ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.16 by root, Mon Nov 12 08:47:14 2007 UTC vs.
Revision 1.141 by root, Wed Apr 2 19:19:33 2008 UTC

…		…
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
8		8
		9	=head2 EXAMPLE PROGRAM
		10
		11	// a single header file is required
		12	#include <ev.h>
		13
		14	// every watcher type has its own typedef'd struct
		15	// with the name ev_<type>
		16	ev_io stdin_watcher;
		17	ev_timer timeout_watcher;
		18
		19	// all watcher callbacks have a similar signature
		20	// this callback is called when data is readable on stdin
		21	static void
		22	stdin_cb (EV_P_ struct ev_io *w, int revents)
		23	{
		24	puts ("stdin ready");
		25	// for one-shot events, one must manually stop the watcher
		26	// with its corresponding stop function.
		27	ev_io_stop (EV_A_ w);
		28
		29	// this causes all nested ev_loop's to stop iterating
		30	ev_unloop (EV_A_ EVUNLOOP_ALL);
		31	}
		32
		33	// another callback, this time for a time-out
		34	static void
		35	timeout_cb (EV_P_ struct ev_timer *w, int revents)
		36	{
		37	puts ("timeout");
		38	// this causes the innermost ev_loop to stop iterating
		39	ev_unloop (EV_A_ EVUNLOOP_ONE);
		40	}
		41
		42	int
		43	main (void)
		44	{
		45	// use the default event loop unless you have special needs
		46	struct ev_loop *loop = ev_default_loop (0);
		47
		48	// initialise an io watcher, then start it
		49	// this one will watch for stdin to become readable
		50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		51	ev_io_start (loop, &stdin_watcher);
		52
		53	// initialise a timer watcher, then start it
		54	// simple non-repeating 5.5 second timeout
		55	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		56	ev_timer_start (loop, &timeout_watcher);
		57
		58	// now wait for events to arrive
		59	ev_loop (loop, 0);
		60
		61	// unloop was called, so exit
		62	return 0;
		63	}
		64
9	=head1 DESCRIPTION	65	=head1 DESCRIPTION
10		66
		67	The newest version of this document is also available as an html-formatted
		68	web page you might find easier to navigate when reading it for the first
		69	time: L<http://cvs.schmorp.de/libev/ev.html>.
		70
11	Libev is an event loop: you register interest in certain events (such as a	71	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	72	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	73	these event sources and provide your program with events.
14		74
15	To do this, it must take more or less complete control over your process	75	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	76	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	77	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	79	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	80	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	81	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	82	watcher.
23		83
24	=head1 FEATURES	84	=head2 FEATURES
25		85
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	86	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	87	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	88	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
29	events (related to SIGCHLD), and event watchers dealing with the event	89	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	90	with customised rescheduling (C<ev_periodic>), synchronous signals
		91	(C<ev_signal>), process status change events (C<ev_child>), and event
		92	watchers dealing with the event loop mechanism itself (C<ev_idle>,
		93	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
		94	file watchers (C<ev_stat>) and even limited support for fork events
		95	(C<ev_fork>).
		96
		97	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	99	for example).
33		100
34	=head1 CONVENTIONS	101	=head2 CONVENTIONS
35		102
36	Libev is very configurable. In this manual the default configuration	103	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	104	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	105	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	106	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	107	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 *>)	108	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	109	this argument.
43		110
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	111	=head2 TIME REPRESENTATION
45		112
46	Libev represents time as a single floating point number, representing the	113	Libev represents time as a single floating point number, representing the
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	114	(fractional) number of seconds since the (POSIX) epoch (somewhere near
48	the beginning of 1970, details are complicated, don't ask). This type is	115	the beginning of 1970, details are complicated, don't ask). This type is
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	116	called C<ev_tstamp>, which is what you should use too. It usually aliases
50	to the double type in C.	117	to the C<double> type in C, and when you need to do any calculations on
		118	it, you should treat it as some floatingpoint value. Unlike the name
		119	component C<stamp> might indicate, it is also used for time differences
		120	throughout libev.
		121
		122	=head1 GLOBAL FUNCTIONS
		123
		124	These functions can be called anytime, even before initialising the
		125	library in any way.
51		126
52	=over 4	127	=over 4
53		128
54	=item ev_tstamp ev_time ()	129	=item ev_tstamp ev_time ()
55		130
56	Returns the current time as libev would use it.	131	Returns the current time as libev would use it. Please note that the
		132	C<ev_now> function is usually faster and also often returns the timestamp
		133	you actually want to know.
		134
		135	=item ev_sleep (ev_tstamp interval)
		136
		137	Sleep for the given interval: The current thread will be blocked until
		138	either it is interrupted or the given time interval has passed. Basically
		139	this is a subsecond-resolution C<sleep ()>.
57		140
58	=item int ev_version_major ()	141	=item int ev_version_major ()
59		142
60	=item int ev_version_minor ()	143	=item int ev_version_minor ()
61		144
62	You can find out the major and minor version numbers of the library	145	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	146	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	147	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	148	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
66	version of the library your program was compiled against.	149	version of the library your program was compiled against.
67		150
		151	These version numbers refer to the ABI version of the library, not the
		152	release version.
		153
68	Usually, it's a good idea to terminate if the major versions mismatch,	154	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	155	as this indicates an incompatible change. Minor versions are usually
70	compatible to older versions, so a larger minor version alone is usually	156	compatible to older versions, so a larger minor version alone is usually
71	not a problem.	157	not a problem.
72		158
		159	Example: Make sure we haven't accidentally been linked against the wrong
		160	version.
		161
		162	assert (("libev version mismatch",
		163	ev_version_major () == EV_VERSION_MAJOR
		164	&& ev_version_minor () >= EV_VERSION_MINOR));
		165
		166	=item unsigned int ev_supported_backends ()
		167
		168	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		169	value) compiled into this binary of libev (independent of their
		170	availability on the system you are running on). See C<ev_default_loop> for
		171	a description of the set values.
		172
		173	Example: make sure we have the epoll method, because yeah this is cool and
		174	a must have and can we have a torrent of it please!!!11
		175
		176	assert (("sorry, no epoll, no sex",
		177	ev_supported_backends () & EVBACKEND_EPOLL));
		178
		179	=item unsigned int ev_recommended_backends ()
		180
		181	Return the set of all backends compiled into this binary of libev and also
		182	recommended for this platform. This set is often smaller than the one
		183	returned by C<ev_supported_backends>, as for example kqueue is broken on
		184	most BSDs and will not be autodetected unless you explicitly request it
		185	(assuming you know what you are doing). This is the set of backends that
		186	libev will probe for if you specify no backends explicitly.
		187
		188	=item unsigned int ev_embeddable_backends ()
		189
		190	Returns the set of backends that are embeddable in other event loops. This
		191	is the theoretical, all-platform, value. To find which backends
		192	might be supported on the current system, you would need to look at
		193	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		194	recommended ones.
