ev_TYPE *watcher)
		1187
		1188	Returns the callback currently set on the watcher.
		1189
		1190	=item ev_cb_set (ev_TYPE *watcher, callback)
		1191
		1192	Change the callback. You can change the callback at virtually any time
		1193	(modulo threads).
		1194
		1195	=item ev_set_priority (ev_TYPE *watcher, priority)
		1196
		1197	=item int ev_priority (ev_TYPE *watcher)
		1198
		1199	Set and query the priority of the watcher. The priority is a small
		1200	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1201	(default: C<-2>). Pending watchers with higher priority will be invoked
		1202	before watchers with lower priority, but priority will not keep watchers
		1203	from being executed (except for C<ev_idle> watchers).
		1204
		1205	If you need to suppress invocation when higher priority events are pending
		1206	you need to look at C<ev_idle> watchers, which provide this functionality.
		1207
		1208	You I<must not> change the priority of a watcher as long as it is active or
		1209	pending.
		1210
		1211	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1212	fine, as long as you do not mind that the priority value you query might
		1213	or might not have been clamped to the valid range.
		1214
		1215	The default priority used by watchers when no priority has been set is
		1216	always C<0>, which is supposed to not be too high and not be too low :).
		1217
		1218	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1219	priorities.
		1220
		1221	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1222
		1223	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1224	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1225	can deal with that fact, as both are simply passed through to the
		1226	callback.
		1227
		1228	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1229
		1230	If the watcher is pending, this function clears its pending status and
		1231	returns its C<revents> bitset (as if its callback was invoked). If the
		1232	watcher isn't pending it does nothing and returns C<0>.
		1233
		1234	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1235	callback to be invoked, which can be accomplished with this function.
		1236
		1237	=back
		1238
		1239
372	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1240	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
373		1241
374	Each watcher has, by default, a member C<void *data> that you can change	1242	Each watcher has, by default, a member C<void *data> that you can change
375	and read at any time, libev will completely ignore it. This cna be used	1243	and read at any time: libev will completely ignore it. This can be used
376	to associate arbitrary data with your watcher. If you need more data and	1244	to associate arbitrary data with your watcher. If you need more data and
377	don't want to allocate memory and store a pointer to it in that data	1245	don't want to allocate memory and store a pointer to it in that data
378	member, you can also "subclass" the watcher type and provide your own	1246	member, you can also "subclass" the watcher type and provide your own
379	data:	1247	data:
380		1248
381	struct my_io	1249	struct my_io
382	{	1250	{
383	struct ev_io io;	1251	ev_io io;
384	int otherfd;	1252	int otherfd;
385	void *somedata;	1253	void *somedata;
386	struct whatever *mostinteresting;	1254	struct whatever *mostinteresting;
387	}	1255	};
		1256
		1257	...
		1258	struct my_io w;
		1259	ev_io_init (&w.io, my_cb, fd, EV_READ);
388		1260
389	And since your callback will be called with a pointer to the watcher, you	1261	And since your callback will be called with a pointer to the watcher, you
390	can cast it back to your own type:	1262	can cast it back to your own type:
391		1263
392	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1264	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
393	{	1265	{
394	struct my_io *w = (struct my_io *)w_;	1266	struct my_io *w = (struct my_io *)w_;
395	...	1267	...
396	}	1268	}
397		1269
398	More interesting and less C-conformant ways of catsing your callback type	1270	More interesting and less C-conformant ways of casting your callback type
399	have been omitted....	1271	instead have been omitted.
		1272
		1273	Another common scenario is to use some data structure with multiple
		1274	embedded watchers:
		1275
		1276	struct my_biggy
		1277	{
		1278	int some_data;
		1279	ev_timer t1;
		1280	ev_timer t2;
		1281	}
		1282
		1283	In this case getting the pointer to C<my_biggy> is a bit more
		1284	complicated: Either you store the address of your C<my_biggy> struct
		1285	in the C<data> member of the watcher (for woozies), or you need to use
		1286	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1287	programmers):
		1288
		1289	#include <stddef.h>
		1290
		1291	static void
		1292	t1_cb (EV_P_ ev_timer *w, int revents)
		1293	{
		1294	struct my_biggy big = (struct my_biggy *)
		1295	(((char *)w) - offsetof (struct my_biggy, t1));
		1296	}
		1297
		1298	static void
		1299	t2_cb (EV_P_ ev_timer *w, int revents)
		1300	{
		1301	struct my_biggy big = (struct my_biggy *)
		1302	(((char *)w) - offsetof (struct my_biggy, t2));
		1303	}
		1304
		1305	=head2 WATCHER PRIORITY MODELS
		1306
		1307	Many event loops support I<watcher priorities>, which are usually small
		1308	integers that influence the ordering of event callback invocation
		1309	between watchers in some way, all else being equal.
		1310
		1311	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1312	description for the more technical details such as the actual priority
		1313	range.
		1314
		1315	There are two common ways how these these priorities are being interpreted
		1316	by event loops:
		1317
		1318	In the more common lock-out model, higher priorities "lock out" invocation
		1319	of lower priority watchers, which means as long as higher priority
		1320	watchers receive events, lower priority watchers are not being invoked.
		1321
		1322	The less common only-for-ordering model uses priorities solely to order
		1323	callback invocation within a single event loop iteration: Higher priority
		1324	watchers are invoked before lower priority ones, but they all get invoked
		1325	before polling for new events.
		1326
		1327	Libev uses the second (only-for-ordering) model for all its watchers
		1328	except for idle watchers (which use the lock-out model).
		1329
		1330	The rationale behind this is that implementing the lock-out model for
		1331	watchers is not well supported by most kernel interfaces, and most event
		1332	libraries will just poll for the same events again and again as long as
		1333	their callbacks have not been executed, which is very inefficient in the
		1334	common case of one high-priority watcher locking out a mass of lower
		1335	priority ones.
		1336
		1337	Static (ordering) priorities are most useful when you have two or more
		1338	watchers handling the same resource: a typical usage example is having an
		1339	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1340	timeouts. Under load, data might be received while the program handles
		1341	other jobs, but since timers normally get invoked first, the timeout
		1342	handler will be executed before checking for data. In that case, giving
		1343	the timer a lower priority than the I/O watcher ensures that I/O will be
		1344	handled first even under adverse conditions (which is usually, but not
		1345	always, what you want).
		1346
		1347	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1348	will only be executed when no same or higher priority watchers have
		1349	received events, they can be used to implement the "lock-out" model when
		1350	required.
		1351
		1352	For example, to emulate how many other event libraries handle priorities,
		1353	you can associate an C<ev_idle> watcher to each such watcher, and in
		1354	the normal watcher callback, you just start the idle watcher. The real
		1355	processing is done in the idle watcher callback. This causes libev to
		1356	continously poll and process kernel event data for the watcher, but when
		1357	the lock-out case is known to be rare (which in turn is rare :), this is
		1358	workable.
		1359
		1360	Usually, however, the lock-out model implemented that way will perform
		1361	miserably under the type of load it was designed to handle. In that case,
		1362	it might be preferable to stop the real watcher before starting the
		1363	idle watcher, so the kernel will not have to process the event in case
		1364	the actual processing will be delayed for considerable time.
		1365
		1366	Here is an example of an I/O watcher that should run at a strictly lower
		1367	priority than the default, and which should only process data when no
		1368	other events are pending:
		1369
		1370	ev_idle idle; // actual processing watcher
		1371	ev_io io; // actual event watcher
		1372
		1373	static void
		1374	io_cb (EV_P_ ev_io *w, int revents)
		1375	{
		1376	// stop the I/O watcher, we received the event, but
		1377	// are not yet ready to handle it.
		1378	ev_io_stop (EV_A_ w);
		1379
		1380	// start the idle watcher to ahndle the actual event.
		1381	// it will not be executed as long as other watchers
		1382	// with the default priority are receiving events.
		1383	ev_idle_start (EV_A_ &idle);
		1384	}
		1385
		1386	static void
		1387	idle_cb (EV_P_ ev_idle *w, int revents)
		1388	{
		1389	// actual processing
		1390	read (STDIN_FILENO, ...);
		1391
		1392	// have to start the I/O watcher again, as
		1393	// we have handled the event
		1394	ev_io_start (EV_P_ &io);
		1395	}
		1396
		1397	// initialisation
		1398	ev_idle_init (&idle, idle_cb);
		1399	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1400	ev_io_start (EV_DEFAULT_ &io);
		1401
		1402	In the "real" world, it might also be beneficial to start a timer, so that
		1403	low-priority connections can not be locked out forever under load. This
		1404	enables your program to keep a lower latency for important connections
		1405	during short periods of high load, while not completely locking out less
		1406	important ones.
400		1407
401		1408
402	=head1 WATCHER TYPES	1409	=head1 WATCHER TYPES
403		1410
404	This section describes each watcher in detail, but will not repeat	1411	This section describes each watcher in detail, but will not repeat
405	information given in the last section.	1412	information given in the last section. Any initialisation/set macros,
		1413	functions and members specific to the watcher type are explained.
406		1414
		1415	Members are additionally marked with either I<[read-only]>, meaning that,
		1416	while the watcher is active, you can look at the member and expect some
		1417	sensible content, but you must not modify it (you can modify it while the
		1418	watcher is stopped to your hearts content), or I<[read-write]>, which
		1419	means you can expect it to have some sensible content while the watcher
		1420	is active, but you can also modify it. Modifying it may not do something
		1421	sensible or take immediate effect (or do anything at all), but libev will
		1422	not crash or malfunction in any way.
		1423
		1424
407	=head2 struct ev_io - is my file descriptor readable or writable	1425	=head2 C<ev_io> - is this file descriptor readable or writable?
408		1426
409	I/O watchers check whether a file descriptor is readable or writable	1427	I/O watchers check whether a file descriptor is readable or writable
410	in each iteration of the event loop (This behaviour is called	1428	in each iteration of the event loop, or, more precisely, when reading
411	level-triggering because you keep receiving events as long as the	1429	would not block the process and writing would at least be able to write
412	condition persists. Remember you cna stop the watcher if you don't want to	1430	some data. This behaviour is called level-triggering because you keep
413	act on the event and neither want to receive future events).	1431	receiving events as long as the condition persists. Remember you can stop
		1432	the watcher if you don't want to act on the event and neither want to
		1433	receive future events.
414		1434
415	In general you can register as many read and/or write event watchers oer	1435	In general you can register as many read and/or write event watchers per
416	fd as you want (as long as you don't confuse yourself). Setting all file	1436	fd as you want (as long as you don't confuse yourself). Setting all file
417	descriptors to non-blocking mode is also usually a good idea (but not	1437	descriptors to non-blocking mode is also usually a good idea (but not
418	required if you know what you are doing).	1438	required if you know what you are doing).
419		1439
420	You have to be careful with dup'ed file descriptors, though. Some backends	1440	If you cannot use non-blocking mode, then force the use of a
421	(the linux epoll backend is a notable example) cannot handle dup'ed file	1441	known-to-be-good backend (at the time of this writing, this includes only
422	descriptors correctly if you register interest in two or more fds pointing	1442	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
423	to the same file/socket etc. description.	1443	descriptors for which non-blocking operation makes no sense (such as
		1444	files) - libev doesn't guarentee any specific behaviour in that case.
424		1445
425	If you must do this, then force the use of a known-to-be-good backend	1446	Another thing you have to watch out for is that it is quite easy to
426	(at the time of this writing, this includes only EVMETHOD_SELECT and	1447	receive "spurious" readiness notifications, that is your callback might
427	EVMETHOD_POLL).	1448	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1449	because there is no data. Not only are some backends known to create a
		1450	lot of those (for example Solaris ports), it is very easy to get into
		1451	this situation even with a relatively standard program structure. Thus
		1452	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1453	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1454
		1455	If you cannot run the fd in non-blocking mode (for example you should
		1456	not play around with an Xlib connection), then you have to separately
		1457	re-test whether a file descriptor is really ready with a known-to-be good
		1458	interface such as poll (fortunately in our Xlib example, Xlib already
		1459	does this on its own, so its quite safe to use). Some people additionally
		1460	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1461	indefinitely.
		1462
		1463	But really, best use non-blocking mode.
		1464
		1465	=head3 The special problem of disappearing file descriptors
		1466
		1467	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1468	descriptor (either due to calling C<close> explicitly or any other means,
		1469	such as C<dup2>). The reason is that you register interest in some file
		1470	descriptor, but when it goes away, the operating system will silently drop
		1471	this interest. If another file descriptor with the same number then is
		1472	registered with libev, there is no efficient way to see that this is, in
		1473	fact, a different file descriptor.
		1474
		1475	To avoid having to explicitly tell libev about such cases, libev follows
		1476	the following policy: Each time C<ev_io_set> is being called, libev
		1477	will assume that this is potentially a new file descriptor, otherwise
		1478	it is assumed that the file descriptor stays the same. That means that
		1479	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1480	descriptor even if the file descriptor number itself did not change.
		1481
		1482	This is how one would do it normally anyway, the important point is that
		1483	the libev application should not optimise around libev but should leave
		1484	optimisations to libev.
		1485
		1486	=head3 The special problem of dup'ed file descriptors
		1487
		1488	Some backends (e.g. epoll), cannot register events for file descriptors,
		1489	but only events for the underlying file descriptions. That means when you
		1490	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1491	events for them, only one file descriptor might actually receive events.
		1492
		1493	There is no workaround possible except not registering events
		1494	for potentially C<dup ()>'ed file descriptors, or to resort to
		1495	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1496
		1497	=head3 The special problem of fork
		1498
		1499	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1500	useless behaviour. Libev fully supports fork, but needs to be told about
		1501	it in the child.
		1502
		1503	To support fork in your programs, you either have to call
		1504	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1505	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1506	C<EVBACKEND_POLL>.
		1507
		1508	=head3 The special problem of SIGPIPE
		1509
		1510	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1511	when writing to a pipe whose other end has been closed, your program gets
		1512	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1513	this is sensible behaviour, for daemons, this is usually undesirable.
		1514
		1515	So when you encounter spurious, unexplained daemon exits, make sure you
		1516	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1517	somewhere, as that would have given you a big clue).
		1518
		1519
		1520	=head3 Watcher-Specific Functions
428		1521
429	=over 4	1522	=over 4
430		1523
431	=item ev_io_init (ev_io *, callback, int fd, int events)	1524	=item ev_io_init (ev_io *, callback, int fd, int events)
432		1525
433	=item ev_io_set (ev_io *, int fd, int events)	1526	=item ev_io_set (ev_io *, int fd, int events)
434		1527
435	Configures an ev_io watcher. The fd is the file descriptor to rceeive	1528	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
436	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1529	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
437	EV_WRITE> to receive the given events.	1530	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1531
		1532	=item int fd [read-only]
		1533
		1534	The file descriptor being watched.
		1535
		1536	=item int events [read-only]
		1537
		1538	The events being watched.
438		1539
439	=back	1540	=back
440		1541
		1542	=head3 Examples
		1543
		1544	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1545	readable, but only once. Since it is likely line-buffered, you could
		1546	attempt to read a whole line in the callback.
		1547
		1548	static void
		1549	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1550	{
		1551	ev_io_stop (loop, w);
		1552	.. read from stdin here (or from w->fd) and handle any I/O errors
		1553	}
		1554
		1555	...
		1556	struct ev_loop *loop = ev_default_init (0);
		1557	ev_io stdin_readable;
		1558	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1559	ev_io_start (loop, &stdin_readable);
		1560	ev_loop (loop, 0);
		1561
		1562
441	=head2 struct ev_timer - relative and optionally recurring timeouts	1563	=head2 C<ev_timer> - relative and optionally repeating timeouts
442		1564
443	Timer watchers are simple relative timers that generate an event after a	1565	Timer watchers are simple relative timers that generate an event after a
444	given time, and optionally repeating in regular intervals after that.	1566	given time, and optionally repeating in regular intervals after that.
445		1567
446	The timers are based on real time, that is, if you register an event that	1568	The timers are based on real time, that is, if you register an event that
447	times out after an hour and youreset your system clock to last years	1569	times out after an hour and you reset your system clock to January last
448	time, it will still time out after (roughly) and hour. "Roughly" because	1570	year, it will still time out after (roughly) one hour. "Roughly" because
449	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1571	detecting time jumps is hard, and some inaccuracies are unavoidable (the
450	monotonic clock option helps a lot here).	1572	monotonic clock option helps a lot here).