		195
		196	See the description of C<ev_embed> watchers for more info.
		197
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	198	=item ev_set_allocator (void *(*cb)(void *ptr, long size))
74		199
75	Sets the allocation function to use (the prototype is similar to the	200	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	201	semantics is identical - to the realloc C function). It is used to
77	and free memory (no surprises here). If it returns zero when memory	202	allocate and free memory (no surprises here). If it returns zero when
78	needs to be allocated, the library might abort or take some potentially	203	memory needs to be allocated, the library might abort or take some
79	destructive action. The default is your system realloc function.	204	potentially destructive action. The default is your system realloc
		205	function.
80		206
81	You could override this function in high-availability programs to, say,	207	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,	208	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.	209	or even to sleep a while and retry until some memory is available.
		210
		211	Example: Replace the libev allocator with one that waits a bit and then
		212	retries).
		213
		214	static void *
		215	persistent_realloc (void *ptr, size_t size)
		216	{
		217	for (;;)
		218	{
		219	void *newptr = realloc (ptr, size);
		220
		221	if (newptr)
		222	return newptr;
		223
		224	sleep (60);
		225	}
		226	}
		227
		228	...
		229	ev_set_allocator (persistent_realloc);
84		230
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	231	=item ev_set_syserr_cb (void (*cb)(const char *msg));
86		232
87	Set the callback function to call on a retryable syscall error (such	233	Set the callback function to call on a retryable syscall error (such
88	as failed select, poll, epoll_wait). The message is a printable string	234	as failed select, poll, epoll_wait). The message is a printable string
…		…
90	callback is set, then libev will expect it to remedy the sitution, no	236	callback is set, then libev will expect it to remedy the sitution, no
91	matter what, when it returns. That is, libev will generally retry the	237	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	238	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	239	(such as abort).
94		240
		241	Example: This is basically the same thing that libev does internally, too.
		242
		243	static void
		244	fatal_error (const char *msg)
		245	{
		246	perror (msg);
		247	abort ();
		248	}
		249
		250	...
		251	ev_set_syserr_cb (fatal_error);
		252
95	=back	253	=back
96		254
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	255	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
98		256
99	An event loop is described by a C<struct ev_loop *>. The library knows two	257	An event loop is described by a C<struct ev_loop *>. The library knows two
100	types of such loops, the I<default> loop, which supports signals and child	258	types of such loops, the I<default> loop, which supports signals and child
101	events, and dynamically created loops which do not.	259	events, and dynamically created loops which do not.
102		260
103	If you use threads, a common model is to run the default event loop	261	If you use threads, a common model is to run the default event loop
104	in your main thread (or in a separate thrad) and for each thread you	262	in your main thread (or in a separate thread) and for each thread you
105	create, you also create another event loop. Libev itself does no locking	263	create, you also create another event loop. Libev itself does no locking
106	whatsoever, so if you mix calls to the same event loop in different	264	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	265	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because it's hideous and inefficient).	266	done correctly, because it's hideous and inefficient).
109		267
…		…
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	270	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		271
114	This will initialise the default event loop if it hasn't been initialised	272	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	273	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	274	false. If it already was initialised it simply returns it (and ignores the
117	flags).	275	flags. If that is troubling you, check C<ev_backend ()> afterwards).
118		276
119	If you don't know what event loop to use, use the one returned from this	277	If you don't know what event loop to use, use the one returned from this
120	function.	278	function.
121		279
		280	Note that this function is I<not> thread-safe, so if you want to use it
		281	from multiple threads, you have to lock (note also that this is unlikely,
		282	as loops cannot bes hared easily between threads anyway).
		283
		284	The default loop is the only loop that can handle C<ev_signal> and
		285	C<ev_child> watchers, and to do this, it always registers a handler
		286	for C<SIGCHLD>. If this is a problem for your app you can either
		287	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		288	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		289	C<ev_default_init>.
		290
122	The flags argument can be used to specify special behaviour or specific	291	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).	292	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
124		293
125	It supports the following flags:	294	The following flags are supported:
126		295
127	=over 4	296	=over 4
128		297
129	=item C<EVFLAG_AUTO>	298	=item C<EVFLAG_AUTO>
130		299
…		…
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	307	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	308	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	309	useful to try out specific backends to test their performance, or to work
141	around bugs.	310	around bugs.
142		311
		312	=item C<EVFLAG_FORKCHECK>
		313
		314	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
		315	a fork, you can also make libev check for a fork in each iteration by
		316	enabling this flag.
		317
		318	This works by calling C<getpid ()> on every iteration of the loop,
		319	and thus this might slow down your event loop if you do a lot of loop
		320	iterations and little real work, but is usually not noticeable (on my
		321	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		322	without a syscall and thus I<very> fast, but my GNU/Linux system also has
		323	C<pthread_atfork> which is even faster).
		324
		325	The big advantage of this flag is that you can forget about fork (and
		326	forget about forgetting to tell libev about forking) when you use this
		327	flag.
		328
		329	This flag setting cannot be overriden or specified in the C<LIBEV_FLAGS>
		330	environment variable.
		331
143	=item C<EVMETHOD_SELECT> (portable select backend)	332	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		333
		334	This is your standard select(2) backend. Not I<completely> standard, as
		335	libev tries to roll its own fd_set with no limits on the number of fds,
		336	but if that fails, expect a fairly low limit on the number of fds when
		337	using this backend. It doesn't scale too well (O(highest_fd)), but its
		338	usually the fastest backend for a low number of (low-numbered :) fds.
		339
		340	To get good performance out of this backend you need a high amount of
		341	parallelity (most of the file descriptors should be busy). If you are
		342	writing a server, you should C<accept ()> in a loop to accept as many
		343	connections as possible during one iteration. You might also want to have
		344	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		345	readyness notifications you get per iteration.
		346
145	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	347	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		348
147	=item C<EVMETHOD_EPOLL> (linux only)	349	And this is your standard poll(2) backend. It's more complicated
		350	than select, but handles sparse fds better and has no artificial
		351	limit on the number of fds you can use (except it will slow down
		352	considerably with a lot of inactive fds). It scales similarly to select,
		353	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		354	performance tips.
148		355
149	=item C<EVMETHOD_KQUEUE> (some bsds only)	356	=item C<EVBACKEND_EPOLL> (value 4, Linux)
150		357
151	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	358	For few fds, this backend is a bit little slower than poll and select,
		359	but it scales phenomenally better. While poll and select usually scale
		360	like O(total_fds) where n is the total number of fds (or the highest fd),
		361	epoll scales either O(1) or O(active_fds). The epoll design has a number
		362	of shortcomings, such as silently dropping events in some hard-to-detect
		363	cases and requiring a syscall per fd change, no fork support and bad
		364	support for dup.
152		365
153	=item C<EVMETHOD_PORT> (solaris 10 only)	366	While stopping, setting and starting an I/O watcher in the same iteration
		367	will result in some caching, there is still a syscall per such incident
		368	(because the fd could point to a different file description now), so its
		369	best to avoid that. Also, C<dup ()>'ed file descriptors might not work
		370	very well if you register events for both fds.
		371
		372	Please note that epoll sometimes generates spurious notifications, so you
		373	need to use non-blocking I/O or other means to avoid blocking when no data
		374	(or space) is available.
		375
		376	Best performance from this backend is achieved by not unregistering all
		377	watchers for a file descriptor until it has been closed, if possible, i.e.
		378	keep at least one watcher active per fd at all times.
		379
		380	While nominally embeddeble in other event loops, this feature is broken in
		381	all kernel versions tested so far.
		382
		383	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		384
		385	Kqueue deserves special mention, as at the time of this writing, it
		386	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		387	with anything but sockets and pipes, except on Darwin, where of course
		388	it's completely useless). For this reason it's not being "autodetected"
		389	unless you explicitly specify it explicitly in the flags (i.e. using
		390	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		391	system like NetBSD.
		392
		393	You still can embed kqueue into a normal poll or select backend and use it
		394	only for sockets (after having made sure that sockets work with kqueue on
		395	the target platform). See C<ev_embed> watchers for more info.
		396
		397	It scales in the same way as the epoll backend, but the interface to the
		398	kernel is more efficient (which says nothing about its actual speed, of
		399	course). While stopping, setting and starting an I/O watcher does never
		400	cause an extra syscall as with C<EVBACKEND_EPOLL>, it still adds up to
		401	two event changes per incident, support for C<fork ()> is very bad and it
		402	drops fds silently in similarly hard-to-detect cases.
		403
		404	This backend usually performs well under most conditions.
		405
		406	While nominally embeddable in other event loops, this doesn't work
		407	everywhere, so you might need to test for this. And since it is broken
		408	almost everywhere, you should only use it when you have a lot of sockets
		409	(for which it usually works), by embedding it into another event loop
		410	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and using it only for
		411	sockets.
		412
		413	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		414
		415	This is not implemented yet (and might never be, unless you send me an
		416	implementation). According to reports, C</dev/poll> only supports sockets
		417	and is not embeddable, which would limit the usefulness of this backend
		418	immensely.
		419
		420	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		421
		422	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		423	it's really slow, but it still scales very well (O(active_fds)).
		424
		425	Please note that solaris event ports can deliver a lot of spurious
		426	notifications, so you need to use non-blocking I/O or other means to avoid
		427	blocking when no data (or space) is available.
		428
		429	While this backend scales well, it requires one system call per active
		430	file descriptor per loop iteration. For small and medium numbers of file
		431	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		432	might perform better.
		433
		434	On the positive side, ignoring the spurious readyness notifications, this
		435	backend actually performed to specification in all tests and is fully
		436	embeddable, which is a rare feat among the OS-specific backends.
		437
		438	=item C<EVBACKEND_ALL>
		439
		440	Try all backends (even potentially broken ones that wouldn't be tried
		441	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		442	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		443
		444	It is definitely not recommended to use this flag.
		445
		446	=back
154		447
155	If one or more of these are ored into the flags value, then only these	448	If one or more of these are ored into the flags value, then only these
156	backends will be tried (in the reverse order as given here). If one are	449	backends will be tried (in the reverse order as listed here). If none are
157	specified, any backend will do.	450	specified, all backends in C<ev_recommended_backends ()> will be tried.
158		451
159	=back	452	The most typical usage is like this:
		453
		454	if (!ev_default_loop (0))
		455	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		456
		457	Restrict libev to the select and poll backends, and do not allow
		458	environment settings to be taken into account:
		459
		460	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		461
		462	Use whatever libev has to offer, but make sure that kqueue is used if
		463	available (warning, breaks stuff, best use only with your own private
		464	event loop and only if you know the OS supports your types of fds):
		465
		466	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
160		467
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	468	=item struct ev_loop *ev_loop_new (unsigned int flags)
162		469
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	470	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	471	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	472	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).	473	undefined behaviour (or a failed assertion if assertions are enabled).
167		474
		475	Note that this function I<is> thread-safe, and the recommended way to use
		476	libev with threads is indeed to create one loop per thread, and using the
		477	default loop in the "main" or "initial" thread.
		478
		479	Example: Try to create a event loop that uses epoll and nothing else.
		480
		481	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		482	if (!epoller)
		483	fatal ("no epoll found here, maybe it hides under your chair");
		484
168	=item ev_default_destroy ()	485	=item ev_default_destroy ()
169		486
170	Destroys the default loop again (frees all memory and kernel state	487	Destroys the default loop again (frees all memory and kernel state
171	etc.). This stops all registered event watchers (by not touching them in	488	etc.). None of the active event watchers will be stopped in the normal
172	any way whatsoever, although you cannot rely on this :).	489	sense, so e.g. C<ev_is_active> might still return true. It is your
		490	responsibility to either stop all watchers cleanly yoursef I<before>
		491	calling this function, or cope with the fact afterwards (which is usually
		492	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		493	for example).
		494
		495	Note that certain global state, such as signal state, will not be freed by
		496	this function, and related watchers (such as signal and child watchers)
		497	would need to be stopped manually.
		498
		499	In general it is not advisable to call this function except in the
		500	rare occasion where you really need to free e.g. the signal handling
		501	pipe fds. If you need dynamically allocated loops it is better to use
		502	C<ev_loop_new> and C<ev_loop_destroy>).
173		503
174	=item ev_loop_destroy (loop)	504	=item ev_loop_destroy (loop)
175		505
176	Like C<ev_default_destroy>, but destroys an event loop created by an	506	Like C<ev_default_destroy>, but destroys an event loop created by an
177	earlier call to C<ev_loop_new>.	507	earlier call to C<ev_loop_new>.
178		508
179	=item ev_default_fork ()	509	=item ev_default_fork ()
180		510
		511	This function sets a flag that causes subsequent C<ev_loop> iterations
181	This function reinitialises the kernel state for backends that have	512	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	513	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	514	the child process (or both child and parent, but that again makes little
184	again makes little sense).	515	sense). You I<must> call it in the child before using any of the libev
		516	functions, and it will only take effect at the next C<ev_loop> iteration.
185		517
186	You I<must> call this function after forking if and only if you want to	518	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	519	process if and only if you want to use the event library in the child. If
188	have to call it.	520	you just fork+exec, you don't have to call it at all.
189		521
190	The function itself is quite fast and it's usually not a problem to call	522	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	523	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>:	524	quite nicely into a call to C<pthread_atfork>:
193		525
…		…
197		529
198	Like C<ev_default_fork>, but acts on an event loop created by	530	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	531	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.	532	after fork, and how you do this is entirely your own problem.
201		533
		534	=item int ev_is_default_loop (loop)
		535
		536	Returns true when the given loop actually is the default loop, false otherwise.
		537
		538	=item unsigned int ev_loop_count (loop)
		539
		540	Returns the count of loop iterations for the loop, which is identical to
		541	the number of times libev did poll for new events. It starts at C<0> and
		542	happily wraps around with enough iterations.
		543
		544	This value can sometimes be useful as a generation counter of sorts (it
		545	"ticks" the number of loop iterations), as it roughly corresponds with
		546	C<ev_prepare> and C<ev_check> calls.
		547
202	=item unsigned int ev_method (loop)	548	=item unsigned int ev_backend (loop)
203		549
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	550	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	551	use.
206		552
207	=item ev_tstamp ev_now (loop)	553	=item ev_tstamp ev_now (loop)
208		554
209	Returns the current "event loop time", which is the time the event loop	555	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	556	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	557	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	558	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).	559	event occurring (or more correctly, libev finding out about it).