451		1573
		1574	The callback is guaranteed to be invoked only I<after> its timeout has
		1575	passed (not I<at>, so on systems with very low-resolution clocks this
		1576	might introduce a small delay). If multiple timers become ready during the
		1577	same loop iteration then the ones with earlier time-out values are invoked
		1578	before ones of the same priority with later time-out values (but this is
		1579	no longer true when a callback calls C<ev_loop> recursively).
		1580
		1581	=head3 Be smart about timeouts
		1582
		1583	Many real-world problems involve some kind of timeout, usually for error
		1584	recovery. A typical example is an HTTP request - if the other side hangs,
		1585	you want to raise some error after a while.
		1586
		1587	What follows are some ways to handle this problem, from obvious and
		1588	inefficient to smart and efficient.
		1589
		1590	In the following, a 60 second activity timeout is assumed - a timeout that
		1591	gets reset to 60 seconds each time there is activity (e.g. each time some
		1592	data or other life sign was received).
		1593
452	=over 4	1594	=over 4
453		1595
		1596	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1597
		1598	This is the most obvious, but not the most simple way: In the beginning,
		1599	start the watcher:
		1600
		1601	ev_timer_init (timer, callback, 60., 0.);
		1602	ev_timer_start (loop, timer);
		1603
		1604	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1605	and start it again:
		1606
		1607	ev_timer_stop (loop, timer);
		1608	ev_timer_set (timer, 60., 0.);
		1609	ev_timer_start (loop, timer);
		1610
		1611	This is relatively simple to implement, but means that each time there is
		1612	some activity, libev will first have to remove the timer from its internal
		1613	data structure and then add it again. Libev tries to be fast, but it's
		1614	still not a constant-time operation.
		1615
		1616	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1617
		1618	This is the easiest way, and involves using C<ev_timer_again> instead of
		1619	C<ev_timer_start>.
		1620
		1621	To implement this, configure an C<ev_timer> with a C<repeat> value
		1622	of C<60> and then call C<ev_timer_again> at start and each time you
		1623	successfully read or write some data. If you go into an idle state where
		1624	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1625	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1626
		1627	That means you can ignore both the C<ev_timer_start> function and the
		1628	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1629	member and C<ev_timer_again>.
		1630
		1631	At start:
		1632
		1633	ev_init (timer, callback);
		1634	timer->repeat = 60.;
		1635	ev_timer_again (loop, timer);
		1636
		1637	Each time there is some activity:
		1638
		1639	ev_timer_again (loop, timer);
		1640
		1641	It is even possible to change the time-out on the fly, regardless of
		1642	whether the watcher is active or not:
		1643
		1644	timer->repeat = 30.;
		1645	ev_timer_again (loop, timer);
		1646
		1647	This is slightly more efficient then stopping/starting the timer each time
		1648	you want to modify its timeout value, as libev does not have to completely
		1649	remove and re-insert the timer from/into its internal data structure.
		1650
		1651	It is, however, even simpler than the "obvious" way to do it.
		1652
		1653	=item 3. Let the timer time out, but then re-arm it as required.
		1654
		1655	This method is more tricky, but usually most efficient: Most timeouts are
		1656	relatively long compared to the intervals between other activity - in
		1657	our example, within 60 seconds, there are usually many I/O events with
		1658	associated activity resets.
		1659
		1660	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1661	but remember the time of last activity, and check for a real timeout only
		1662	within the callback:
		1663
		1664	ev_tstamp last_activity; // time of last activity
		1665
		1666	static void
		1667	callback (EV_P_ ev_timer *w, int revents)
		1668	{
		1669	ev_tstamp now = ev_now (EV_A);
		1670	ev_tstamp timeout = last_activity + 60.;
		1671
		1672	// if last_activity + 60. is older than now, we did time out
		1673	if (timeout < now)
		1674	{
		1675	// timeout occured, take action
		1676	}
		1677	else
		1678	{
		1679	// callback was invoked, but there was some activity, re-arm
		1680	// the watcher to fire in last_activity + 60, which is
		1681	// guaranteed to be in the future, so "again" is positive:
		1682	w->repeat = timeout - now;
		1683	ev_timer_again (EV_A_ w);
		1684	}
		1685	}
		1686
		1687	To summarise the callback: first calculate the real timeout (defined
		1688	as "60 seconds after the last activity"), then check if that time has
		1689	been reached, which means something I<did>, in fact, time out. Otherwise
		1690	the callback was invoked too early (C<timeout> is in the future), so
		1691	re-schedule the timer to fire at that future time, to see if maybe we have
		1692	a timeout then.
		1693
		1694	Note how C<ev_timer_again> is used, taking advantage of the
		1695	C<ev_timer_again> optimisation when the timer is already running.
		1696
		1697	This scheme causes more callback invocations (about one every 60 seconds
		1698	minus half the average time between activity), but virtually no calls to
		1699	libev to change the timeout.
		1700
		1701	To start the timer, simply initialise the watcher and set C<last_activity>
		1702	to the current time (meaning we just have some activity :), then call the
		1703	callback, which will "do the right thing" and start the timer:
		1704
		1705	ev_init (timer, callback);
		1706	last_activity = ev_now (loop);
		1707	callback (loop, timer, EV_TIMEOUT);
		1708
		1709	And when there is some activity, simply store the current time in
		1710	C<last_activity>, no libev calls at all:
		1711
		1712	last_actiivty = ev_now (loop);
		1713
		1714	This technique is slightly more complex, but in most cases where the
		1715	time-out is unlikely to be triggered, much more efficient.
		1716
		1717	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1718	callback :) - just change the timeout and invoke the callback, which will
		1719	fix things for you.
		1720
		1721	=item 4. Wee, just use a double-linked list for your timeouts.
		1722
		1723	If there is not one request, but many thousands (millions...), all
		1724	employing some kind of timeout with the same timeout value, then one can
		1725	do even better:
		1726
		1727	When starting the timeout, calculate the timeout value and put the timeout
		1728	at the I<end> of the list.
		1729
		1730	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1731	the list is expected to fire (for example, using the technique #3).
		1732
		1733	When there is some activity, remove the timer from the list, recalculate
		1734	the timeout, append it to the end of the list again, and make sure to
		1735	update the C<ev_timer> if it was taken from the beginning of the list.
		1736
		1737	This way, one can manage an unlimited number of timeouts in O(1) time for
		1738	starting, stopping and updating the timers, at the expense of a major
		1739	complication, and having to use a constant timeout. The constant timeout
		1740	ensures that the list stays sorted.
		1741
		1742	=back
		1743
		1744	So which method the best?
		1745
		1746	Method #2 is a simple no-brain-required solution that is adequate in most
		1747	situations. Method #3 requires a bit more thinking, but handles many cases
		1748	better, and isn't very complicated either. In most case, choosing either
		1749	one is fine, with #3 being better in typical situations.
		1750
		1751	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1752	rather complicated, but extremely efficient, something that really pays
		1753	off after the first million or so of active timers, i.e. it's usually
		1754	overkill :)
		1755
		1756	=head3 The special problem of time updates
		1757
		1758	Establishing the current time is a costly operation (it usually takes at
		1759	least two system calls): EV therefore updates its idea of the current
		1760	time only before and after C<ev_loop> collects new events, which causes a
		1761	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1762	lots of events in one iteration.
		1763
		1764	The relative timeouts are calculated relative to the C<ev_now ()>
		1765	time. This is usually the right thing as this timestamp refers to the time
		1766	of the event triggering whatever timeout you are modifying/starting. If
		1767	you suspect event processing to be delayed and you I<need> to base the
		1768	timeout on the current time, use something like this to adjust for this:
		1769
		1770	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
		1771
		1772	If the event loop is suspended for a long time, you can also force an
		1773	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1774	()>.
		1775
		1776	=head3 The special problems of suspended animation
		1777
		1778	When you leave the server world it is quite customary to hit machines that
		1779	can suspend/hibernate - what happens to the clocks during such a suspend?
		1780
		1781	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		1782	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		1783	to run until the system is suspended, but they will not advance while the
		1784	system is suspended. That means, on resume, it will be as if the program
		1785	was frozen for a few seconds, but the suspend time will not be counted
		1786	towards C<ev_timer> when a monotonic clock source is used. The real time
		1787	clock advanced as expected, but if it is used as sole clocksource, then a
		1788	long suspend would be detected as a time jump by libev, and timers would
		1789	be adjusted accordingly.
		1790
		1791	I would not be surprised to see different behaviour in different between
		1792	operating systems, OS versions or even different hardware.
		1793
		1794	The other form of suspend (job control, or sending a SIGSTOP) will see a
		1795	time jump in the monotonic clocks and the realtime clock. If the program
		1796	is suspended for a very long time, and monotonic clock sources are in use,
		1797	then you can expect C<ev_timer>s to expire as the full suspension time
		1798	will be counted towards the timers. When no monotonic clock source is in
		1799	use, then libev will again assume a timejump and adjust accordingly.
		1800
		1801	It might be beneficial for this latter case to call C<ev_suspend>
		1802	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		1803	deterministic behaviour in this case (you can do nothing against
		1804	C<SIGSTOP>).
		1805
		1806	=head3 Watcher-Specific Functions and Data Members
		1807
		1808	=over 4
		1809
454	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1810	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
455		1811
456	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1812	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
457		1813
458	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1814	Configure the timer to trigger after C<after> seconds. If C<repeat>
459	C<0.>, then it will automatically be stopped. If it is positive, then the	1815	is C<0.>, then it will automatically be stopped once the timeout is
460	timer will automatically be configured to trigger again C<repeat> seconds	1816	reached. If it is positive, then the timer will automatically be
461	later, again, and again, until stopped manually.	1817	configured to trigger again C<repeat> seconds later, again, and again,
		1818	until stopped manually.
462		1819
463	The timer itself will do a best-effort at avoiding drift, that is, if you	1820	The timer itself will do a best-effort at avoiding drift, that is, if
464	configure a timer to trigger every 10 seconds, then it will trigger at	1821	you configure a timer to trigger every 10 seconds, then it will normally
465	exactly 10 second intervals. If, however, your program cannot keep up with	1822	trigger at exactly 10 second intervals. If, however, your program cannot
466	the timer (ecause it takes longer than those 10 seconds to do stuff) the	1823	keep up with the timer (because it takes longer than those 10 seconds to
467	timer will not fire more than once per event loop iteration.	1824	do stuff) the timer will not fire more than once per event loop iteration.
468		1825
469	=item ev_timer_again (loop)	1826	=item ev_timer_again (loop, ev_timer *)
470		1827
471	This will act as if the timer timed out and restart it again if it is	1828	This will act as if the timer timed out and restart it again if it is
472	repeating. The exact semantics are:	1829	repeating. The exact semantics are:
473		1830
		1831	If the timer is pending, its pending status is cleared.
		1832
474	If the timer is started but nonrepeating, stop it.	1833	If the timer is started but non-repeating, stop it (as if it timed out).
475		1834
476	If the timer is repeating, either start it if necessary (with the repeat	1835	If the timer is repeating, either start it if necessary (with the
477	value), or reset the running timer to the repeat value.	1836	C<repeat> value), or reset the running timer to the C<repeat> value.
478		1837
479	This sounds a bit complicated, but here is a useful and typical	1838	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
480	example: Imagine you have a tcp connection and you want a so-called idle	1839	usage example.
481	timeout, that is, you want to be called when there have been, say, 60	1840
482	seconds of inactivity on the socket. The easiest way to do this is to	1841	=item ev_timer_remaining (loop, ev_timer *)
483	configure an ev_timer with after=repeat=60 and calling ev_timer_again each	1842
484	time you successfully read or write some data. If you go into an idle	1843	Returns the remaining time until a timer fires. If the timer is active,
485	state where you do not expect data to travel on the socket, you can stop	1844	then this time is relative to the current event loop time, otherwise it's
486	the timer, and again will automatically restart it if need be.	1845	the timeout value currently configured.
		1846
		1847	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		1848	C<5>. When the timer is started and one second passes, C<ev_timer_remain>
		1849	will return C<4>. When the timer expires and is restarted, it will return
		1850	roughly C<7> (likely slightly less as callback invocation takes some time,
		1851	too), and so on.
		1852
		1853	=item ev_tstamp repeat [read-write]
		1854
		1855	The current C<repeat> value. Will be used each time the watcher times out
		1856	or C<ev_timer_again> is called, and determines the next timeout (if any),
		1857	which is also when any modifications are taken into account.
487		1858
488	=back	1859	=back
489		1860
		1861	=head3 Examples
		1862
		1863	Example: Create a timer that fires after 60 seconds.
		1864
		1865	static void
		1866	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1867	{
		1868	.. one minute over, w is actually stopped right here
		1869	}
		1870
		1871	ev_timer mytimer;
		1872	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1873	ev_timer_start (loop, &mytimer);
		1874
		1875	Example: Create a timeout timer that times out after 10 seconds of
		1876	inactivity.
		1877
		1878	static void
		1879	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1880	{
		1881	.. ten seconds without any activity
		1882	}
		1883
		1884	ev_timer mytimer;
		1885	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1886	ev_timer_again (&mytimer); /* start timer */
		1887	ev_loop (loop, 0);
		1888
		1889	// and in some piece of code that gets executed on any "activity":
		1890	// reset the timeout to start ticking again at 10 seconds
		1891	ev_timer_again (&mytimer);
		1892
		1893
490	=head2 ev_periodic - to cron or not to cron it	1894	=head2 C<ev_periodic> - to cron or not to cron?
491		1895
492	Periodic watchers are also timers of a kind, but they are very versatile	1896	Periodic watchers are also timers of a kind, but they are very versatile
493	(and unfortunately a bit complex).	1897	(and unfortunately a bit complex).
494		1898
495	Unlike ev_timer's, they are not based on real time (or relative time)	1899	Unlike C<ev_timer>, periodic watchers are not based on real time (or
496	but on wallclock time (absolute time). You can tell a periodic watcher	1900	relative time, the physical time that passes) but on wall clock time
497	to trigger "at" some specific point in time. For example, if you tell a	1901	(absolute time, the thing you can read on your calender or clock). The
498	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1902	difference is that wall clock time can run faster or slower than real
499	+ 10.>) and then reset your system clock to the last year, then it will	1903	time, and time jumps are not uncommon (e.g. when you adjust your
500	take a year to trigger the event (unlike an ev_timer, which would trigger	1904	wrist-watch).
501	roughly 10 seconds later and of course not if you reset your system time
502	again).
503		1905
504	They can also be used to implement vastly more complex timers, such as	1906	You can tell a periodic watcher to trigger after some specific point
		1907	in time: for example, if you tell a periodic watcher to trigger "in 10
		1908	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1909	not a delay) and then reset your system clock to January of the previous
		1910	year, then it will take a year or more to trigger the event (unlike an
		1911	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1912	it, as it uses a relative timeout).
		1913
		1914	C<ev_periodic> watchers can also be used to implement vastly more complex
505	triggering an event on eahc midnight, local time.	1915	timers, such as triggering an event on each "midnight, local time", or
		1916	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1917	those cannot react to time jumps.
		1918
		1919	As with timers, the callback is guaranteed to be invoked only when the
		1920	point in time where it is supposed to trigger has passed. If multiple
		1921	timers become ready during the same loop iteration then the ones with
		1922	earlier time-out values are invoked before ones with later time-out values
		1923	(but this is no longer true when a callback calls C<ev_loop> recursively).
		1924
		1925	=head3 Watcher-Specific Functions and Data Members
506		1926
507	=over 4	1927	=over 4
508		1928
509	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1929	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
510		1930
511	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1931	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
512		1932
513	Lots of arguments, lets sort it out... There are basically three modes of	1933	Lots of arguments, let's sort it out... There are basically three modes of
514	operation, and we will explain them from simplest to complex:	1934	operation, and we will explain them from simplest to most complex:
515
516		1935
517	=over 4	1936	=over 4
518		1937
519	=item * absolute timer (interval = reschedule_cb = 0)	1938	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
520		1939
521	In this configuration the watcher triggers an event at the wallclock time	1940	In this configuration the watcher triggers an event after the wall clock
522	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1941	time C<offset> has passed. It will not repeat and will not adjust when a
523	that is, if it is to be run at January 1st 2011 then it will run when the	1942	time jump occurs, that is, if it is to be run at January 1st 2011 then it
524	system time reaches or surpasses this time.	1943	will be stopped and invoked when the system clock reaches or surpasses
		1944	this point in time.