214		560
215	=item ev_loop (loop, int flags)	561	=item ev_loop (loop, int flags)
216		562
217	Finally, this is it, the event handler. This function usually is called	563	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	564	after you initialised all your watchers and you want to start handling
219	events.	565	events.
220		566
221	If the flags argument is specified as 0, it will not return until either	567	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.	568	either no event watchers are active anymore or C<ev_unloop> was called.
		569
		570	Please note that an explicit C<ev_unloop> is usually better than
		571	relying on all watchers to be stopped when deciding when a program has
		572	finished (especially in interactive programs), but having a program that
		573	automatically loops as long as it has to and no longer by virtue of
		574	relying on its watchers stopping correctly is a thing of beauty.
223		575
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	576	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	577	those events and any outstanding ones, but will not block your process in
226	case there are no events and will return after one iteration of the loop.	578	case there are no events and will return after one iteration of the loop.
227		579
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	580	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	581	neccessary) and will handle those and any outstanding ones. It will block
230	your process until at least one new event arrives, and will return after	582	your process until at least one new event arrives, and will return after
231	one iteration of the loop.	583	one iteration of the loop. This is useful if you are waiting for some
		584	external event in conjunction with something not expressible using other
		585	libev watchers. However, a pair of C<ev_prepare>/C<ev_check> watchers is
		586	usually a better approach for this kind of thing.
232		587
233	This flags value could be used to implement alternative looping	588	Here are the gory details of what C<ev_loop> does:
234	constructs, but the C<prepare> and C<check> watchers provide a better and	589
235	more generic mechanism.	590	- Before the first iteration, call any pending watchers.
		591	* If EVFLAG_FORKCHECK was used, check for a fork.
		592	- If a fork was detected, queue and call all fork watchers.
		593	- Queue and call all prepare watchers.
		594	- If we have been forked, recreate the kernel state.
		595	- Update the kernel state with all outstanding changes.
		596	- Update the "event loop time".
		597	- Calculate for how long to sleep or block, if at all
		598	(active idle watchers, EVLOOP_NONBLOCK or not having
		599	any active watchers at all will result in not sleeping).
		600	- Sleep if the I/O and timer collect interval say so.
		601	- Block the process, waiting for any events.
		602	- Queue all outstanding I/O (fd) events.
		603	- Update the "event loop time" and do time jump handling.
		604	- Queue all outstanding timers.
		605	- Queue all outstanding periodics.
		606	- If no events are pending now, queue all idle watchers.
		607	- Queue all check watchers.
		608	- Call all queued watchers in reverse order (i.e. check watchers first).
		609	Signals and child watchers are implemented as I/O watchers, and will
		610	be handled here by queueing them when their watcher gets executed.
		611	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		612	were used, or there are no active watchers, return, otherwise
		613	continue with step *.
		614
		615	Example: Queue some jobs and then loop until no events are outstanding
		616	anymore.
		617
		618	... queue jobs here, make sure they register event watchers as long
		619	... as they still have work to do (even an idle watcher will do..)
		620	ev_loop (my_loop, 0);
		621	... jobs done. yeah!
236		622
237	=item ev_unloop (loop, how)	623	=item ev_unloop (loop, how)
238		624
239	Can be used to make a call to C<ev_loop> return early (but only after it	625	Can be used to make a call to C<ev_loop> return early (but only after it
240	has processed all outstanding events). The C<how> argument must be either	626	has processed all outstanding events). The C<how> argument must be either
241	C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or	627	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
242	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	628	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
		629
		630	This "unloop state" will be cleared when entering C<ev_loop> again.
243		631
244	=item ev_ref (loop)	632	=item ev_ref (loop)
245		633
246	=item ev_unref (loop)	634	=item ev_unref (loop)
247		635
…		…
252	returning, ev_unref() after starting, and ev_ref() before stopping it. For	640	returning, ev_unref() after starting, and ev_ref() before stopping it. For
253	example, libev itself uses this for its internal signal pipe: It is not	641	example, libev itself uses this for its internal signal pipe: It is not
254	visible to the libev user and should not keep C<ev_loop> from exiting if	642	visible to the libev user and should not keep C<ev_loop> from exiting if
255	no event watchers registered by it are active. It is also an excellent	643	no event watchers registered by it are active. It is also an excellent
256	way to do this for generic recurring timers or from within third-party	644	way to do this for generic recurring timers or from within third-party
257	libraries. Just remember to I<unref after start> and I<ref before stop>.	645	libraries. Just remember to I<unref after start> and I<ref before stop>
		646	(but only if the watcher wasn't active before, or was active before,
		647	respectively).
		648
		649	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		650	running when nothing else is active.
		651
		652	struct ev_signal exitsig;
		653	ev_signal_init (&exitsig, sig_cb, SIGINT);
		654	ev_signal_start (loop, &exitsig);
		655	evf_unref (loop);
		656
		657	Example: For some weird reason, unregister the above signal handler again.
		658
		659	ev_ref (loop);
		660	ev_signal_stop (loop, &exitsig);
		661
		662	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		663
		664	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		665
		666	These advanced functions influence the time that libev will spend waiting
		667	for events. Both are by default C<0>, meaning that libev will try to
		668	invoke timer/periodic callbacks and I/O callbacks with minimum latency.
		669
		670	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		671	allows libev to delay invocation of I/O and timer/periodic callbacks to
		672	increase efficiency of loop iterations.
		673
		674	The background is that sometimes your program runs just fast enough to
		675	handle one (or very few) event(s) per loop iteration. While this makes
		676	the program responsive, it also wastes a lot of CPU time to poll for new
		677	events, especially with backends like C<select ()> which have a high
		678	overhead for the actual polling but can deliver many events at once.
		679
		680	By setting a higher I<io collect interval> you allow libev to spend more
		681	time collecting I/O events, so you can handle more events per iteration,
		682	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		683	C<ev_timer>) will be not affected. Setting this to a non-null value will
		684	introduce an additional C<ev_sleep ()> call into most loop iterations.
		685
		686	Likewise, by setting a higher I<timeout collect interval> you allow libev
		687	to spend more time collecting timeouts, at the expense of increased
		688	latency (the watcher callback will be called later). C<ev_io> watchers
		689	will not be affected. Setting this to a non-null value will not introduce
		690	any overhead in libev.
		691
		692	Many (busy) programs can usually benefit by setting the io collect
		693	interval to a value near C<0.1> or so, which is often enough for
		694	interactive servers (of course not for games), likewise for timeouts. It
		695	usually doesn't make much sense to set it to a lower value than C<0.01>,
		696	as this approsaches the timing granularity of most systems.
258		697
259	=back	698	=back
		699
260		700
261	=head1 ANATOMY OF A WATCHER	701	=head1 ANATOMY OF A WATCHER
262		702
263	A watcher is a structure that you create and register to record your	703	A watcher is a structure that you create and register to record your
264	interest in some event. For instance, if you want to wait for STDIN to	704	interest in some event. For instance, if you want to wait for STDIN to
…		…
297	*) >>), and you can stop watching for events at any time by calling the	737	*) >>), and you can stop watching for events at any time by calling the
298	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	738	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.