525		1945
526	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1946	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
527		1947
528	In this mode the watcher will always be scheduled to time out at the next	1948	In this mode the watcher will always be scheduled to time out at the next
529	C<at + N * interval> time (for some integer N) and then repeat, regardless	1949	C<offset + N * interval> time (for some integer N, which can also be
530	of any time jumps.	1950	negative) and then repeat, regardless of any time jumps. The C<offset>
		1951	argument is merely an offset into the C<interval> periods.
531		1952
532	This can be used to create timers that do not drift with respect to system	1953	This can be used to create timers that do not drift with respect to the
533	time:	1954	system clock, for example, here is an C<ev_periodic> that triggers each
		1955	hour, on the hour (with respect to UTC):
534		1956
535	ev_periodic_set (&periodic, 0., 3600., 0);	1957	ev_periodic_set (&periodic, 0., 3600., 0);
536		1958
537	This doesn't mean there will always be 3600 seconds in between triggers,	1959	This doesn't mean there will always be 3600 seconds in between triggers,
538	but only that the the callback will be called when the system time shows a	1960	but only that the callback will be called when the system time shows a
539	full hour (UTC), or more correct, when the system time is evenly divisible	1961	full hour (UTC), or more correctly, when the system time is evenly divisible
540	by 3600.	1962	by 3600.
541		1963
542	Another way to think about it (for the mathematically inclined) is that	1964	Another way to think about it (for the mathematically inclined) is that
543	ev_periodic will try to run the callback in this mode at the next possible	1965	C<ev_periodic> will try to run the callback in this mode at the next possible
544	time where C<time = at (mod interval)>, regardless of any time jumps.	1966	time where C<time = offset (mod interval)>, regardless of any time jumps.
545		1967
		1968	For numerical stability it is preferable that the C<offset> value is near
		1969	C<ev_now ()> (the current time), but there is no range requirement for
		1970	this value, and in fact is often specified as zero.
		1971
		1972	Note also that there is an upper limit to how often a timer can fire (CPU
		1973	speed for example), so if C<interval> is very small then timing stability
		1974	will of course deteriorate. Libev itself tries to be exact to be about one
		1975	millisecond (if the OS supports it and the machine is fast enough).
		1976
546	=item * manual reschedule mode (reschedule_cb = callback)	1977	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
547		1978
548	In this mode the values for C<interval> and C<at> are both being	1979	In this mode the values for C<interval> and C<offset> are both being
549	ignored. Instead, each time the periodic watcher gets scheduled, the	1980	ignored. Instead, each time the periodic watcher gets scheduled, the
550	reschedule callback will be called with the watcher as first, and the	1981	reschedule callback will be called with the watcher as first, and the
551	current time as second argument.	1982	current time as second argument.
552		1983
553	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	1984	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
554	periodic watcher, ever, or make any event loop modificstions>. If you need	1985	or make ANY other event loop modifications whatsoever, unless explicitly
555	to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards.	1986	allowed by documentation here>.
556		1987
		1988	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1989	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1990	only event loop modification you are allowed to do).
		1991
557	Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1992	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
558	ev_tstamp now)>, e.g.:	1993	*w, ev_tstamp now)>, e.g.:
559		1994
		1995	static ev_tstamp
560	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1996	my_rescheduler (ev_periodic *w, ev_tstamp now)
561	{	1997	{
562	return now + 60.;	1998	return now + 60.;
563	}	1999	}
564		2000
565	It must return the next time to trigger, based on the passed time value	2001	It must return the next time to trigger, based on the passed time value
566	(that is, the lowest time value larger than to the second argument). It	2002	(that is, the lowest time value larger than to the second argument). It
567	will usually be called just before the callback will be triggered, but	2003	will usually be called just before the callback will be triggered, but
568	might be called at other times, too.	2004	might be called at other times, too.
569		2005
		2006	NOTE: I<< This callback must always return a time that is higher than or
		2007	equal to the passed C<now> value >>.
		2008
570	This can be used to create very complex timers, such as a timer that	2009	This can be used to create very complex timers, such as a timer that
571	triggers on each midnight, local time. To do this, you would calculate the	2010	triggers on "next midnight, local time". To do this, you would calculate the
572	next midnight after C<now> and return the timestamp value for this. How you do this	2011	next midnight after C<now> and return the timestamp value for this. How
573	is, again, up to you (but it is not trivial).	2012	you do this is, again, up to you (but it is not trivial, which is the main
		2013	reason I omitted it as an example).
574		2014
575	=back	2015	=back
576		2016
577	=item ev_periodic_again (loop, ev_periodic *)	2017	=item ev_periodic_again (loop, ev_periodic *)
578		2018
579	Simply stops and restarts the periodic watcher again. This is only useful	2019	Simply stops and restarts the periodic watcher again. This is only useful
580	when you changed some parameters or the reschedule callback would return	2020	when you changed some parameters or the reschedule callback would return
581	a different time than the last time it was called (e.g. in a crond like	2021	a different time than the last time it was called (e.g. in a crond like
582	program when the crontabs have changed).	2022	program when the crontabs have changed).
583		2023
		2024	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2025
		2026	When active, returns the absolute time that the watcher is supposed
		2027	to trigger next. This is not the same as the C<offset> argument to
		2028	C<ev_periodic_set>, but indeed works even in interval and manual
		2029	rescheduling modes.
		2030
		2031	=item ev_tstamp offset [read-write]
		2032
		2033	When repeating, this contains the offset value, otherwise this is the
		2034	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2035	although libev might modify this value for better numerical stability).
		2036
		2037	Can be modified any time, but changes only take effect when the periodic
		2038	timer fires or C<ev_periodic_again> is being called.
		2039
		2040	=item ev_tstamp interval [read-write]
		2041
		2042	The current interval value. Can be modified any time, but changes only
		2043	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2044	called.
		2045
		2046	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2047
		2048	The current reschedule callback, or C<0>, if this functionality is
		2049	switched off. Can be changed any time, but changes only take effect when
		2050	the periodic timer fires or C<ev_periodic_again> is being called.
		2051
584	=back	2052	=back
585		2053
		2054	=head3 Examples
		2055
		2056	Example: Call a callback every hour, or, more precisely, whenever the
		2057	system time is divisible by 3600. The callback invocation times have
		2058	potentially a lot of jitter, but good long-term stability.
		2059
		2060	static void
		2061	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		2062	{
		2063	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2064	}
		2065
		2066	ev_periodic hourly_tick;
		2067	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2068	ev_periodic_start (loop, &hourly_tick);
		2069
		2070	Example: The same as above, but use a reschedule callback to do it:
		2071
		2072	#include <math.h>
		2073
		2074	static ev_tstamp
		2075	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2076	{
		2077	return now + (3600. - fmod (now, 3600.));
		2078	}
		2079
		2080	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2081
		2082	Example: Call a callback every hour, starting now:
		2083
		2084	ev_periodic hourly_tick;
		2085	ev_periodic_init (&hourly_tick, clock_cb,
		2086	fmod (ev_now (loop), 3600.), 3600., 0);
		2087	ev_periodic_start (loop, &hourly_tick);
		2088
		2089
586	=head2 ev_signal - signal me when a signal gets signalled	2090	=head2 C<ev_signal> - signal me when a signal gets signalled!
587		2091
588	Signal watchers will trigger an event when the process receives a specific	2092	Signal watchers will trigger an event when the process receives a specific
589	signal one or more times. Even though signals are very asynchronous, libev	2093	signal one or more times. Even though signals are very asynchronous, libev
590	will try its best to deliver signals synchronously, i.e. as part of the	2094	will try it's best to deliver signals synchronously, i.e. as part of the
591	normal event processing, like any other event.	2095	normal event processing, like any other event.
592		2096
		2097	If you want signals to be delivered truly asynchronously, just use
		2098	C<sigaction> as you would do without libev and forget about sharing
		2099	the signal. You can even use C<ev_async> from a signal handler to
		2100	synchronously wake up an event loop.
		2101
593	You cna configure as many watchers as you like per signal. Only when the	2102	You can configure as many watchers as you like for the same signal, but
		2103	only within the same loop, i.e. you can watch for C<SIGINT> in your
		2104	default loop and for C<SIGIO> in another loop, but you cannot watch for
		2105	C<SIGINT> in both the default loop and another loop at the same time. At
		2106	the moment, C<SIGCHLD> is permanently tied to the default loop.
		2107
594	first watcher gets started will libev actually register a signal watcher	2108	When the first watcher gets started will libev actually register something
595	with the kernel (thus it coexists with your own signal handlers as long	2109	with the kernel (thus it coexists with your own signal handlers as long as
596	as you don't register any with libev). Similarly, when the last signal	2110	you don't register any with libev for the same signal).
597	watcher for a signal is stopped libev will reset the signal handler to	2111
598	SIG_DFL (regardless of what it was set to before).	2112	Both the signal mask state (C<sigprocmask>) and the signal handler state
		2113	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2114	sotpping it again), that is, libev might or might not block the signal,
		2115	and might or might not set or restore the installed signal handler.
		2116
		2117	If possible and supported, libev will install its handlers with
		2118	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2119	not be unduly interrupted. If you have a problem with system calls getting
		2120	interrupted by signals you can block all signals in an C<ev_check> watcher
		2121	and unblock them in an C<ev_prepare> watcher.
		2122
		2123	=head3 Watcher-Specific Functions and Data Members
599		2124
600	=over 4	2125	=over 4
601		2126
602	=item ev_signal_init (ev_signal *, callback, int signum)	2127	=item ev_signal_init (ev_signal *, callback, int signum)
603		2128
604	=item ev_signal_set (ev_signal *, int signum)	2129	=item ev_signal_set (ev_signal *, int signum)
605		2130
606	Configures the watcher to trigger on the given signal number (usually one	2131	Configures the watcher to trigger on the given signal number (usually one
607	of the C<SIGxxx> constants).	2132	of the C<SIGxxx> constants).
608		2133
		2134	=item int signum [read-only]
		2135
		2136	The signal the watcher watches out for.
		2137
609	=back	2138	=back
610		2139
		2140	=head3 Examples
		2141
		2142	Example: Try to exit cleanly on SIGINT.
		2143
		2144	static void
		2145	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2146	{
		2147	ev_unloop (loop, EVUNLOOP_ALL);
		2148	}
		2149
		2150	ev_signal signal_watcher;
		2151	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2152	ev_signal_start (loop, &signal_watcher);
		2153
		2154
611	=head2 ev_child - wait for pid status changes	2155	=head2 C<ev_child> - watch out for process status changes
612		2156
613	Child watchers trigger when your process receives a SIGCHLD in response to	2157	Child watchers trigger when your process receives a SIGCHLD in response to
614	some child status changes (most typically when a child of yours dies).	2158	some child status changes (most typically when a child of yours dies or
		2159	exits). It is permissible to install a child watcher I<after> the child
		2160	has been forked (which implies it might have already exited), as long
		2161	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2162	forking and then immediately registering a watcher for the child is fine,
		2163	but forking and registering a watcher a few event loop iterations later or
		2164	in the next callback invocation is not.
		2165
		2166	Only the default event loop is capable of handling signals, and therefore
		2167	you can only register child watchers in the default event loop.
		2168
		2169	Due to some design glitches inside libev, child watchers will always be
		2170	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2171	libev)
		2172
		2173	=head3 Process Interaction
		2174
		2175	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2176	initialised. This is necessary to guarantee proper behaviour even if the
		2177	first child watcher is started after the child exits. The occurrence
		2178	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2179	synchronously as part of the event loop processing. Libev always reaps all
		2180	children, even ones not watched.
		2181
		2182	=head3 Overriding the Built-In Processing
		2183
		2184	Libev offers no special support for overriding the built-in child
		2185	processing, but if your application collides with libev's default child
		2186	handler, you can override it easily by installing your own handler for
		2187	C<SIGCHLD> after initialising the default loop, and making sure the
		2188	default loop never gets destroyed. You are encouraged, however, to use an
		2189	event-based approach to child reaping and thus use libev's support for
		2190	that, so other libev users can use C<ev_child> watchers freely.
		2191
		2192	=head3 Stopping the Child Watcher
		2193
		2194	Currently, the child watcher never gets stopped, even when the
		2195	child terminates, so normally one needs to stop the watcher in the
		2196	callback. Future versions of libev might stop the watcher automatically
		2197	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2198	problem).
		2199
		2200	=head3 Watcher-Specific Functions and Data Members
615		2201
616	=over 4	2202	=over 4
617		2203
618	=item ev_child_init (ev_child *, callback, int pid)	2204	=item ev_child_init (ev_child *, callback, int pid, int trace)
619		2205
620	=item ev_child_set (ev_child *, int pid)	2206	=item ev_child_set (ev_child *, int pid, int trace)
621		2207
622	Configures the watcher to wait for status changes of process C<pid> (or	2208	Configures the watcher to wait for status changes of process C<pid> (or
623	I<any> process if C<pid> is specified as C<0>). The callback can look	2209	I<any> process if C<pid> is specified as C<0>). The callback can look
624	at the C<rstatus> member of the C<ev_child> watcher structure to see	2210	at the C<rstatus> member of the C<ev_child> watcher structure to see
625	the status word (use the macros from C<sys/wait.h>). The C<rpid> member	2211	the status word (use the macros from C<sys/wait.h> and see your systems
626	contains the pid of the process causing the status change.	2212	C<waitpid> documentation). The C<rpid> member contains the pid of the
		2213	process causing the status change. C<trace> must be either C<0> (only
		2214	activate the watcher when the process terminates) or C<1> (additionally
		2215	activate the watcher when the process is stopped or continued).
		2216
		2217	=item int pid [read-only]
		2218
		2219	The process id this watcher watches out for, or C<0>, meaning any process id.
		2220
		2221	=item int rpid [read-write]
		2222
		2223	The process id that detected a status change.
		2224
		2225	=item int rstatus [read-write]
		2226
		2227	The process exit/trace status caused by C<rpid> (see your systems
		2228	C<waitpid> and C<sys/wait.h> documentation for details).
627		2229
628	=back	2230	=back
629		2231
		2232	=head3 Examples
		2233
		2234	Example: C<fork()> a new process and install a child handler to wait for
		2235	its completion.
		2236
		2237	ev_child cw;
		2238
		2239	static void
		2240	child_cb (EV_P_ ev_child *w, int revents)
		2241	{
		2242	ev_child_stop (EV_A_ w);
		2243	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2244	}
		2245
		2246	pid_t pid = fork ();
		2247
		2248	if (pid < 0)
		2249	// error
		2250	else if (pid == 0)
		2251	{
		2252	// the forked child executes here
		2253	exit (1);
		2254	}
		2255	else
		2256	{
		2257	ev_child_init (&cw, child_cb, pid, 0);
		2258	ev_child_start (EV_DEFAULT_ &cw);
		2259	}
		2260
		2261
		2262	=head2 C<ev_stat> - did the file attributes just change?
		2263
		2264	This watches a file system path for attribute changes. That is, it calls
		2265	C<stat> on that path in regular intervals (or when the OS says it changed)
		2266	and sees if it changed compared to the last time, invoking the callback if
		2267	it did.
		2268
		2269	The path does not need to exist: changing from "path exists" to "path does
		2270	not exist" is a status change like any other. The condition "path does not
		2271	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2272	C<st_nlink> field being zero (which is otherwise always forced to be at
		2273	least one) and all the other fields of the stat buffer having unspecified
		2274	contents.
		2275
		2276	The path I<must not> end in a slash or contain special components such as
		2277	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2278	your working directory changes, then the behaviour is undefined.
		2279
		2280	Since there is no portable change notification interface available, the
		2281	portable implementation simply calls C<stat(2)> regularly on the path
		2282	to see if it changed somehow. You can specify a recommended polling
		2283	interval for this case. If you specify a polling interval of C<0> (highly
		2284	recommended!) then a I<suitable, unspecified default> value will be used
		2285	(which you can expect to be around five seconds, although this might
		2286	change dynamically). Libev will also impose a minimum interval which is
		2287	currently around C<0.1>, but that's usually overkill.
		2288
		2289	This watcher type is not meant for massive numbers of stat watchers,
		2290	as even with OS-supported change notifications, this can be
		2291	resource-intensive.
		2292
		2293	At the time of this writing, the only OS-specific interface implemented
		2294	is the Linux inotify interface (implementing kqueue support is left as an
		2295	exercise for the reader. Note, however, that the author sees no way of
		2296	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2297
		2298	=head3 ABI Issues (Largefile Support)
		2299
		2300	Libev by default (unless the user overrides this) uses the default
		2301	compilation environment, which means that on systems with large file
		2302	support disabled by default, you get the 32 bit version of the stat
		2303	structure. When using the library from programs that change the ABI to
		2304	use 64 bit file offsets the programs will fail. In that case you have to
		2305	compile libev with the same flags to get binary compatibility. This is
		2306	obviously the case with any flags that change the ABI, but the problem is
		2307	most noticeably displayed with ev_stat and large file support.
		2308
		2309	The solution for this is to lobby your distribution maker to make large
		2310	file interfaces available by default (as e.g. FreeBSD does) and not
		2311	optional. Libev cannot simply switch on large file support because it has
		2312	to exchange stat structures with application programs compiled using the
		2313	default compilation environment.
		2314
		2315	=head3 Inotify and Kqueue
		2316
		2317	When C<inotify (7)> support has been compiled into libev and present at
		2318	runtime, it will be used to speed up change detection where possible. The
		2319	inotify descriptor will be created lazily when the first C<ev_stat>
		2320	watcher is being started.
		2321
		2322	Inotify presence does not change the semantics of C<ev_stat> watchers
		2323	except that changes might be detected earlier, and in some cases, to avoid
		2324	making regular C<stat> calls. Even in the presence of inotify support
		2325	there are many cases where libev has to resort to regular C<stat> polling,
		2326	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2327	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2328	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2329	xfs are fully working) libev usually gets away without polling.
		2330
		2331	There is no support for kqueue, as apparently it cannot be used to
		2332	implement this functionality, due to the requirement of having a file
		2333	descriptor open on the object at all times, and detecting renames, unlinks
		2334	etc. is difficult.
		2335
		2336	=head3 C<stat ()> is a synchronous operation
		2337
		2338	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2339	the process. The exception are C<ev_stat> watchers - those call C<stat
		2340	()>, which is a synchronous operation.
		2341
		2342	For local paths, this usually doesn't matter: unless the system is very
		2343	busy or the intervals between stat's are large, a stat call will be fast,
		2344	as the path data is usually in memory already (except when starting the
		2345	watcher).
		2346
		2347	For networked file systems, calling C<stat ()> can block an indefinite
		2348	time due to network issues, and even under good conditions, a stat call
		2349	often takes multiple milliseconds.
		2350
		2351	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2352	paths, although this is fully supported by libev.
		2353
		2354	=head3 The special problem of stat time resolution
		2355
		2356	The C<stat ()> system call only supports full-second resolution portably,
		2357	and even on systems where the resolution is higher, most file systems
		2358	still only support whole seconds.
		2359
		2360	That means that, if the time is the only thing that changes, you can
		2361	easily miss updates: on the first update, C<ev_stat> detects a change and
		2362	calls your callback, which does something. When there is another update
		2363	within the same second, C<ev_stat> will be unable to detect unless the
		2364	stat data does change in other ways (e.g. file size).
		2365
		2366	The solution to this is to delay acting on a change for slightly more
		2367	than a second (or till slightly after the next full second boundary), using
		2368	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2369	ev_timer_again (loop, w)>).
		2370
		2371	The C<.02> offset is added to work around small timing inconsistencies
		2372	of some operating systems (where the second counter of the current time
		2373	might be be delayed. One such system is the Linux kernel, where a call to
		2374	C<gettimeofday> might return a timestamp with a full second later than
		2375	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2376	update file times then there will be a small window where the kernel uses
		2377	the previous second to update file times but libev might already execute
		2378	the timer callback).
		2379
		2380	=head3 Watcher-Specific Functions and Data Members
		2381
		2382	=over 4
		2383
		2384	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2385
		2386	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2387
		2388	Configures the watcher to wait for status changes of the given
		2389	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2390	be detected and should normally be specified as C<0> to let libev choose
		2391	a suitable value. The memory pointed to by C<path> must point to the same
		2392	path for as long as the watcher is active.
		2393
		2394	The callback will receive an C<EV_STAT> event when a change was detected,
		2395	relative to the attributes at the time the watcher was started (or the
		2396	last change was detected).
		2397
		2398	=item ev_stat_stat (loop, ev_stat *)
		2399
		2400	Updates the stat buffer immediately with new values. If you change the
		2401	watched path in your callback, you could call this function to avoid
		2402	detecting this change (while introducing a race condition if you are not
		2403	the only one changing the path). Can also be useful simply to find out the
		2404	new values.
		2405
		2406	=item ev_statdata attr [read-only]
		2407
		2408	The most-recently detected attributes of the file. Although the type is
		2409	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2410	suitable for your system, but you can only rely on the POSIX-standardised
		2411	members to be present. If the C<st_nlink> member is C<0>, then there was
		2412	some error while C<stat>ing the file.
		2413
		2414	=item ev_statdata prev [read-only]
		2415
		2416	The previous attributes of the file. The callback gets invoked whenever
		2417	C<prev> != C<attr>, or, more precisely, one or more of these members
		2418	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2419	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2420
		2421	=item ev_tstamp interval [read-only]
		2422
		2423	The specified interval.
		2424
		2425	=item const char *path [read-only]
		2426
		2427	The file system path that is being watched.
		2428
		2429	=back
		2430
		2431	=head3 Examples
		2432
		2433	Example: Watch C</etc/passwd> for attribute changes.
		2434
		2435	static void
		2436	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2437	{
		2438	/* /etc/passwd changed in some way */
		2439	if (w->attr.st_nlink)
		2440	{
		2441	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2442	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2443	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2444	}
		2445	else
		2446	/* you shalt not abuse printf for puts */
		2447	puts ("wow, /etc/passwd is not there, expect problems. "
		2448	"if this is windows, they already arrived\n");
		2449	}
		2450
		2451	...
		2452	ev_stat passwd;
		2453
		2454	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2455	ev_stat_start (loop, &passwd);
		2456
		2457	Example: Like above, but additionally use a one-second delay so we do not
		2458	miss updates (however, frequent updates will delay processing, too, so
		2459	one might do the work both on C<ev_stat> callback invocation I<and> on
		2460	C<ev_timer> callback invocation).
		2461
		2462	static ev_stat passwd;
		2463	static ev_timer timer;
		2464
		2465	static void
		2466	timer_cb (EV_P_ ev_timer *w, int revents)
		2467	{
		2468	ev_timer_stop (EV_A_ w);
		2469
		2470	/* now it's one second after the most recent passwd change */
		2471	}
		2472
		2473	static void
		2474	stat_cb (EV_P_ ev_stat *w, int revents)
		2475	{
		2476	/* reset the one-second timer */
		2477	ev_timer_again (EV_A_ &timer);
		2478	}
		2479
		2480	...
		2481	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2482	ev_stat_start (loop, &passwd);
		2483	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2484
		2485
630	=head2 ev_idle - when you've got nothing better to do	2486	=head2 C<ev_idle> - when you've got nothing better to do...
631		2487
632	Idle watchers trigger events when there are no other I/O or timer (or	2488	Idle watchers trigger events when no other events of the same or higher
633	periodic) events pending. That is, as long as your process is busy	2489	priority are pending (prepare, check and other idle watchers do not count
634	handling sockets or timeouts it will not be called. But when your process	2490	as receiving "events").
635	is idle all idle watchers are being called again and again - until	2491
		2492	That is, as long as your process is busy handling sockets or timeouts
		2493	(or even signals, imagine) of the same or higher priority it will not be
		2494	triggered. But when your process is idle (or only lower-priority watchers
		2495	are pending), the idle watchers are being called once per event loop
636	stopped, that is, or your process receives more events.	2496	iteration - until stopped, that is, or your process receives more events
		2497	and becomes busy again with higher priority stuff.
637		2498
638	The most noteworthy effect is that as long as any idle watchers are	2499	The most noteworthy effect is that as long as any idle watchers are
639	active, the process will not block when waiting for new events.	2500	active, the process will not block when waiting for new events.
640		2501
641	Apart from keeping your process non-blocking (which is a useful	2502	Apart from keeping your process non-blocking (which is a useful
642	effect on its own sometimes), idle watchers are a good place to do	2503	effect on its own sometimes), idle watchers are a good place to do
643	"pseudo-background processing", or delay processing stuff to after the	2504	"pseudo-background processing", or delay processing stuff to after the
644	event loop has handled all outstanding events.	2505	event loop has handled all outstanding events.
645		2506
		2507	=head3 Watcher-Specific Functions and Data Members
		2508
646	=over 4	2509	=over 4
647		2510
648	=item ev_idle_init (ev_signal *, callback)	2511	=item ev_idle_init (ev_idle *, callback)
649		2512
650	Initialises and configures the idle watcher - it has no parameters of any	2513	Initialises and configures the idle watcher - it has no parameters of any
651	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2514	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
652	believe me.	2515	believe me.
653		2516
654	=back	2517	=back
655		2518
656	=head2 prepare and check - your hooks into the event loop	2519	=head3 Examples
657		2520
		2521	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2522	callback, free it. Also, use no error checking, as usual.
		2523
		2524	static void
		2525	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2526	{
		2527	free (w);
		2528	// now do something you wanted to do when the program has
		2529	// no longer anything immediate to do.
		2530	}
		2531
		2532	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2533	ev_idle_init (idle_watcher, idle_cb);
		2534	ev_idle_start (loop, idle_watcher);
		2535
		2536
		2537	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
		2538
658	Prepare and check watchers usually (but not always) are used in	2539	Prepare and check watchers are usually (but not always) used in pairs:
659	tandom. Prepare watchers get invoked before the process blocks and check	2540	prepare watchers get invoked before the process blocks and check watchers
660	watchers afterwards.	2541	afterwards.
661		2542
		2543	You I<must not> call C<ev_loop> or similar functions that enter
		2544	the current event loop from either C<ev_prepare> or C<ev_check>
		2545	watchers. Other loops than the current one are fine, however. The
		2546	rationale behind this is that you do not need to check for recursion in
		2547	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2548	C<ev_check> so if you have one watcher of each kind they will always be
		2549	called in pairs bracketing the blocking call.
		2550
662	Their main purpose is to integrate other event mechanisms into libev. This	2551	Their main purpose is to integrate other event mechanisms into libev and
663	could be used, for example, to track variable changes, implement your own	2552	their use is somewhat advanced. They could be used, for example, to track
664	watchers, integrate net-snmp or a coroutine library and lots more.	2553	variable changes, implement your own watchers, integrate net-snmp or a
		2554	coroutine library and lots more. They are also occasionally useful if
		2555	you cache some data and want to flush it before blocking (for example,
		2556	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2557	watcher).
665		2558
666	This is done by examining in each prepare call which file descriptors need	2559	This is done by examining in each prepare call which file descriptors
667	to be watched by the other library, registering ev_io watchers for them	2560	need to be watched by the other library, registering C<ev_io> watchers
668	and starting an ev_timer watcher for any timeouts (many libraries provide	2561	for them and starting an C<ev_timer> watcher for any timeouts (many
669	just this functionality). Then, in the check watcher you check for any	2562	libraries provide exactly this functionality). Then, in the check watcher,
670	events that occured (by making your callbacks set soem flags for example)	2563	you check for any events that occurred (by checking the pending status
671	and call back into the library.	2564	of all watchers and stopping them) and call back into the library. The
		2565	I/O and timer callbacks will never actually be called (but must be valid
		2566	nevertheless, because you never know, you know?).
672		2567
673	As another example, the perl Coro module uses these hooks to integrate	2568	As another example, the Perl Coro module uses these hooks to integrate
674	coroutines into libev programs, by yielding to other active coroutines	2569	coroutines into libev programs, by yielding to other active coroutines
675	during each prepare and only letting the process block if no coroutines	2570	during each prepare and only letting the process block if no coroutines
676	are ready to run.	2571	are ready to run (it's actually more complicated: it only runs coroutines
		2572	with priority higher than or equal to the event loop and one coroutine
		2573	of lower priority, but only once, using idle watchers to keep the event
		2574	loop from blocking if lower-priority coroutines are active, thus mapping
		2575	low-priority coroutines to idle/background tasks).
		2576
		2577	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2578	priority, to ensure that they are being run before any other watchers
		2579	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2580
		2581	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2582	activate ("feed") events into libev. While libev fully supports this, they
		2583	might get executed before other C<ev_check> watchers did their job. As
		2584	C<ev_check> watchers are often used to embed other (non-libev) event
		2585	loops those other event loops might be in an unusable state until their
		2586	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2587	others).
		2588
		2589	=head3 Watcher-Specific Functions and Data Members
677		2590
678	=over 4	2591	=over 4
679		2592
680	=item ev_prepare_init (ev_prepare *, callback)	2593	=item ev_prepare_init (ev_prepare *, callback)
681		2594
682	=item ev_check_init (ev_check *, callback)	2595	=item ev_check_init (ev_check *, callback)
683		2596
684	Initialises and configures the prepare or check watcher - they have no	2597	Initialises and configures the prepare or check watcher - they have no
685	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2598	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
686	macros, but using them is utterly, utterly pointless.	2599	macros, but using them is utterly, utterly, utterly and completely
		2600	pointless.
687		2601
688	=back	2602	=back
689		2603
		2604	=head3 Examples
		2605
		2606	There are a number of principal ways to embed other event loops or modules
		2607	into libev. Here are some ideas on how to include libadns into libev
		2608	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2609	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2610	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2611	Glib event loop).
		2612
		2613	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2614	and in a check watcher, destroy them and call into libadns. What follows
		2615	is pseudo-code only of course. This requires you to either use a low
		2616	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2617	the callbacks for the IO/timeout watchers might not have been called yet.
		2618
		2619	static ev_io iow [nfd];
		2620	static ev_timer tw;
		2621
		2622	static void
		2623	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2624	{
		2625	}
		2626
		2627	// create io watchers for each fd and a timer before blocking
		2628	static void
		2629	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2630	{
		2631	int timeout = 3600000;
		2632	struct pollfd fds [nfd];
		2633	// actual code will need to loop here and realloc etc.
		2634	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2635
		2636	/* the callback is illegal, but won't be called as we stop during check */
		2637	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		2638	ev_timer_start (loop, &tw);
		2639
		2640	// create one ev_io per pollfd
		2641	for (int i = 0; i < nfd; ++i)
		2642	{
		2643	ev_io_init (iow + i, io_cb, fds [i].fd,
		2644	((fds [i].events & POLLIN ? EV_READ : 0)
		2645	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2646
		2647	fds [i].revents = 0;
		2648	ev_io_start (loop, iow + i);
		2649	}
		2650	}
		2651
		2652	// stop all watchers after blocking
		2653	static void
		2654	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2655	{
		2656	ev_timer_stop (loop, &tw);
		2657
		2658	for (int i = 0; i < nfd; ++i)
		2659	{
		2660	// set the relevant poll flags
		2661	// could also call adns_processreadable etc. here
		2662	struct pollfd *fd = fds + i;
		2663	int revents = ev_clear_pending (iow + i);
		2664	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2665	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2666
		2667	// now stop the watcher
		2668	ev_io_stop (loop, iow + i);
		2669	}
		2670
		2671	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2672	}
		2673
		2674	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2675	in the prepare watcher and would dispose of the check watcher.
		2676
		2677	Method 3: If the module to be embedded supports explicit event
		2678	notification (libadns does), you can also make use of the actual watcher
		2679	callbacks, and only destroy/create the watchers in the prepare watcher.
		2680
		2681	static void
		2682	timer_cb (EV_P_ ev_timer *w, int revents)
		2683	{
		2684	adns_state ads = (adns_state)w->data;
		2685	update_now (EV_A);
		2686
		2687	adns_processtimeouts (ads, &tv_now);
		2688	}
		2689
		2690	static void
		2691	io_cb (EV_P_ ev_io *w, int revents)
		2692	{
		2693	adns_state ads = (adns_state)w->data;
		2694	update_now (EV_A);
		2695
		2696	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2697	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2698	}
		2699
		2700	// do not ever call adns_afterpoll
		2701
		2702	Method 4: Do not use a prepare or check watcher because the module you
		2703	want to embed is not flexible enough to support it. Instead, you can
		2704	override their poll function. The drawback with this solution is that the
		2705	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2706	this approach, effectively embedding EV as a client into the horrible
		2707	libglib event loop.