299		739
300	As long as your watcher is active (has been started but not stopped) you	740	As long as your watcher is active (has been started but not stopped) you
301	must not touch the values stored in it. Most specifically you must never	741	must not touch the values stored in it. Most specifically you must never
302	reinitialise it or call its set method.	742	reinitialise it or call its C<set> macro.
303
304	You can check whether an event is active by calling the C<ev_is_active
305	(watcher *)> macro. To see whether an event is outstanding (but the
306	callback for it has not been called yet) you can use the C<ev_is_pending
307	(watcher *)> macro.
308		743
309	Each and every callback receives the event loop pointer as first, the	744	Each and every callback receives the event loop pointer as first, the
310	registered watcher structure as second, and a bitset of received events as	745	registered watcher structure as second, and a bitset of received events as
311	third argument.	746	third argument.
312		747
…		…
336	The signal specified in the C<ev_signal> watcher has been received by a thread.	771	The signal specified in the C<ev_signal> watcher has been received by a thread.
337		772
338	=item C<EV_CHILD>	773	=item C<EV_CHILD>
339		774
340	The pid specified in the C<ev_child> watcher has received a status change.	775	The pid specified in the C<ev_child> watcher has received a status change.
		776
		777	=item C<EV_STAT>
		778
		779	The path specified in the C<ev_stat> watcher changed its attributes somehow.
341		780
342	=item C<EV_IDLE>	781	=item C<EV_IDLE>
343		782
344	The C<ev_idle> watcher has determined that you have nothing better to do.	783	The C<ev_idle> watcher has determined that you have nothing better to do.
345		784
…		…
353	received events. Callbacks of both watcher types can start and stop as	792	received events. Callbacks of both watcher types can start and stop as
354	many watchers as they want, and all of them will be taken into account	793	many watchers as they want, and all of them will be taken into account
355	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	794	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
356	C<ev_loop> from blocking).	795	C<ev_loop> from blocking).
357		796
		797	=item C<EV_EMBED>
		798
		799	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		800
		801	=item C<EV_FORK>
		802
		803	The event loop has been resumed in the child process after fork (see
		804	C<ev_fork>).
		805
		806	=item C<EV_ASYNC>
		807
		808	The given async watcher has been asynchronously notified (see C<ev_async>).
		809
358	=item C<EV_ERROR>	810	=item C<EV_ERROR>
359		811
360	An unspecified error has occured, the watcher has been stopped. This might	812	An unspecified error has occured, the watcher has been stopped. This might
361	happen because the watcher could not be properly started because libev	813	happen because the watcher could not be properly started because libev
362	ran out of memory, a file descriptor was found to be closed or any other	814	ran out of memory, a file descriptor was found to be closed or any other
…		…
368	your callbacks is well-written it can just attempt the operation and cope	820	your callbacks is well-written it can just attempt the operation and cope
369	with the error from read() or write(). This will not work in multithreaded	821	with the error from read() or write(). This will not work in multithreaded
370	programs, though, so beware.	822	programs, though, so beware.
371		823
372	=back	824	=back
		825
		826	=head2 GENERIC WATCHER FUNCTIONS
		827
		828	In the following description, C<TYPE> stands for the watcher type,
		829	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
		830
		831	=over 4
		832
		833	=item C<ev_init> (ev_TYPE *watcher, callback)
		834
		835	This macro initialises the generic portion of a watcher. The contents
		836	of the watcher object can be arbitrary (so C<malloc> will do). Only
		837	the generic parts of the watcher are initialised, you I<need> to call
		838	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		839	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		840	which rolls both calls into one.
		841
		842	You can reinitialise a watcher at any time as long as it has been stopped
		843	(or never started) and there are no pending events outstanding.
		844
		845	The callback is always of type C<void (*)(ev_loop *loop, ev_TYPE *watcher,
		846	int revents)>.
		847
		848	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		849
		850	This macro initialises the type-specific parts of a watcher. You need to
		851	call C<ev_init> at least once before you call this macro, but you can
		852	call C<ev_TYPE_set> any number of times. You must not, however, call this
		853	macro on a watcher that is active (it can be pending, however, which is a
		854	difference to the C<ev_init> macro).
		855
		856	Although some watcher types do not have type-specific arguments
		857	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		858
		859	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		860
		861	This convinience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		862	calls into a single call. This is the most convinient method to initialise
		863	a watcher. The same limitations apply, of course.
		864
		865	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		866
		867	Starts (activates) the given watcher. Only active watchers will receive
		868	events. If the watcher is already active nothing will happen.
		869
		870	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		871
		872	Stops the given watcher again (if active) and clears the pending
		873	status. It is possible that stopped watchers are pending (for example,
		874	non-repeating timers are being stopped when they become pending), but
		875	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If
		876	you want to free or reuse the memory used by the watcher it is therefore a
		877	good idea to always call its C<ev_TYPE_stop> function.
		878
		879	=item bool ev_is_active (ev_TYPE *watcher)
		880
		881	Returns a true value iff the watcher is active (i.e. it has been started
		882	and not yet been stopped). As long as a watcher is active you must not modify
		883	it.
		884
		885	=item bool ev_is_pending (ev_TYPE *watcher)
		886
		887	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		888	events but its callback has not yet been invoked). As long as a watcher
		889	is pending (but not active) you must not call an init function on it (but
		890	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		891	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		892	it).
		893
		894	=item callback ev_cb (ev_TYPE *watcher)
		895
		896	Returns the callback currently set on the watcher.
		897
		898	=item ev_cb_set (ev_TYPE *watcher, callback)
		899
		900	Change the callback. You can change the callback at virtually any time
		901	(modulo threads).
		902
		903	=item ev_set_priority (ev_TYPE *watcher, priority)
		904
		905	=item int ev_priority (ev_TYPE *watcher)
		906
		907	Set and query the priority of the watcher. The priority is a small
		908	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		909	(default: C<-2>). Pending watchers with higher priority will be invoked
		910	before watchers with lower priority, but priority will not keep watchers
		911	from being executed (except for C<ev_idle> watchers).
		912
		913	This means that priorities are I<only> used for ordering callback
		914	invocation after new events have been received. This is useful, for
		915	example, to reduce latency after idling, or more often, to bind two
		916	watchers on the same event and make sure one is called first.
		917
		918	If you need to suppress invocation when higher priority events are pending
		919	you need to look at C<ev_idle> watchers, which provide this functionality.
		920
		921	You I<must not> change the priority of a watcher as long as it is active or
		922	pending.
		923
		924	The default priority used by watchers when no priority has been set is
		925	always C<0>, which is supposed to not be too high and not be too low :).
		926
		927	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		928	fine, as long as you do not mind that the priority value you query might
		929	or might not have been adjusted to be within valid range.
		930
		931	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		932
		933	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		934	C<loop> nor C<revents> need to be valid as long as the watcher callback
		935	can deal with that fact.
		936
		937	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		938
		939	If the watcher is pending, this function returns clears its pending status
		940	and returns its C<revents> bitset (as if its callback was invoked). If the
		941	watcher isn't pending it does nothing and returns C<0>.
		942
		943	=back
		944
373		945
374	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	946	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375		947
376	Each watcher has, by default, a member C<void *data> that you can change	948	Each watcher has, by default, a member C<void *data> that you can change
377	and read at any time, libev will completely ignore it. This can be used	949	and read at any time, libev will completely ignore it. This can be used
…		…
395	{	967	{
396	struct my_io *w = (struct my_io *)w_;	968	struct my_io *w = (struct my_io *)w_;
397	...	969	...