		2708
		2709	static gint
		2710	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2711	{
		2712	int got_events = 0;
		2713
		2714	for (n = 0; n < nfds; ++n)
		2715	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2716
		2717	if (timeout >= 0)
		2718	// create/start timer
		2719
		2720	// poll
		2721	ev_loop (EV_A_ 0);
		2722
		2723	// stop timer again
		2724	if (timeout >= 0)
		2725	ev_timer_stop (EV_A_ &to);
		2726
		2727	// stop io watchers again - their callbacks should have set
		2728	for (n = 0; n < nfds; ++n)
		2729	ev_io_stop (EV_A_ iow [n]);
		2730
		2731	return got_events;
		2732	}
		2733
		2734
		2735	=head2 C<ev_embed> - when one backend isn't enough...
		2736
		2737	This is a rather advanced watcher type that lets you embed one event loop
		2738	into another (currently only C<ev_io> events are supported in the embedded
		2739	loop, other types of watchers might be handled in a delayed or incorrect
		2740	fashion and must not be used).
		2741
		2742	There are primarily two reasons you would want that: work around bugs and
		2743	prioritise I/O.
		2744
		2745	As an example for a bug workaround, the kqueue backend might only support
		2746	sockets on some platform, so it is unusable as generic backend, but you
		2747	still want to make use of it because you have many sockets and it scales
		2748	so nicely. In this case, you would create a kqueue-based loop and embed
		2749	it into your default loop (which might use e.g. poll). Overall operation
		2750	will be a bit slower because first libev has to call C<poll> and then
		2751	C<kevent>, but at least you can use both mechanisms for what they are
		2752	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2753
		2754	As for prioritising I/O: under rare circumstances you have the case where
		2755	some fds have to be watched and handled very quickly (with low latency),
		2756	and even priorities and idle watchers might have too much overhead. In
		2757	this case you would put all the high priority stuff in one loop and all
		2758	the rest in a second one, and embed the second one in the first.
		2759
		2760	As long as the watcher is active, the callback will be invoked every
		2761	time there might be events pending in the embedded loop. The callback
		2762	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		2763	sweep and invoke their callbacks (the callback doesn't need to invoke the
		2764	C<ev_embed_sweep> function directly, it could also start an idle watcher
		2765	to give the embedded loop strictly lower priority for example).
		2766
		2767	You can also set the callback to C<0>, in which case the embed watcher
		2768	will automatically execute the embedded loop sweep whenever necessary.
		2769
		2770	Fork detection will be handled transparently while the C<ev_embed> watcher
		2771	is active, i.e., the embedded loop will automatically be forked when the
		2772	embedding loop forks. In other cases, the user is responsible for calling
		2773	C<ev_loop_fork> on the embedded loop.
		2774
		2775	Unfortunately, not all backends are embeddable: only the ones returned by
		2776	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2777	portable one.
		2778
		2779	So when you want to use this feature you will always have to be prepared
		2780	that you cannot get an embeddable loop. The recommended way to get around
		2781	this is to have a separate variables for your embeddable loop, try to
		2782	create it, and if that fails, use the normal loop for everything.
		2783
		2784	=head3 C<ev_embed> and fork
		2785
		2786	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2787	automatically be applied to the embedded loop as well, so no special
		2788	fork handling is required in that case. When the watcher is not running,
		2789	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2790	as applicable.
		2791
		2792	=head3 Watcher-Specific Functions and Data Members
		2793
		2794	=over 4
		2795
		2796	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2797
		2798	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2799
		2800	Configures the watcher to embed the given loop, which must be
		2801	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2802	invoked automatically, otherwise it is the responsibility of the callback
		2803	to invoke it (it will continue to be called until the sweep has been done,
		2804	if you do not want that, you need to temporarily stop the embed watcher).
		2805
		2806	=item ev_embed_sweep (loop, ev_embed *)
		2807
		2808	Make a single, non-blocking sweep over the embedded loop. This works
		2809	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2810	appropriate way for embedded loops.
		2811
		2812	=item struct ev_loop *other [read-only]
		2813
		2814	The embedded event loop.
		2815
		2816	=back
		2817
		2818	=head3 Examples
		2819
		2820	Example: Try to get an embeddable event loop and embed it into the default
		2821	event loop. If that is not possible, use the default loop. The default
		2822	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2823	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2824	used).
		2825
		2826	struct ev_loop *loop_hi = ev_default_init (0);
		2827	struct ev_loop *loop_lo = 0;
		2828	ev_embed embed;
		2829
		2830	// see if there is a chance of getting one that works
		2831	// (remember that a flags value of 0 means autodetection)
		2832	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2833	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2834	: 0;
		2835
		2836	// if we got one, then embed it, otherwise default to loop_hi
		2837	if (loop_lo)
		2838	{
		2839	ev_embed_init (&embed, 0, loop_lo);
		2840	ev_embed_start (loop_hi, &embed);
		2841	}
		2842	else
		2843	loop_lo = loop_hi;
		2844
		2845	Example: Check if kqueue is available but not recommended and create
		2846	a kqueue backend for use with sockets (which usually work with any
		2847	kqueue implementation). Store the kqueue/socket-only event loop in
		2848	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2849
		2850	struct ev_loop *loop = ev_default_init (0);
		2851	struct ev_loop *loop_socket = 0;
		2852	ev_embed embed;
		2853
		2854	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2855	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2856	{
		2857	ev_embed_init (&embed, 0, loop_socket);
		2858	ev_embed_start (loop, &embed);
		2859	}
		2860
		2861	if (!loop_socket)
		2862	loop_socket = loop;
		2863
		2864	// now use loop_socket for all sockets, and loop for everything else
		2865
		2866
		2867	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2868
		2869	Fork watchers are called when a C<fork ()> was detected (usually because
		2870	whoever is a good citizen cared to tell libev about it by calling
		2871	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2872	event loop blocks next and before C<ev_check> watchers are being called,
		2873	and only in the child after the fork. If whoever good citizen calling
		2874	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2875	handlers will be invoked, too, of course.
		2876
		2877	=head3 The special problem of life after fork - how is it possible?
		2878
		2879	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2880	up/change the process environment, followed by a call to C<exec()>. This
		2881	sequence should be handled by libev without any problems.
		2882
		2883	This changes when the application actually wants to do event handling
		2884	in the child, or both parent in child, in effect "continuing" after the
		2885	fork.
		2886
		2887	The default mode of operation (for libev, with application help to detect
		2888	forks) is to duplicate all the state in the child, as would be expected
		2889	when I<either> the parent I<or> the child process continues.
		2890
		2891	When both processes want to continue using libev, then this is usually the
		2892	wrong result. In that case, usually one process (typically the parent) is
		2893	supposed to continue with all watchers in place as before, while the other
		2894	process typically wants to start fresh, i.e. without any active watchers.
		2895
		2896	The cleanest and most efficient way to achieve that with libev is to
		2897	simply create a new event loop, which of course will be "empty", and
		2898	use that for new watchers. This has the advantage of not touching more
		2899	memory than necessary, and thus avoiding the copy-on-write, and the
		2900	disadvantage of having to use multiple event loops (which do not support
		2901	signal watchers).
		2902
		2903	When this is not possible, or you want to use the default loop for
		2904	other reasons, then in the process that wants to start "fresh", call
		2905	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2906	the default loop will "orphan" (not stop) all registered watchers, so you
		2907	have to be careful not to execute code that modifies those watchers. Note
		2908	also that in that case, you have to re-register any signal watchers.
		2909
		2910	=head3 Watcher-Specific Functions and Data Members
		2911
		2912	=over 4
		2913
		2914	=item ev_fork_init (ev_signal *, callback)
		2915
		2916	Initialises and configures the fork watcher - it has no parameters of any
		2917	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2918	believe me.
		2919
		2920	=back
		2921
		2922
		2923	=head2 C<ev_async> - how to wake up another event loop
		2924
		2925	In general, you cannot use an C<ev_loop> from multiple threads or other
		2926	asynchronous sources such as signal handlers (as opposed to multiple event
		2927	loops - those are of course safe to use in different threads).
		2928
		2929	Sometimes, however, you need to wake up another event loop you do not
		2930	control, for example because it belongs to another thread. This is what
		2931	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2932	can signal it by calling C<ev_async_send>, which is thread- and signal
		2933	safe.
		2934
		2935	This functionality is very similar to C<ev_signal> watchers, as signals,
		2936	too, are asynchronous in nature, and signals, too, will be compressed
		2937	(i.e. the number of callback invocations may be less than the number of
		2938	C<ev_async_sent> calls).
		2939
		2940	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2941	just the default loop.
		2942
		2943	=head3 Queueing
		2944
		2945	C<ev_async> does not support queueing of data in any way. The reason
		2946	is that the author does not know of a simple (or any) algorithm for a
		2947	multiple-writer-single-reader queue that works in all cases and doesn't
		2948	need elaborate support such as pthreads.
		2949
		2950	That means that if you want to queue data, you have to provide your own
		2951	queue. But at least I can tell you how to implement locking around your
		2952	queue:
		2953
		2954	=over 4
		2955
		2956	=item queueing from a signal handler context
		2957
		2958	To implement race-free queueing, you simply add to the queue in the signal
		2959	handler but you block the signal handler in the watcher callback. Here is
		2960	an example that does that for some fictitious SIGUSR1 handler:
		2961
		2962	static ev_async mysig;
		2963
		2964	static void
		2965	sigusr1_handler (void)
		2966	{
		2967	sometype data;
		2968
		2969	// no locking etc.
		2970	queue_put (data);
		2971	ev_async_send (EV_DEFAULT_ &mysig);
		2972	}
		2973
		2974	static void
		2975	mysig_cb (EV_P_ ev_async *w, int revents)
		2976	{
		2977	sometype data;
		2978	sigset_t block, prev;
		2979
		2980	sigemptyset (&block);
		2981	sigaddset (&block, SIGUSR1);
		2982	sigprocmask (SIG_BLOCK, &block, &prev);
		2983
		2984	while (queue_get (&data))
		2985	process (data);
		2986
		2987	if (sigismember (&prev, SIGUSR1)
		2988	sigprocmask (SIG_UNBLOCK, &block, 0);
		2989	}
		2990
		2991	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2992	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2993	either...).
		2994
		2995	=item queueing from a thread context
		2996
		2997	The strategy for threads is different, as you cannot (easily) block
		2998	threads but you can easily preempt them, so to queue safely you need to
		2999	employ a traditional mutex lock, such as in this pthread example:
		3000
		3001	static ev_async mysig;
		3002	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3003
		3004	static void
		3005	otherthread (void)
		3006	{
		3007	// only need to lock the actual queueing operation
		3008	pthread_mutex_lock (&mymutex);
		3009	queue_put (data);
		3010	pthread_mutex_unlock (&mymutex);
		3011
		3012	ev_async_send (EV_DEFAULT_ &mysig);
		3013	}
		3014
		3015	static void
		3016	mysig_cb (EV_P_ ev_async *w, int revents)
		3017	{
		3018	pthread_mutex_lock (&mymutex);
		3019
		3020	while (queue_get (&data))
		3021	process (data);
		3022
		3023	pthread_mutex_unlock (&mymutex);
		3024	}
		3025
		3026	=back
		3027
		3028
		3029	=head3 Watcher-Specific Functions and Data Members
		3030
		3031	=over 4
		3032
		3033	=item ev_async_init (ev_async *, callback)
		3034
		3035	Initialises and configures the async watcher - it has no parameters of any
		3036	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3037	trust me.
		3038
		3039	=item ev_async_send (loop, ev_async *)
		3040
		3041	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3042	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		3043	C<ev_feed_event>, this call is safe to do from other threads, signal or
		3044	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		3045	section below on what exactly this means).
		3046
		3047	Note that, as with other watchers in libev, multiple events might get
		3048	compressed into a single callback invocation (another way to look at this
		3049	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		3050	reset when the event loop detects that).
		3051
		3052	This call incurs the overhead of a system call only once per event loop
		3053	iteration, so while the overhead might be noticeable, it doesn't apply to
		3054	repeated calls to C<ev_async_send> for the same event loop.
		3055
		3056	=item bool = ev_async_pending (ev_async *)
		3057
		3058	Returns a non-zero value when C<ev_async_send> has been called on the
		3059	watcher but the event has not yet been processed (or even noted) by the
		3060	event loop.
		3061
		3062	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3063	the loop iterates next and checks for the watcher to have become active,
		3064	it will reset the flag again. C<ev_async_pending> can be used to very
		3065	quickly check whether invoking the loop might be a good idea.
		3066
		3067	Not that this does I<not> check whether the watcher itself is pending,
		3068	only whether it has been requested to make this watcher pending: there
		3069	is a time window between the event loop checking and resetting the async
		3070	notification, and the callback being invoked.
		3071
		3072	=back
		3073
		3074
690	=head1 OTHER FUNCTIONS	3075	=head1 OTHER FUNCTIONS
691		3076
692	There are some other fucntions of possible interest. Described. Here. Now.	3077	There are some other functions of possible interest. Described. Here. Now.
693		3078
694	=over 4	3079	=over 4
695		3080
696	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3081	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
697		3082
698	This function combines a simple timer and an I/O watcher, calls your	3083	This function combines a simple timer and an I/O watcher, calls your
699	callback on whichever event happens first and automatically stop both	3084	callback on whichever event happens first and automatically stops both
700	watchers. This is useful if you want to wait for a single event on an fd	3085	watchers. This is useful if you want to wait for a single event on an fd
701	or timeout without havign to allocate/configure/start/stop/free one or	3086	or timeout without having to allocate/configure/start/stop/free one or
702	more watchers yourself.	3087	more watchers yourself.
703		3088
704	If C<fd> is less than 0, then no I/O watcher will be started and events is	3089	If C<fd> is less than 0, then no I/O watcher will be started and the
705	ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set	3090	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
706	will be craeted and started.	3091	the given C<fd> and C<events> set will be created and started.
707		3092
708	If C<timeout> is less than 0, then no timeout watcher will be	3093	If C<timeout> is less than 0, then no timeout watcher will be
709	started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat	3094	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
710	= 0) will be started.	3095	repeat = 0) will be started. C<0> is a valid timeout.
711		3096
712	The callback has the type C<void (*cb)(int revents, void *arg)> and	3097	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
713	gets passed an events set (normally a combination of EV_ERROR, EV_READ,	3098	passed an C<revents> set like normal event callbacks (a combination of
714	EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>:	3099	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
		3100	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3101	a timeout and an io event at the same time - you probably should give io
		3102	events precedence.
715		3103
		3104	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3105
716	static void stdin_ready (int revents, void *arg)	3106	static void stdin_ready (int revents, void *arg)
717	{	3107	{
718	if (revents & EV_TIMEOUT)
719	/* doh, nothing entered */
720	else if (revents & EV_READ)	3108	if (revents & EV_READ)
721	/* stdin might have data for us, joy! */	3109	/* stdin might have data for us, joy! */;
		3110	else if (revents & EV_TIMEOUT)
		3111	/* doh, nothing entered */;
722	}	3112	}
723		3113
724	ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0);	3114	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
725		3115
726	=item ev_feed_event (loop, watcher, int events)	3116	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
727		3117
728	Feeds the given event set into the event loop, as if the specified event	3118	Feeds the given event set into the event loop, as if the specified event
729	has happened for the specified watcher (which must be a pointer to an	3119	had happened for the specified watcher (which must be a pointer to an
730	initialised but not necessarily active event watcher).	3120	initialised but not necessarily started event watcher).
731		3121
732	=item ev_feed_fd_event (loop, int fd, int revents)	3122	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
733		3123
734	Feed an event on the given fd, as if a file descriptor backend detected it.	3124	Feed an event on the given fd, as if a file descriptor backend detected
		3125	the given events it.
735		3126
736	=item ev_feed_signal_event (loop, int signum)	3127	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
737		3128
738	Feed an event as if the given signal occured (loop must be the default loop!).	3129	Feed an event as if the given signal occurred (C<loop> must be the default
		3130	loop!).
739		3131
740	=back	3132	=back
741		3133
		3134
		3135	=head1 LIBEVENT EMULATION
		3136
		3137	Libev offers a compatibility emulation layer for libevent. It cannot
		3138	emulate the internals of libevent, so here are some usage hints:
		3139
		3140	=over 4
		3141
		3142	=item * Use it by including <event.h>, as usual.
		3143
		3144	=item * The following members are fully supported: ev_base, ev_callback,
		3145	ev_arg, ev_fd, ev_res, ev_events.
		3146
		3147	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		3148	maintained by libev, it does not work exactly the same way as in libevent (consider
		3149	it a private API).
		3150
		3151	=item * Priorities are not currently supported. Initialising priorities
		3152	will fail and all watchers will have the same priority, even though there
		3153	is an ev_pri field.
		3154
		3155	=item * In libevent, the last base created gets the signals, in libev, the
		3156	first base created (== the default loop) gets the signals.
		3157
		3158	=item * Other members are not supported.
		3159
		3160	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		3161	to use the libev header file and library.
		3162
		3163	=back
		3164
		3165	=head1 C++ SUPPORT
		3166
		3167	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		3168	you to use some convenience methods to start/stop watchers and also change
		3169	the callback model to a model using method callbacks on objects.
		3170
		3171	To use it,
		3172
		3173	#include <ev++.h>
		3174
		3175	This automatically includes F<ev.h> and puts all of its definitions (many
		3176	of them macros) into the global namespace. All C++ specific things are
		3177	put into the C<ev> namespace. It should support all the same embedding
		3178	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		3179
		3180	Care has been taken to keep the overhead low. The only data member the C++
		3181	classes add (compared to plain C-style watchers) is the event loop pointer
		3182	that the watcher is associated with (or no additional members at all if
		3183	you disable C<EV_MULTIPLICITY> when embedding libev).
		3184
		3185	Currently, functions, and static and non-static member functions can be
		3186	used as callbacks. Other types should be easy to add as long as they only
		3187	need one additional pointer for context. If you need support for other
		3188	types of functors please contact the author (preferably after implementing
		3189	it).
		3190
		3191	Here is a list of things available in the C<ev> namespace:
		3192
		3193	=over 4
		3194
		3195	=item C<ev::READ>, C<ev::WRITE> etc.
		3196
		3197	These are just enum values with the same values as the C<EV_READ> etc.
		3198	macros from F<ev.h>.
		3199
		3200	=item C<ev::tstamp>, C<ev::now>
		3201
		3202	Aliases to the same types/functions as with the C<ev_> prefix.
		3203
		3204	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		3205
		3206	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		3207	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		3208	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		3209	defines by many implementations.
		3210
		3211	All of those classes have these methods:
		3212
		3213	=over 4
		3214
		3215	=item ev::TYPE::TYPE ()
		3216
		3217	=item ev::TYPE::TYPE (struct ev_loop *)
		3218
		3219	=item ev::TYPE::~TYPE
		3220
		3221	The constructor (optionally) takes an event loop to associate the watcher
		3222	with. If it is omitted, it will use C<EV_DEFAULT>.
		3223
		3224	The constructor calls C<ev_init> for you, which means you have to call the
		3225	C<set> method before starting it.
		3226
		3227	It will not set a callback, however: You have to call the templated C<set>
		3228	method to set a callback before you can start the watcher.
		3229
		3230	(The reason why you have to use a method is a limitation in C++ which does
		3231	not allow explicit template arguments for constructors).
		3232
		3233	The destructor automatically stops the watcher if it is active.
		3234
		3235	=item w->set<class, &class::method> (object *)
		3236
		3237	This method sets the callback method to call. The method has to have a
		3238	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		3239	first argument and the C<revents> as second. The object must be given as
		3240	parameter and is stored in the C<data> member of the watcher.
		3241
		3242	This method synthesizes efficient thunking code to call your method from
		3243	the C callback that libev requires. If your compiler can inline your
		3244	callback (i.e. it is visible to it at the place of the C<set> call and
		3245	your compiler is good :), then the method will be fully inlined into the
		3246	thunking function, making it as fast as a direct C callback.
		3247
		3248	Example: simple class declaration and watcher initialisation
		3249
		3250	struct myclass
		3251	{
		3252	void io_cb (ev::io &w, int revents) { }
		3253	}
		3254
		3255	myclass obj;
		3256	ev::io iow;
		3257	iow.set <myclass, &myclass::io_cb> (&obj);
		3258
		3259	=item w->set (object *)
		3260
		3261	This is an B<experimental> feature that might go away in a future version.
		3262
		3263	This is a variation of a method callback - leaving out the method to call
		3264	will default the method to C<operator ()>, which makes it possible to use
		3265	functor objects without having to manually specify the C<operator ()> all
		3266	the time. Incidentally, you can then also leave out the template argument
		3267	list.
		3268
		3269	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3270	int revents)>.
		3271
		3272	See the method-C<set> above for more details.
		3273
		3274	Example: use a functor object as callback.
		3275
		3276	struct myfunctor
		3277	{
		3278	void operator() (ev::io &w, int revents)
		3279	{
		3280	...
		3281	}
		3282	}
		3283
		3284	myfunctor f;
		3285
		3286	ev::io w;
		3287	w.set (&f);
		3288
		3289	=item w->set<function> (void *data = 0)
		3290
		3291	Also sets a callback, but uses a static method or plain function as
		3292	callback. The optional C<data> argument will be stored in the watcher's
		3293	C<data> member and is free for you to use.
		3294
		3295	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		3296
		3297	See the method-C<set> above for more details.
		3298
		3299	Example: Use a plain function as callback.
		3300
		3301	static void io_cb (ev::io &w, int revents) { }
		3302	iow.set <io_cb> ();
		3303
		3304	=item w->set (struct ev_loop *)
		3305
		3306	Associates a different C<struct ev_loop> with this watcher. You can only
		3307	do this when the watcher is inactive (and not pending either).
		3308
		3309	=item w->set ([arguments])
		3310
		3311	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		3312	called at least once. Unlike the C counterpart, an active watcher gets
		3313	automatically stopped and restarted when reconfiguring it with this
		3314	method.
		3315
		3316	=item w->start ()
		3317
		3318	Starts the watcher. Note that there is no C<loop> argument, as the
		3319	constructor already stores the event loop.
		3320
		3321	=item w->stop ()
		3322
		3323	Stops the watcher if it is active. Again, no C<loop> argument.
		3324
		3325	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		3326
		3327	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		3328	C<ev_TYPE_again> function.
		3329
		3330	=item w->sweep () (C<ev::embed> only)
		3331
		3332	Invokes C<ev_embed_sweep>.
		3333
		3334	=item w->update () (C<ev::stat> only)
		3335
		3336	Invokes C<ev_stat_stat>.
		3337
		3338	=back
		3339
		3340	=back
		3341
		3342	Example: Define a class with an IO and idle watcher, start one of them in
		3343	the constructor.
		3344
		3345	class myclass
		3346	{
		3347	ev::io io ; void io_cb (ev::io &w, int revents);
		3348	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		3349
		3350	myclass (int fd)
		3351	{
		3352	io .set <myclass, &myclass::io_cb > (this);
		3353	idle.set <myclass, &myclass::idle_cb> (this);
		3354
		3355	io.start (fd, ev::READ);
		3356	}
		3357	};
		3358
		3359
		3360	=head1 OTHER LANGUAGE BINDINGS
		3361
		3362	Libev does not offer other language bindings itself, but bindings for a
		3363	number of languages exist in the form of third-party packages. If you know
		3364	any interesting language binding in addition to the ones listed here, drop
		3365	me a note.
		3366
		3367	=over 4
		3368
		3369	=item Perl
		3370
		3371	The EV module implements the full libev API and is actually used to test
		3372	libev. EV is developed together with libev. Apart from the EV core module,
		3373	there are additional modules that implement libev-compatible interfaces
		3374	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		3375	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		3376	and C<EV::Glib>).
		3377
		3378	It can be found and installed via CPAN, its homepage is at
		3379	L<http://software.schmorp.de/pkg/EV>.
		3380
		3381	=item Python
		3382
		3383	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		3384	seems to be quite complete and well-documented.
		3385
		3386	=item Ruby
		3387
		3388	Tony Arcieri has written a ruby extension that offers access to a subset
		3389	of the libev API and adds file handle abstractions, asynchronous DNS and
		3390	more on top of it. It can be found via gem servers. Its homepage is at
		3391	L<http://rev.rubyforge.org/>.
		3392
		3393	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3394	makes rev work even on mingw.
		3395
		3396	=item Haskell
		3397
		3398	A haskell binding to libev is available at
		3399	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3400
		3401	=item D
		3402
		3403	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3404	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3405
		3406	=item Ocaml
		3407
		3408	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3409	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3410
		3411	=back
		3412
		3413
		3414	=head1 MACRO MAGIC
		3415
		3416	Libev can be compiled with a variety of options, the most fundamental
		3417	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3418	functions and callbacks have an initial C<struct ev_loop *> argument.
		3419
		3420	To make it easier to write programs that cope with either variant, the
		3421	following macros are defined:
		3422
		3423	=over 4
		3424
		3425	=item C<EV_A>, C<EV_A_>
		3426
		3427	This provides the loop I<argument> for functions, if one is required ("ev
		3428	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3429	C<EV_A_> is used when other arguments are following. Example:
		3430
		3431	ev_unref (EV_A);
		3432	ev_timer_add (EV_A_ watcher);
		3433	ev_loop (EV_A_ 0);
		3434
		3435	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3436	which is often provided by the following macro.
		3437
		3438	=item C<EV_P>, C<EV_P_>
		3439
		3440	This provides the loop I<parameter> for functions, if one is required ("ev
		3441	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3442	C<EV_P_> is used when other parameters are following. Example:
		3443
		3444	// this is how ev_unref is being declared
		3445	static void ev_unref (EV_P);
		3446
		3447	// this is how you can declare your typical callback
		3448	static void cb (EV_P_ ev_timer *w, int revents)
		3449
		3450	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3451	suitable for use with C<EV_A>.
		3452
		3453	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3454
		3455	Similar to the other two macros, this gives you the value of the default
		3456	loop, if multiple loops are supported ("ev loop default").
		3457
		3458	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3459
		3460	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3461	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3462	is undefined when the default loop has not been initialised by a previous
		3463	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3464
		3465	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3466	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3467
		3468	=back
		3469
		3470	Example: Declare and initialise a check watcher, utilising the above
		3471	macros so it will work regardless of whether multiple loops are supported
		3472	or not.
		3473
		3474	static void
		3475	check_cb (EV_P_ ev_timer *w, int revents)
		3476	{
		3477	ev_check_stop (EV_A_ w);
		3478	}
		3479
		3480	ev_check check;
		3481	ev_check_init (&check, check_cb);
		3482	ev_check_start (EV_DEFAULT_ &check);
		3483	ev_loop (EV_DEFAULT_ 0);
		3484
		3485	=head1 EMBEDDING
		3486
		3487	Libev can (and often is) directly embedded into host
		3488	applications. Examples of applications that embed it include the Deliantra
		3489	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3490	and rxvt-unicode.
		3491
		3492	The goal is to enable you to just copy the necessary files into your
		3493	source directory without having to change even a single line in them, so
		3494	you can easily upgrade by simply copying (or having a checked-out copy of
		3495	libev somewhere in your source tree).
		3496
		3497	=head2 FILESETS
		3498
		3499	Depending on what features you need you need to include one or more sets of files
		3500	in your application.
		3501
		3502	=head3 CORE EVENT LOOP
		3503
		3504	To include only the libev core (all the C<ev_*> functions), with manual
		3505	configuration (no autoconf):
		3506
		3507	#define EV_STANDALONE 1
		3508	#include "ev.c"
		3509
		3510	This will automatically include F<ev.h>, too, and should be done in a
		3511	single C source file only to provide the function implementations. To use
		3512	it, do the same for F<ev.h> in all files wishing to use this API (best
		3513	done by writing a wrapper around F<ev.h> that you can include instead and
		3514	where you can put other configuration options):
		3515
		3516	#define EV_STANDALONE 1
		3517	#include "ev.h"
		3518
		3519	Both header files and implementation files can be compiled with a C++
		3520	compiler (at least, that's a stated goal, and breakage will be treated
		3521	as a bug).
		3522
		3523	You need the following files in your source tree, or in a directory
		3524	in your include path (e.g. in libev/ when using -Ilibev):
		3525
		3526	ev.h
		3527	ev.c
		3528	ev_vars.h
		3529	ev_wrap.h
		3530
		3531	ev_win32.c required on win32 platforms only
		3532
		3533	ev_select.c only when select backend is enabled (which is enabled by default)
		3534	ev_poll.c only when poll backend is enabled (disabled by default)
		3535	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3536	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3537	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3538
		3539	F<ev.c> includes the backend files directly when enabled, so you only need
		3540	to compile this single file.
		3541
		3542	=head3 LIBEVENT COMPATIBILITY API
		3543
		3544	To include the libevent compatibility API, also include:
		3545
		3546	#include "event.c"
		3547
		3548	in the file including F<ev.c>, and:
		3549
		3550	#include "event.h"
		3551
		3552	in the files that want to use the libevent API. This also includes F<ev.h>.
		3553
		3554	You need the following additional files for this:
		3555
		3556	event.h
		3557	event.c
		3558
		3559	=head3 AUTOCONF SUPPORT
		3560
		3561	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3562	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3563	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3564	include F<config.h> and configure itself accordingly.
		3565
		3566	For this of course you need the m4 file:
		3567
		3568	libev.m4
		3569
		3570	=head2 PREPROCESSOR SYMBOLS/MACROS
		3571
		3572	Libev can be configured via a variety of preprocessor symbols you have to
		3573	define before including any of its files. The default in the absence of
		3574	autoconf is documented for every option.
		3575
		3576	=over 4
		3577
		3578	=item EV_STANDALONE
		3579
		3580	Must always be C<1> if you do not use autoconf configuration, which
		3581	keeps libev from including F<config.h>, and it also defines dummy
		3582	implementations for some libevent functions (such as logging, which is not
		3583	supported). It will also not define any of the structs usually found in
		3584	F<event.h> that are not directly supported by the libev core alone.
		3585
		3586	In stanbdalone mode, libev will still try to automatically deduce the
		3587	configuration, but has to be more conservative.
		3588
		3589	=item EV_USE_MONOTONIC
		3590
		3591	If defined to be C<1>, libev will try to detect the availability of the
		3592	monotonic clock option at both compile time and runtime. Otherwise no
		3593	use of the monotonic clock option will be attempted. If you enable this,
		3594	you usually have to link against librt or something similar. Enabling it
		3595	when the functionality isn't available is safe, though, although you have
		3596	to make sure you link against any libraries where the C<clock_gettime>
		3597	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		3598
		3599	=item EV_USE_REALTIME
		3600
		3601	If defined to be C<1>, libev will try to detect the availability of the
		3602	real-time clock option at compile time (and assume its availability
		3603	at runtime if successful). Otherwise no use of the real-time clock
		3604	option will be attempted. This effectively replaces C<gettimeofday>
		3605	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		3606	correctness. See the note about libraries in the description of
		3607	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3608	C<EV_USE_CLOCK_SYSCALL>.
		3609
		3610	=item EV_USE_CLOCK_SYSCALL
		3611
		3612	If defined to be C<1>, libev will try to use a direct syscall instead
		3613	of calling the system-provided C<clock_gettime> function. This option
		3614	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3615	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3616	programs needlessly. Using a direct syscall is slightly slower (in
		3617	theory), because no optimised vdso implementation can be used, but avoids
		3618	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3619	higher, as it simplifies linking (no need for C<-lrt>).
		3620
		3621	=item EV_USE_NANOSLEEP
		3622
		3623	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3624	and will use it for delays. Otherwise it will use C<select ()>.
		3625
		3626	=item EV_USE_EVENTFD
		3627
		3628	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3629	available and will probe for kernel support at runtime. This will improve
		3630	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3631	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3632	2.7 or newer, otherwise disabled.
		3633
		3634	=item EV_USE_SELECT
		3635
		3636	If undefined or defined to be C<1>, libev will compile in support for the
		3637	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3638	other method takes over, select will be it. Otherwise the select backend
		3639	will not be compiled in.
		3640
		3641	=item EV_SELECT_USE_FD_SET
		3642
		3643	If defined to C<1>, then the select backend will use the system C<fd_set>
		3644	structure. This is useful if libev doesn't compile due to a missing
		3645	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		3646	on exotic systems. This usually limits the range of file descriptors to
		3647	some low limit such as 1024 or might have other limitations (winsocket
		3648	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		3649	configures the maximum size of the C<fd_set>.