398	}	970	}
399		971
400	More interesting and less C-conformant ways of catsing your callback type	972	More interesting and less C-conformant ways of casting your callback type
401	have been omitted....	973	instead have been omitted.
		974
		975	Another common scenario is having some data structure with multiple
		976	watchers:
		977
		978	struct my_biggy
		979	{
		980	int some_data;
		981	ev_timer t1;
		982	ev_timer t2;
		983	}
		984
		985	In this case getting the pointer to C<my_biggy> is a bit more complicated,
		986	you need to use C<offsetof>:
		987
		988	#include <stddef.h>
		989
		990	static void
		991	t1_cb (EV_P_ struct ev_timer *w, int revents)
		992	{
		993	struct my_biggy big = (struct my_biggy *
		994	(((char *)w) - offsetof (struct my_biggy, t1));
		995	}
		996
		997	static void
		998	t2_cb (EV_P_ struct ev_timer *w, int revents)
		999	{
		1000	struct my_biggy big = (struct my_biggy *
		1001	(((char *)w) - offsetof (struct my_biggy, t2));
		1002	}
402		1003
403		1004
404	=head1 WATCHER TYPES	1005	=head1 WATCHER TYPES
405		1006
406	This section describes each watcher in detail, but will not repeat	1007	This section describes each watcher in detail, but will not repeat
407	information given in the last section.	1008	information given in the last section. Any initialisation/set macros,
		1009	functions and members specific to the watcher type are explained.
408		1010
		1011	Members are additionally marked with either I<[read-only]>, meaning that,
		1012	while the watcher is active, you can look at the member and expect some
		1013	sensible content, but you must not modify it (you can modify it while the
		1014	watcher is stopped to your hearts content), or I<[read-write]>, which
		1015	means you can expect it to have some sensible content while the watcher
		1016	is active, but you can also modify it. Modifying it may not do something
		1017	sensible or take immediate effect (or do anything at all), but libev will
		1018	not crash or malfunction in any way.
		1019
		1020
409	=head2 C<ev_io> - is this file descriptor readable or writable	1021	=head2 C<ev_io> - is this file descriptor readable or writable?
410		1022
411	I/O watchers check whether a file descriptor is readable or writable	1023	I/O watchers check whether a file descriptor is readable or writable
412	in each iteration of the event loop (This behaviour is called	1024	in each iteration of the event loop, or, more precisely, when reading
413	level-triggering because you keep receiving events as long as the	1025	would not block the process and writing would at least be able to write
414	condition persists. Remember you can stop the watcher if you don't want to	1026	some data. This behaviour is called level-triggering because you keep
415	act on the event and neither want to receive future events).	1027	receiving events as long as the condition persists. Remember you can stop
		1028	the watcher if you don't want to act on the event and neither want to
		1029	receive future events.
416		1030
417	In general you can register as many read and/or write event watchers oer	1031	In general you can register as many read and/or write event watchers per
418	fd as you want (as long as you don't confuse yourself). Setting all file	1032	fd as you want (as long as you don't confuse yourself). Setting all file
419	descriptors to non-blocking mode is also usually a good idea (but not	1033	descriptors to non-blocking mode is also usually a good idea (but not
420	required if you know what you are doing).	1034	required if you know what you are doing).
421		1035
422	You have to be careful with dup'ed file descriptors, though. Some backends
423	(the linux epoll backend is a notable example) cannot handle dup'ed file
424	descriptors correctly if you register interest in two or more fds pointing
425	to the same file/socket etc. description.
426
427	If you must do this, then force the use of a known-to-be-good backend	1036	If you must do this, then force the use of a known-to-be-good backend
428	(at the time of this writing, this includes only EVMETHOD_SELECT and	1037	(at the time of this writing, this includes only C<EVBACKEND_SELECT> and
429	EVMETHOD_POLL).	1038	C<EVBACKEND_POLL>).
		1039
		1040	Another thing you have to watch out for is that it is quite easy to
		1041	receive "spurious" readyness notifications, that is your callback might
		1042	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1043	because there is no data. Not only are some backends known to create a
		1044	lot of those (for example solaris ports), it is very easy to get into
		1045	this situation even with a relatively standard program structure. Thus
		1046	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1047	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1048
		1049	If you cannot run the fd in non-blocking mode (for example you should not
		1050	play around with an Xlib connection), then you have to seperately re-test
		1051	whether a file descriptor is really ready with a known-to-be good interface
		1052	such as poll (fortunately in our Xlib example, Xlib already does this on
		1053	its own, so its quite safe to use).
		1054
		1055	=head3 The special problem of disappearing file descriptors
		1056
		1057	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1058	descriptor (either by calling C<close> explicitly or by any other means,
		1059	such as C<dup>). The reason is that you register interest in some file
		1060	descriptor, but when it goes away, the operating system will silently drop
		1061	this interest. If another file descriptor with the same number then is
		1062	registered with libev, there is no efficient way to see that this is, in
		1063	fact, a different file descriptor.
		1064
		1065	To avoid having to explicitly tell libev about such cases, libev follows
		1066	the following policy: Each time C<ev_io_set> is being called, libev
		1067	will assume that this is potentially a new file descriptor, otherwise
		1068	it is assumed that the file descriptor stays the same. That means that
		1069	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1070	descriptor even if the file descriptor number itself did not change.
		1071
		1072	This is how one would do it normally anyway, the important point is that
		1073	the libev application should not optimise around libev but should leave
		1074	optimisations to libev.
		1075
		1076	=head3 The special problem of dup'ed file descriptors
		1077
		1078	Some backends (e.g. epoll), cannot register events for file descriptors,
		1079	but only events for the underlying file descriptions. That means when you
		1080	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1081	events for them, only one file descriptor might actually receive events.
		1082
		1083	There is no workaround possible except not registering events
		1084	for potentially C<dup ()>'ed file descriptors, or to resort to
		1085	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1086
		1087	=head3 The special problem of fork
		1088
		1089	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1090	useless behaviour. Libev fully supports fork, but needs to be told about
		1091	it in the child.
		1092
		1093	To support fork in your programs, you either have to call
		1094	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1095	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1096	C<EVBACKEND_POLL>.
		1097
		1098	=head3 The special problem of SIGPIPE
		1099
		1100	While not really specific to libev, it is easy to forget about SIGPIPE:
		1101	when reading from a pipe whose other end has been closed, your program
		1102	gets send a SIGPIPE, which, by default, aborts your program. For most
		1103	programs this is sensible behaviour, for daemons, this is usually
		1104	undesirable.
		1105
		1106	So when you encounter spurious, unexplained daemon exits, make sure you
		1107	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1108	somewhere, as that would have given you a big clue).
		1109
		1110
		1111	=head3 Watcher-Specific Functions
430		1112
431	=over 4	1113	=over 4
432		1114
433	=item ev_io_init (ev_io *, callback, int fd, int events)	1115	=item ev_io_init (ev_io *, callback, int fd, int events)
434		1116
435	=item ev_io_set (ev_io *, int fd, int events)	1117	=item ev_io_set (ev_io *, int fd, int events)
436		1118
437	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1119	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
438	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1120	rceeive events for and events is either C<EV_READ>, C<EV_WRITE> or
439	EV_WRITE> to receive the given events.	1121	C<EV_READ | EV_WRITE> to receive the given events.
		1122
		1123	=item int fd [read-only]
		1124
		1125	The file descriptor being watched.
		1126
		1127	=item int events [read-only]
		1128
		1129	The events being watched.
440		1130
441	=back	1131	=back
442		1132
		1133	=head3 Examples
		1134
		1135	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1136	readable, but only once. Since it is likely line-buffered, you could
		1137	attempt to read a whole line in the callback.
		1138
		1139	static void
		1140	stdin_readable_cb (struct ev_loop *loop, struct ev_io *w, int revents)
		1141	{
		1142	ev_io_stop (loop, w);
		1143	.. read from stdin here (or from w->fd) and haqndle any I/O errors
		1144	}
		1145
		1146	...
		1147	struct ev_loop *loop = ev_default_init (0);
		1148	struct ev_io stdin_readable;
		1149	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1150	ev_io_start (loop, &stdin_readable);
		1151	ev_loop (loop, 0);
		1152
		1153
443	=head2 C<ev_timer> - relative and optionally recurring timeouts	1154	=head2 C<ev_timer> - relative and optionally repeating timeouts
444		1155
445	Timer watchers are simple relative timers that generate an event after a	1156	Timer watchers are simple relative timers that generate an event after a
446	given time, and optionally repeating in regular intervals after that.	1157	given time, and optionally repeating in regular intervals after that.
447		1158
448	The timers are based on real time, that is, if you register an event that	1159	The timers are based on real time, that is, if you register an event that
449	times out after an hour and youreset your system clock to last years	1160	times out after an hour and you reset your system clock to last years
450	time, it will still time out after (roughly) and hour. "Roughly" because	1161	time, it will still time out after (roughly) and hour. "Roughly" because
451	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1162	detecting time jumps is hard, and some inaccuracies are unavoidable (the
452	monotonic clock option helps a lot here).	1163	monotonic clock option helps a lot here).
453		1164
454	The relative timeouts are calculated relative to the C<ev_now ()>	1165	The relative timeouts are calculated relative to the C<ev_now ()>
455	time. This is usually the right thing as this timestamp refers to the time	1166	time. This is usually the right thing as this timestamp refers to the time
456	of the event triggering whatever timeout you are modifying/starting. If	1167	of the event triggering whatever timeout you are modifying/starting. If
457	you suspect event processing to be delayed and you *need* to base the timeout	1168	you suspect event processing to be delayed and you I<need> to base the timeout
458	ion the current time, use something like this to adjust for this:	1169	on the current time, use something like this to adjust for this:
459		1170
460	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1171	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
		1172
		1173	The callback is guarenteed to be invoked only when its timeout has passed,
		1174	but if multiple timers become ready during the same loop iteration then
		1175	order of execution is undefined.
		1176
		1177	=head3 Watcher-Specific Functions and Data Members
461		1178
462	=over 4	1179	=over 4
463		1180
464	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1181	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
465		1182
…		…
471	later, again, and again, until stopped manually.	1188	later, again, and again, until stopped manually.
472		1189
473	The timer itself will do a best-effort at avoiding drift, that is, if you	1190	The timer itself will do a best-effort at avoiding drift, that is, if you
474	configure a timer to trigger every 10 seconds, then it will trigger at	1191	configure a timer to trigger every 10 seconds, then it will trigger at
475	exactly 10 second intervals. If, however, your program cannot keep up with	1192	exactly 10 second intervals. If, however, your program cannot keep up with
476	the timer (ecause it takes longer than those 10 seconds to do stuff) the	1193	the timer (because it takes longer than those 10 seconds to do stuff) the
477	timer will not fire more than once per event loop iteration.	1194	timer will not fire more than once per event loop iteration.
478		1195
479	=item ev_timer_again (loop)	1196	=item ev_timer_again (loop, ev_timer *)
480		1197
481	This will act as if the timer timed out and restart it again if it is	1198	This will act as if the timer timed out and restart it again if it is
482	repeating. The exact semantics are:	1199	repeating. The exact semantics are:
483		1200
		1201	If the timer is pending, its pending status is cleared.
		1202
484	If the timer is started but nonrepeating, stop it.	1203	If the timer is started but nonrepeating, stop it (as if it timed out).
485		1204
486	If the timer is repeating, either start it if necessary (with the repeat	1205	If the timer is repeating, either start it if necessary (with the
487	value), or reset the running timer to the repeat value.	1206	C<repeat> value), or reset the running timer to the C<repeat> value.
488		1207
489	This sounds a bit complicated, but here is a useful and typical	1208	This sounds a bit complicated, but here is a useful and typical
490	example: Imagine you have a tcp connection and you want a so-called idle	1209	example: Imagine you have a tcp connection and you want a so-called idle
491	timeout, that is, you want to be called when there have been, say, 60	1210	timeout, that is, you want to be called when there have been, say, 60
492	seconds of inactivity on the socket. The easiest way to do this is to	1211	seconds of inactivity on the socket. The easiest way to do this is to
493	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1212	configure an C<ev_timer> with a C<repeat> value of C<60> and then call
494	time you successfully read or write some data. If you go into an idle	1213	C<ev_timer_again> each time you successfully read or write some data. If
495	state where you do not expect data to travel on the socket, you can stop	1214	you go into an idle state where you do not expect data to travel on the
496	the timer, and again will automatically restart it if need be.	1215	socket, you can C<ev_timer_stop> the timer, and C<ev_timer_again> will
		1216	automatically restart it if need be.
		1217
		1218	That means you can ignore the C<after> value and C<ev_timer_start>
		1219	altogether and only ever use the C<repeat> value and C<ev_timer_again>:
		1220
		1221	ev_timer_init (timer, callback, 0., 5.);
		1222	ev_timer_again (loop, timer);
		1223	...
		1224	timer->again = 17.;
		1225	ev_timer_again (loop, timer);
		1226	...
		1227	timer->again = 10.;
		1228	ev_timer_again (loop, timer);
		1229
		1230	This is more slightly efficient then stopping/starting the timer each time
		1231	you want to modify its timeout value.
		1232
		1233	=item ev_tstamp repeat [read-write]
		1234
		1235	The current C<repeat> value. Will be used each time the watcher times out
		1236	or C<ev_timer_again> is called and determines the next timeout (if any),
		1237	which is also when any modifications are taken into account.
497		1238
498	=back	1239	=back
499		1240
		1241	=head3 Examples
		1242
		1243	Example: Create a timer that fires after 60 seconds.
		1244
		1245	static void
		1246	one_minute_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
		1247	{
		1248	.. one minute over, w is actually stopped right here
		1249	}
		1250
		1251	struct ev_timer mytimer;
		1252	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1253	ev_timer_start (loop, &mytimer);
		1254
		1255	Example: Create a timeout timer that times out after 10 seconds of
		1256	inactivity.