		3650
		3651	=item EV_SELECT_IS_WINSOCKET
		3652
		3653	When defined to C<1>, the select backend will assume that
		3654	select/socket/connect etc. don't understand file descriptors but
		3655	wants osf handles on win32 (this is the case when the select to
		3656	be used is the winsock select). This means that it will call
		3657	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3658	it is assumed that all these functions actually work on fds, even
		3659	on win32. Should not be defined on non-win32 platforms.
		3660
		3661	=item EV_FD_TO_WIN32_HANDLE
		3662
		3663	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3664	file descriptors to socket handles. When not defining this symbol (the
		3665	default), then libev will call C<_get_osfhandle>, which is usually
		3666	correct. In some cases, programs use their own file descriptor management,
		3667	in which case they can provide this function to map fds to socket handles.
		3668
		3669	=item EV_USE_POLL
		3670
		3671	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3672	backend. Otherwise it will be enabled on non-win32 platforms. It
		3673	takes precedence over select.
		3674
		3675	=item EV_USE_EPOLL
		3676
		3677	If defined to be C<1>, libev will compile in support for the Linux
		3678	C<epoll>(7) backend. Its availability will be detected at runtime,
		3679	otherwise another method will be used as fallback. This is the preferred
		3680	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3681	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3682
		3683	=item EV_USE_KQUEUE
		3684
		3685	If defined to be C<1>, libev will compile in support for the BSD style
		3686	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3687	otherwise another method will be used as fallback. This is the preferred
		3688	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3689	supports some types of fds correctly (the only platform we found that
		3690	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3691	not be used unless explicitly requested. The best way to use it is to find
		3692	out whether kqueue supports your type of fd properly and use an embedded
		3693	kqueue loop.
		3694
		3695	=item EV_USE_PORT
		3696
		3697	If defined to be C<1>, libev will compile in support for the Solaris
		3698	10 port style backend. Its availability will be detected at runtime,
		3699	otherwise another method will be used as fallback. This is the preferred
		3700	backend for Solaris 10 systems.
		3701
		3702	=item EV_USE_DEVPOLL
		3703
		3704	Reserved for future expansion, works like the USE symbols above.
		3705
		3706	=item EV_USE_INOTIFY
		3707
		3708	If defined to be C<1>, libev will compile in support for the Linux inotify
		3709	interface to speed up C<ev_stat> watchers. Its actual availability will
		3710	be detected at runtime. If undefined, it will be enabled if the headers
		3711	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3712
		3713	=item EV_ATOMIC_T
		3714
		3715	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3716	access is atomic with respect to other threads or signal contexts. No such
		3717	type is easily found in the C language, so you can provide your own type
		3718	that you know is safe for your purposes. It is used both for signal handler "locking"
		3719	as well as for signal and thread safety in C<ev_async> watchers.
		3720
		3721	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3722	(from F<signal.h>), which is usually good enough on most platforms.
		3723
		3724	=item EV_H
		3725
		3726	The name of the F<ev.h> header file used to include it. The default if
		3727	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3728	used to virtually rename the F<ev.h> header file in case of conflicts.
		3729
		3730	=item EV_CONFIG_H
		3731
		3732	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3733	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3734	C<EV_H>, above.
		3735
		3736	=item EV_EVENT_H
		3737
		3738	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3739	of how the F<event.h> header can be found, the default is C<"event.h">.
		3740
		3741	=item EV_PROTOTYPES
		3742
		3743	If defined to be C<0>, then F<ev.h> will not define any function
		3744	prototypes, but still define all the structs and other symbols. This is
		3745	occasionally useful if you want to provide your own wrapper functions
		3746	around libev functions.
		3747
		3748	=item EV_MULTIPLICITY
		3749
		3750	If undefined or defined to C<1>, then all event-loop-specific functions
		3751	will have the C<struct ev_loop *> as first argument, and you can create
		3752	additional independent event loops. Otherwise there will be no support
		3753	for multiple event loops and there is no first event loop pointer
		3754	argument. Instead, all functions act on the single default loop.
		3755
		3756	=item EV_MINPRI
		3757
		3758	=item EV_MAXPRI
		3759
		3760	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3761	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3762	provide for more priorities by overriding those symbols (usually defined
		3763	to be C<-2> and C<2>, respectively).
		3764
		3765	When doing priority-based operations, libev usually has to linearly search
		3766	all the priorities, so having many of them (hundreds) uses a lot of space
		3767	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3768	fine.
		3769
		3770	If your embedding application does not need any priorities, defining these
		3771	both to C<0> will save some memory and CPU.
		3772
		3773	=item EV_PERIODIC_ENABLE
		3774
		3775	If undefined or defined to be C<1>, then periodic timers are supported. If
		3776	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3777	code.
		3778
		3779	=item EV_IDLE_ENABLE
		3780
		3781	If undefined or defined to be C<1>, then idle watchers are supported. If
		3782	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3783	code.
		3784
		3785	=item EV_EMBED_ENABLE
		3786
		3787	If undefined or defined to be C<1>, then embed watchers are supported. If
		3788	defined to be C<0>, then they are not. Embed watchers rely on most other
		3789	watcher types, which therefore must not be disabled.
		3790
		3791	=item EV_STAT_ENABLE
		3792
		3793	If undefined or defined to be C<1>, then stat watchers are supported. If
		3794	defined to be C<0>, then they are not.
		3795
		3796	=item EV_FORK_ENABLE
		3797
		3798	If undefined or defined to be C<1>, then fork watchers are supported. If
		3799	defined to be C<0>, then they are not.
		3800
		3801	=item EV_ASYNC_ENABLE
		3802
		3803	If undefined or defined to be C<1>, then async watchers are supported. If
		3804	defined to be C<0>, then they are not.
		3805
		3806	=item EV_MINIMAL
		3807
		3808	If you need to shave off some kilobytes of code at the expense of some
		3809	speed (but with the full API), define this symbol to C<1>. Currently this
		3810	is used to override some inlining decisions, saves roughly 30% code size
		3811	on amd64. It also selects a much smaller 2-heap for timer management over
		3812	the default 4-heap.
		3813
		3814	You can save even more by disabling watcher types you do not need
		3815	and setting C<EV_MAXPRI> == C<EV_MINPRI>. Also, disabling C<assert>
		3816	(C<-DNDEBUG>) will usually reduce code size a lot.
		3817
		3818	Defining C<EV_MINIMAL> to C<2> will additionally reduce the core API to
		3819	provide a bare-bones event library. See C<ev.h> for details on what parts
		3820	of the API are still available, and do not complain if this subset changes
		3821	over time.
		3822
		3823	=item EV_NSIG
		3824
		3825	The highest supported signal number, +1 (or, the number of
		3826	signals): Normally, libev tries to deduce the maximum number of signals
		3827	automatically, but sometimes this fails, in which case it can be
		3828	specified. Also, using a lower number than detected (C<32> should be
		3829	good for about any system in existance) can save some memory, as libev
		3830	statically allocates some 12-24 bytes per signal number.
		3831
		3832	=item EV_PID_HASHSIZE
		3833
		3834	C<ev_child> watchers use a small hash table to distribute workload by
		3835	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3836	than enough. If you need to manage thousands of children you might want to
		3837	increase this value (I<must> be a power of two).
		3838
		3839	=item EV_INOTIFY_HASHSIZE
		3840
		3841	C<ev_stat> watchers use a small hash table to distribute workload by
		3842	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3843	usually more than enough. If you need to manage thousands of C<ev_stat>
		3844	watchers you might want to increase this value (I<must> be a power of
		3845	two).
		3846
		3847	=item EV_USE_4HEAP
		3848
		3849	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3850	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3851	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3852	faster performance with many (thousands) of watchers.
		3853
		3854	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3855	(disabled).
		3856
		3857	=item EV_HEAP_CACHE_AT
		3858
		3859	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3860	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3861	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3862	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3863	but avoids random read accesses on heap changes. This improves performance
		3864	noticeably with many (hundreds) of watchers.
		3865
		3866	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3867	(disabled).
		3868
		3869	=item EV_VERIFY
		3870
		3871	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3872	be done: If set to C<0>, no internal verification code will be compiled
		3873	in. If set to C<1>, then verification code will be compiled in, but not
		3874	called. If set to C<2>, then the internal verification code will be
		3875	called once per loop, which can slow down libev. If set to C<3>, then the
		3876	verification code will be called very frequently, which will slow down
		3877	libev considerably.
		3878
		3879	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3880	C<0>.
		3881
		3882	=item EV_COMMON
		3883
		3884	By default, all watchers have a C<void *data> member. By redefining
		3885	this macro to a something else you can include more and other types of
		3886	members. You have to define it each time you include one of the files,
		3887	though, and it must be identical each time.
		3888
		3889	For example, the perl EV module uses something like this:
		3890
		3891	#define EV_COMMON \
		3892	SV *self; /* contains this struct */ \
		3893	SV *cb_sv, *fh /* note no trailing ";" */
		3894
		3895	=item EV_CB_DECLARE (type)
		3896
		3897	=item EV_CB_INVOKE (watcher, revents)
		3898
		3899	=item ev_set_cb (ev, cb)
		3900
		3901	Can be used to change the callback member declaration in each watcher,
		3902	and the way callbacks are invoked and set. Must expand to a struct member
		3903	definition and a statement, respectively. See the F<ev.h> header file for
		3904	their default definitions. One possible use for overriding these is to
		3905	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3906	method calls instead of plain function calls in C++.
		3907
		3908	=back
		3909
		3910	=head2 EXPORTED API SYMBOLS
		3911
		3912	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3913	exported symbols, you can use the provided F<Symbol.*> files which list
		3914	all public symbols, one per line:
		3915
		3916	Symbols.ev for libev proper
		3917	Symbols.event for the libevent emulation
		3918
		3919	This can also be used to rename all public symbols to avoid clashes with
		3920	multiple versions of libev linked together (which is obviously bad in
		3921	itself, but sometimes it is inconvenient to avoid this).
		3922
		3923	A sed command like this will create wrapper C<#define>'s that you need to
		3924	include before including F<ev.h>:
		3925
		3926	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3927
		3928	This would create a file F<wrap.h> which essentially looks like this:
		3929
		3930	#define ev_backend myprefix_ev_backend
		3931	#define ev_check_start myprefix_ev_check_start
		3932	#define ev_check_stop myprefix_ev_check_stop
		3933	...
		3934
		3935	=head2 EXAMPLES
		3936
		3937	For a real-world example of a program the includes libev
		3938	verbatim, you can have a look at the EV perl module
		3939	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3940	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3941	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3942	will be compiled. It is pretty complex because it provides its own header
		3943	file.
		3944
		3945	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3946	that everybody includes and which overrides some configure choices:
		3947
		3948	#define EV_MINIMAL 1
		3949	#define EV_USE_POLL 0
		3950	#define EV_MULTIPLICITY 0
		3951	#define EV_PERIODIC_ENABLE 0
		3952	#define EV_STAT_ENABLE 0
		3953	#define EV_FORK_ENABLE 0
		3954	#define EV_CONFIG_H <config.h>
		3955	#define EV_MINPRI 0
		3956	#define EV_MAXPRI 0
		3957
		3958	#include "ev++.h"
		3959
		3960	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3961
		3962	#include "ev_cpp.h"
		3963	#include "ev.c"
		3964
		3965	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3966
		3967	=head2 THREADS AND COROUTINES
		3968
		3969	=head3 THREADS
		3970
		3971	All libev functions are reentrant and thread-safe unless explicitly
		3972	documented otherwise, but libev implements no locking itself. This means
		3973	that you can use as many loops as you want in parallel, as long as there
		3974	are no concurrent calls into any libev function with the same loop
		3975	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3976	of course): libev guarantees that different event loops share no data
		3977	structures that need any locking.
		3978
		3979	Or to put it differently: calls with different loop parameters can be done
		3980	concurrently from multiple threads, calls with the same loop parameter
		3981	must be done serially (but can be done from different threads, as long as
		3982	only one thread ever is inside a call at any point in time, e.g. by using
		3983	a mutex per loop).
		3984
		3985	Specifically to support threads (and signal handlers), libev implements
		3986	so-called C<ev_async> watchers, which allow some limited form of
		3987	concurrency on the same event loop, namely waking it up "from the
		3988	outside".
		3989
		3990	If you want to know which design (one loop, locking, or multiple loops
		3991	without or something else still) is best for your problem, then I cannot
		3992	help you, but here is some generic advice:
		3993
		3994	=over 4
		3995
		3996	=item * most applications have a main thread: use the default libev loop
		3997	in that thread, or create a separate thread running only the default loop.
		3998
		3999	This helps integrating other libraries or software modules that use libev
		4000	themselves and don't care/know about threading.
		4001
		4002	=item * one loop per thread is usually a good model.
		4003
		4004	Doing this is almost never wrong, sometimes a better-performance model
		4005	exists, but it is always a good start.
		4006
		4007	=item * other models exist, such as the leader/follower pattern, where one
		4008	loop is handed through multiple threads in a kind of round-robin fashion.
		4009
		4010	Choosing a model is hard - look around, learn, know that usually you can do
		4011	better than you currently do :-)
		4012
		4013	=item * often you need to talk to some other thread which blocks in the
		4014	event loop.
		4015
		4016	C<ev_async> watchers can be used to wake them up from other threads safely
		4017	(or from signal contexts...).
		4018
		4019	An example use would be to communicate signals or other events that only
		4020	work in the default loop by registering the signal watcher with the
		4021	default loop and triggering an C<ev_async> watcher from the default loop
		4022	watcher callback into the event loop interested in the signal.
		4023
		4024	=back
		4025
		4026	=head4 THREAD LOCKING EXAMPLE
		4027
		4028	Here is a fictitious example of how to run an event loop in a different
		4029	thread than where callbacks are being invoked and watchers are
		4030	created/added/removed.
		4031
		4032	For a real-world example, see the C<EV::Loop::Async> perl module,
		4033	which uses exactly this technique (which is suited for many high-level
		4034	languages).
		4035
		4036	The example uses a pthread mutex to protect the loop data, a condition
		4037	variable to wait for callback invocations, an async watcher to notify the
		4038	event loop thread and an unspecified mechanism to wake up the main thread.
		4039
		4040	First, you need to associate some data with the event loop:
		4041
		4042	typedef struct {
		4043	mutex_t lock; /* global loop lock */
		4044	ev_async async_w;
		4045	thread_t tid;
		4046	cond_t invoke_cv;
		4047	} userdata;
		4048
		4049	void prepare_loop (EV_P)
		4050	{
		4051	// for simplicity, we use a static userdata struct.
		4052	static userdata u;
		4053
		4054	ev_async_init (&u->async_w, async_cb);
		4055	ev_async_start (EV_A_ &u->async_w);
		4056
		4057	pthread_mutex_init (&u->lock, 0);
		4058	pthread_cond_init (&u->invoke_cv, 0);
		4059
		4060	// now associate this with the loop
		4061	ev_set_userdata (EV_A_ u);
		4062	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		4063	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		4064
		4065	// then create the thread running ev_loop
		4066	pthread_create (&u->tid, 0, l_run, EV_A);
		4067	}
		4068
		4069	The callback for the C<ev_async> watcher does nothing: the watcher is used
		4070	solely to wake up the event loop so it takes notice of any new watchers
		4071	that might have been added:
		4072
		4073	static void
		4074	async_cb (EV_P_ ev_async *w, int revents)
		4075	{
		4076	// just used for the side effects
		4077	}
		4078
		4079	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		4080	protecting the loop data, respectively.
		4081
		4082	static void
		4083	l_release (EV_P)
		4084	{
		4085	userdata *u = ev_userdata (EV_A);
		4086	pthread_mutex_unlock (&u->lock);
		4087	}
		4088
		4089	static void
		4090	l_acquire (EV_P)
		4091	{
		4092	userdata *u = ev_userdata (EV_A);
		4093	pthread_mutex_lock (&u->lock);
		4094	}
		4095
		4096	The event loop thread first acquires the mutex, and then jumps straight
		4097	into C<ev_loop>:
		4098
		4099	void *
		4100	l_run (void *thr_arg)
		4101	{
		4102	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		4103
		4104	l_acquire (EV_A);
		4105	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		4106	ev_loop (EV_A_ 0);
		4107	l_release (EV_A);
		4108
		4109	return 0;
		4110	}
		4111
		4112	Instead of invoking all pending watchers, the C<l_invoke> callback will
		4113	signal the main thread via some unspecified mechanism (signals? pipe
		4114	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		4115	have been called (in a while loop because a) spurious wakeups are possible
		4116	and b) skipping inter-thread-communication when there are no pending
		4117	watchers is very beneficial):
		4118
		4119	static void
		4120	l_invoke (EV_P)
		4121	{
		4122	userdata *u = ev_userdata (EV_A);
		4123
		4124	while (ev_pending_count (EV_A))
		4125	{
		4126	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		4127	pthread_cond_wait (&u->invoke_cv, &u->lock);
		4128	}
		4129	}
		4130
		4131	Now, whenever the main thread gets told to invoke pending watchers, it
		4132	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		4133	thread to continue:
		4134
		4135	static void
		4136	real_invoke_pending (EV_P)
		4137	{
		4138	userdata *u = ev_userdata (EV_A);
		4139
		4140	pthread_mutex_lock (&u->lock);
		4141	ev_invoke_pending (EV_A);
		4142	pthread_cond_signal (&u->invoke_cv);
		4143	pthread_mutex_unlock (&u->lock);
		4144	}
		4145
		4146	Whenever you want to start/stop a watcher or do other modifications to an
		4147	event loop, you will now have to lock:
		4148
		4149	ev_timer timeout_watcher;
		4150	userdata *u = ev_userdata (EV_A);
		4151
		4152	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		4153
		4154	pthread_mutex_lock (&u->lock);
		4155	ev_timer_start (EV_A_ &timeout_watcher);
		4156	ev_async_send (EV_A_ &u->async_w);
		4157	pthread_mutex_unlock (&u->lock);
		4158
		4159	Note that sending the C<ev_async> watcher is required because otherwise
		4160	an event loop currently blocking in the kernel will have no knowledge
		4161	about the newly added timer. By waking up the loop it will pick up any new
		4162	watchers in the next event loop iteration.
		4163
		4164	=head3 COROUTINES
		4165
		4166	Libev is very accommodating to coroutines ("cooperative threads"):
		4167	libev fully supports nesting calls to its functions from different
		4168	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		4169	different coroutines, and switch freely between both coroutines running
		4170	the loop, as long as you don't confuse yourself). The only exception is
		4171	that you must not do this from C<ev_periodic> reschedule callbacks.
		4172
		4173	Care has been taken to ensure that libev does not keep local state inside
		4174	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		4175	they do not call any callbacks.
		4176
		4177	=head2 COMPILER WARNINGS
		4178
		4179	Depending on your compiler and compiler settings, you might get no or a
		4180	lot of warnings when compiling libev code. Some people are apparently
		4181	scared by this.
		4182
		4183	However, these are unavoidable for many reasons. For one, each compiler
		4184	has different warnings, and each user has different tastes regarding
		4185	warning options. "Warn-free" code therefore cannot be a goal except when
		4186	targeting a specific compiler and compiler-version.
		4187
		4188	Another reason is that some compiler warnings require elaborate
		4189	workarounds, or other changes to the code that make it less clear and less
		4190	maintainable.
		4191
		4192	And of course, some compiler warnings are just plain stupid, or simply
		4193	wrong (because they don't actually warn about the condition their message
		4194	seems to warn about). For example, certain older gcc versions had some
		4195	warnings that resulted an extreme number of false positives. These have
		4196	been fixed, but some people still insist on making code warn-free with
		4197	such buggy versions.
		4198
		4199	While libev is written to generate as few warnings as possible,
		4200	"warn-free" code is not a goal, and it is recommended not to build libev
		4201	with any compiler warnings enabled unless you are prepared to cope with
		4202	them (e.g. by ignoring them). Remember that warnings are just that:
		4203	warnings, not errors, or proof of bugs.
		4204
		4205
		4206	=head2 VALGRIND
		4207
		4208	Valgrind has a special section here because it is a popular tool that is
		4209	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		4210
		4211	If you think you found a bug (memory leak, uninitialised data access etc.)
		4212	in libev, then check twice: If valgrind reports something like:
		4213
		4214	==2274== definitely lost: 0 bytes in 0 blocks.
		4215	==2274== possibly lost: 0 bytes in 0 blocks.
		4216	==2274== still reachable: 256 bytes in 1 blocks.
		4217
		4218	Then there is no memory leak, just as memory accounted to global variables
		4219	is not a memleak - the memory is still being referenced, and didn't leak.
		4220
		4221	Similarly, under some circumstances, valgrind might report kernel bugs
		4222	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		4223	although an acceptable workaround has been found here), or it might be
		4224	confused.
		4225
		4226	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		4227	make it into some kind of religion.
		4228
		4229	If you are unsure about something, feel free to contact the mailing list
		4230	with the full valgrind report and an explanation on why you think this
		4231	is a bug in libev (best check the archives, too :). However, don't be
		4232	annoyed when you get a brisk "this is no bug" answer and take the chance
		4233	of learning how to interpret valgrind properly.
		4234
		4235	If you need, for some reason, empty reports from valgrind for your project
		4236	I suggest using suppression lists.
		4237
		4238
		4239	=head1 PORTABILITY NOTES
		4240
		4241	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		4242
		4243	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		4244	requires, and its I/O model is fundamentally incompatible with the POSIX
		4245	model. Libev still offers limited functionality on this platform in
		4246	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		4247	descriptors. This only applies when using Win32 natively, not when using
		4248	e.g. cygwin.
		4249
		4250	Lifting these limitations would basically require the full
		4251	re-implementation of the I/O system. If you are into these kinds of
		4252	things, then note that glib does exactly that for you in a very portable
		4253	way (note also that glib is the slowest event library known to man).
		4254
		4255	There is no supported compilation method available on windows except
		4256	embedding it into other applications.
		4257
		4258	Sensible signal handling is officially unsupported by Microsoft - libev
		4259	tries its best, but under most conditions, signals will simply not work.
		4260
		4261	Not a libev limitation but worth mentioning: windows apparently doesn't
		4262	accept large writes: instead of resulting in a partial write, windows will
		4263	either accept everything or return C<ENOBUFS> if the buffer is too large,
		4264	so make sure you only write small amounts into your sockets (less than a
		4265	megabyte seems safe, but this apparently depends on the amount of memory
		4266	available).
		4267
		4268	Due to the many, low, and arbitrary limits on the win32 platform and
		4269	the abysmal performance of winsockets, using a large number of sockets
		4270	is not recommended (and not reasonable). If your program needs to use
		4271	more than a hundred or so sockets, then likely it needs to use a totally
		4272	different implementation for windows, as libev offers the POSIX readiness
		4273	notification model, which cannot be implemented efficiently on windows
		4274	(due to Microsoft monopoly games).
		4275
		4276	A typical way to use libev under windows is to embed it (see the embedding
		4277	section for details) and use the following F<evwrap.h> header file instead
		4278	of F<ev.h>:
		4279
		4280	#define EV_STANDALONE /* keeps ev from requiring config.h */
		4281	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		4282
		4283	#include "ev.h"
		4284
		4285	And compile the following F<evwrap.c> file into your project (make sure
		4286	you do I<not> compile the F<ev.c> or any other embedded source files!):
		4287
		4288	#include "evwrap.h"
		4289	#include "ev.c"
		4290
		4291	=over 4
		4292
		4293	=item The winsocket select function
		4294
		4295	The winsocket C<select> function doesn't follow POSIX in that it
		4296	requires socket I<handles> and not socket I<file descriptors> (it is
		4297	also extremely buggy). This makes select very inefficient, and also
		4298	requires a mapping from file descriptors to socket handles (the Microsoft
		4299	C runtime provides the function C<_open_osfhandle> for this). See the
		4300	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		4301	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		4302
		4303	The configuration for a "naked" win32 using the Microsoft runtime
		4304	libraries and raw winsocket select is:
		4305
		4306	#define EV_USE_SELECT 1
		4307	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		4308
		4309	Note that winsockets handling of fd sets is O(n), so you can easily get a
		4310	complexity in the O(n²) range when using win32.
		4311
		4312	=item Limited number of file descriptors
		4313
		4314	Windows has numerous arbitrary (and low) limits on things.
		4315
		4316	Early versions of winsocket's select only supported waiting for a maximum
		4317	of C<64> handles (probably owning to the fact that all windows kernels
		4318	can only wait for C<64> things at the same time internally; Microsoft
		4319	recommends spawning a chain of threads and wait for 63 handles and the
		4320	previous thread in each. Sounds great!).
		4321
		4322	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		4323	to some high number (e.g. C<2048>) before compiling the winsocket select
		4324	call (which might be in libev or elsewhere, for example, perl and many
		4325	other interpreters do their own select emulation on windows).
		4326
		4327	Another limit is the number of file descriptors in the Microsoft runtime
		4328	libraries, which by default is C<64> (there must be a hidden I<64>
		4329	fetish or something like this inside Microsoft). You can increase this
		4330	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		4331	(another arbitrary limit), but is broken in many versions of the Microsoft
		4332	runtime libraries. This might get you to about C<512> or C<2048> sockets
		4333	(depending on windows version and/or the phase of the moon). To get more,
		4334	you need to wrap all I/O functions and provide your own fd management, but
		4335	the cost of calling select (O(n²)) will likely make this unworkable.
		4336
		4337	=back
		4338
		4339	=head2 PORTABILITY REQUIREMENTS
		4340
		4341	In addition to a working ISO-C implementation and of course the
		4342	backend-specific APIs, libev relies on a few additional extensions:
		4343
		4344	=over 4
		4345
		4346	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		4347	calling conventions regardless of C<ev_watcher_type *>.
		4348
		4349	Libev assumes not only that all watcher pointers have the same internal
		4350	structure (guaranteed by POSIX but not by ISO C for example), but it also
		4351	assumes that the same (machine) code can be used to call any watcher
		4352	callback: The watcher callbacks have different type signatures, but libev
		4353	calls them using an C<ev_watcher *> internally.
		4354
		4355	=item C<sig_atomic_t volatile> must be thread-atomic as well
		4356
		4357	The type C<sig_atomic_t volatile> (or whatever is defined as
		4358	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		4359	threads. This is not part of the specification for C<sig_atomic_t>, but is
		4360	believed to be sufficiently portable.
		4361
		4362	=item C<sigprocmask> must work in a threaded environment
		4363
		4364	Libev uses C<sigprocmask> to temporarily block signals. This is not
		4365	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		4366	pthread implementations will either allow C<sigprocmask> in the "main
		4367	thread" or will block signals process-wide, both behaviours would
		4368	be compatible with libev. Interaction between C<sigprocmask> and
		4369	C<pthread_sigmask> could complicate things, however.
		4370
		4371	The most portable way to handle signals is to block signals in all threads
		4372	except the initial one, and run the default loop in the initial thread as
		4373	well.
		4374
		4375	=item C<long> must be large enough for common memory allocation sizes
		4376
		4377	To improve portability and simplify its API, libev uses C<long> internally
		4378	instead of C<size_t> when allocating its data structures. On non-POSIX
		4379	systems (Microsoft...) this might be unexpectedly low, but is still at
		4380	least 31 bits everywhere, which is enough for hundreds of millions of
		4381	watchers.
		4382
		4383	=item C<double> must hold a time value in seconds with enough accuracy
		4384
		4385	The type C<double> is used to represent timestamps. It is required to
		4386	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		4387	enough for at least into the year 4000. This requirement is fulfilled by
		4388	implementations implementing IEEE 754, which is basically all existing
		4389	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4390	2200.
		4391
		4392	=back
		4393
		4394	If you know of other additional requirements drop me a note.
		4395
		4396
		4397	=head1 ALGORITHMIC COMPLEXITIES
		4398
		4399	In this section the complexities of (many of) the algorithms used inside
		4400	libev will be documented. For complexity discussions about backends see
		4401	the documentation for C<ev_default_init>.
		4402
		4403	All of the following are about amortised time: If an array needs to be
		4404	extended, libev needs to realloc and move the whole array, but this
		4405	happens asymptotically rarer with higher number of elements, so O(1) might
		4406	mean that libev does a lengthy realloc operation in rare cases, but on
		4407	average it is much faster and asymptotically approaches constant time.
		4408
		4409	=over 4
		4410
		4411	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		4412
		4413	This means that, when you have a watcher that triggers in one hour and
		4414	there are 100 watchers that would trigger before that, then inserting will
		4415	have to skip roughly seven (C<ld 100>) of these watchers.
		4416
		4417	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		4418
		4419	That means that changing a timer costs less than removing/adding them,
		4420	as only the relative motion in the event queue has to be paid for.
		4421
		4422	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		4423
		4424	These just add the watcher into an array or at the head of a list.
		4425
		4426	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		4427
		4428	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		4429
		4430	These watchers are stored in lists, so they need to be walked to find the
		4431	correct watcher to remove. The lists are usually short (you don't usually
		4432	have many watchers waiting for the same fd or signal: one is typical, two
		4433	is rare).
		4434
		4435	=item Finding the next timer in each loop iteration: O(1)
		4436
		4437	By virtue of using a binary or 4-heap, the next timer is always found at a
		4438	fixed position in the storage array.
		4439
		4440	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		4441
		4442	A change means an I/O watcher gets started or stopped, which requires
		4443	libev to recalculate its status (and possibly tell the kernel, depending
		4444	on backend and whether C<ev_io_set> was used).
		4445
		4446	=item Activating one watcher (putting it into the pending state): O(1)
		4447
		4448	=item Priority handling: O(number_of_priorities)
		4449
		4450	Priorities are implemented by allocating some space for each
		4451	priority. When doing priority-based operations, libev usually has to
		4452	linearly search all the priorities, but starting/stopping and activating
		4453	watchers becomes O(1) with respect to priority handling.
		4454
		4455	=item Sending an ev_async: O(1)
		4456
		4457	=item Processing ev_async_send: O(number_of_async_watchers)
		4458
		4459	=item Processing signals: O(max_signal_number)
		4460
		4461	Sending involves a system call I<iff> there were no other C<ev_async_send>
		4462	calls in the current loop iteration. Checking for async and signal events
		4463	involves iterating over all running async watchers or all signal numbers.
		4464
		4465	=back
		4466
		4467
		4468	=head1 GLOSSARY
		4469
		4470	=over 4
		4471
		4472	=item active
		4473
		4474	A watcher is active as long as it has been started (has been attached to
		4475	an event loop) but not yet stopped (disassociated from the event loop).
		4476
		4477	=item application
		4478
		4479	In this document, an application is whatever is using libev.
		4480
		4481	=item callback
		4482
		4483	The address of a function that is called when some event has been
		4484	detected. Callbacks are being passed the event loop, the watcher that
		4485	received the event, and the actual event bitset.
		4486
		4487	=item callback invocation
		4488
		4489	The act of calling the callback associated with a watcher.
		4490
		4491	=item event
		4492
		4493	A change of state of some external event, such as data now being available
		4494	for reading on a file descriptor, time having passed or simply not having
		4495	any other events happening anymore.
		4496
		4497	In libev, events are represented as single bits (such as C<EV_READ> or
		4498	C<EV_TIMEOUT>).
		4499
		4500	=item event library
		4501
		4502	A software package implementing an event model and loop.
		4503
		4504	=item event loop
		4505
		4506	An entity that handles and processes external events and converts them
		4507	into callback invocations.
		4508
		4509	=item event model
		4510
		4511	The model used to describe how an event loop handles and processes
		4512	watchers and events.
		4513
		4514	=item pending
		4515
		4516	A watcher is pending as soon as the corresponding event has been detected,
		4517	and stops being pending as soon as the watcher will be invoked or its
		4518	pending status is explicitly cleared by the application.
		4519
		4520	A watcher can be pending, but not active. Stopping a watcher also clears
		4521	its pending status.
		4522
		4523	=item real time
		4524
		4525	The physical time that is observed. It is apparently strictly monotonic :)
		4526
		4527	=item wall-clock time
		4528
		4529	The time and date as shown on clocks. Unlike real time, it can actually
		4530	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4531	clock.
		4532
		4533	=item watcher
		4534
		4535	A data structure that describes interest in certain events. Watchers need
		4536	to be started (attached to an event loop) before they can receive events.
		4537
		4538	=item watcher invocation
		4539
		4540	The act of calling the callback associated with a watcher.
		4541
		4542	=back
		4543
742	=head1 AUTHOR	4544	=head1 AUTHOR
743		4545
744	Marc Lehmann <libev@schmorp.de>.	4546	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
745		4547

Diff Legend

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