		1257
		1258	static void
		1259	timeout_cb (struct ev_loop *loop, struct ev_timer *w, int revents)
		1260	{
		1261	.. ten seconds without any activity
		1262	}
		1263
		1264	struct ev_timer mytimer;
		1265	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1266	ev_timer_again (&mytimer); /* start timer */
		1267	ev_loop (loop, 0);
		1268
		1269	// and in some piece of code that gets executed on any "activity":
		1270	// reset the timeout to start ticking again at 10 seconds
		1271	ev_timer_again (&mytimer);
		1272
		1273
500	=head2 C<ev_periodic> - to cron or not to cron	1274	=head2 C<ev_periodic> - to cron or not to cron?
501		1275
502	Periodic watchers are also timers of a kind, but they are very versatile	1276	Periodic watchers are also timers of a kind, but they are very versatile
503	(and unfortunately a bit complex).	1277	(and unfortunately a bit complex).
504		1278
505	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1279	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
506	but on wallclock time (absolute time). You can tell a periodic watcher	1280	but on wallclock time (absolute time). You can tell a periodic watcher
507	to trigger "at" some specific point in time. For example, if you tell a	1281	to trigger "at" some specific point in time. For example, if you tell a
508	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1282	periodic watcher to trigger in 10 seconds (by specifiying e.g. C<ev_now ()
509	+ 10.>) and then reset your system clock to the last year, then it will	1283	+ 10.>) and then reset your system clock to the last year, then it will
510	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1284	take a year to trigger the event (unlike an C<ev_timer>, which would trigger
511	roughly 10 seconds later and of course not if you reset your system time	1285	roughly 10 seconds later).
512	again).
513		1286
514	They can also be used to implement vastly more complex timers, such as	1287	They can also be used to implement vastly more complex timers, such as
515	triggering an event on eahc midnight, local time.	1288	triggering an event on each midnight, local time or other, complicated,
		1289	rules.
		1290
		1291	As with timers, the callback is guarenteed to be invoked only when the
		1292	time (C<at>) has been passed, but if multiple periodic timers become ready
		1293	during the same loop iteration then order of execution is undefined.
		1294
		1295	=head3 Watcher-Specific Functions and Data Members
516		1296
517	=over 4	1297	=over 4
518		1298
519	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1299	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
520		1300
521	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1301	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
522		1302
523	Lots of arguments, lets sort it out... There are basically three modes of	1303	Lots of arguments, lets sort it out... There are basically three modes of
524	operation, and we will explain them from simplest to complex:	1304	operation, and we will explain them from simplest to complex:
525		1305
526
527	=over 4	1306	=over 4
528		1307
529	=item * absolute timer (interval = reschedule_cb = 0)	1308	=item * absolute timer (at = time, interval = reschedule_cb = 0)
530		1309
531	In this configuration the watcher triggers an event at the wallclock time	1310	In this configuration the watcher triggers an event at the wallclock time
532	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1311	C<at> and doesn't repeat. It will not adjust when a time jump occurs,
533	that is, if it is to be run at January 1st 2011 then it will run when the	1312	that is, if it is to be run at January 1st 2011 then it will run when the
534	system time reaches or surpasses this time.	1313	system time reaches or surpasses this time.
535		1314
536	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1315	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
537		1316
538	In this mode the watcher will always be scheduled to time out at the next	1317	In this mode the watcher will always be scheduled to time out at the next
539	C<at + N * interval> time (for some integer N) and then repeat, regardless	1318	C<at + N * interval> time (for some integer N, which can also be negative)
540	of any time jumps.	1319	and then repeat, regardless of any time jumps.
541		1320
542	This can be used to create timers that do not drift with respect to system	1321	This can be used to create timers that do not drift with respect to system
543	time:	1322	time:
544		1323
545	ev_periodic_set (&periodic, 0., 3600., 0);	1324	ev_periodic_set (&periodic, 0., 3600., 0);
…		…
551		1330
552	Another way to think about it (for the mathematically inclined) is that	1331	Another way to think about it (for the mathematically inclined) is that
553	C<ev_periodic> will try to run the callback in this mode at the next possible	1332	C<ev_periodic> will try to run the callback in this mode at the next possible
554	time where C<time = at (mod interval)>, regardless of any time jumps.	1333	time where C<time = at (mod interval)>, regardless of any time jumps.
555		1334
		1335	For numerical stability it is preferable that the C<at> value is near
		1336	C<ev_now ()> (the current time), but there is no range requirement for
		1337	this value.
		1338
556	=item * manual reschedule mode (reschedule_cb = callback)	1339	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
557		1340
558	In this mode the values for C<interval> and C<at> are both being	1341	In this mode the values for C<interval> and C<at> are both being
559	ignored. Instead, each time the periodic watcher gets scheduled, the	1342	ignored. Instead, each time the periodic watcher gets scheduled, the
560	reschedule callback will be called with the watcher as first, and the	1343	reschedule callback will be called with the watcher as first, and the
561	current time as second argument.	1344	current time as second argument.
562		1345
563	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	1346	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
564	periodic watcher, ever, or make any event loop modifications>. If you need	1347	ever, or make any event loop modifications>. If you need to stop it,
565	to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards.	1348	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
566		1349	starting an C<ev_prepare> watcher, which is legal).
567	Also, I<< this callback must always return a time that is later than the
568	passed C<now> value >>. Not even C<now> itself will be ok.
569		1350
570	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1351	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
571	ev_tstamp now)>, e.g.:	1352	ev_tstamp now)>, e.g.:
572		1353
573	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1354	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
…		…
578	It must return the next time to trigger, based on the passed time value	1359	It must return the next time to trigger, based on the passed time value
579	(that is, the lowest time value larger than to the second argument). It	1360	(that is, the lowest time value larger than to the second argument). It
580	will usually be called just before the callback will be triggered, but	1361	will usually be called just before the callback will be triggered, but
581	might be called at other times, too.	1362	might be called at other times, too.
582		1363
		1364	NOTE: I<< This callback must always return a time that is later than the
		1365	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.
		1366
583	This can be used to create very complex timers, such as a timer that	1367	This can be used to create very complex timers, such as a timer that
584	triggers on each midnight, local time. To do this, you would calculate the	1368	triggers on each midnight, local time. To do this, you would calculate the
585	next midnight after C<now> and return the timestamp value for this. How you do this	1369	next midnight after C<now> and return the timestamp value for this. How
586	is, again, up to you (but it is not trivial).	1370	you do this is, again, up to you (but it is not trivial, which is the main
		1371	reason I omitted it as an example).
587		1372
588	=back	1373	=back
589		1374
590	=item ev_periodic_again (loop, ev_periodic *)	1375	=item ev_periodic_again (loop, ev_periodic *)
591		1376
592	Simply stops and restarts the periodic watcher again. This is only useful	1377	Simply stops and restarts the periodic watcher again. This is only useful
593	when you changed some parameters or the reschedule callback would return	1378	when you changed some parameters or the reschedule callback would return
594	a different time than the last time it was called (e.g. in a crond like	1379	a different time than the last time it was called (e.g. in a crond like
595	program when the crontabs have changed).	1380	program when the crontabs have changed).
596		1381
		1382	=item ev_tstamp offset [read-write]
		1383
		1384	When repeating, this contains the offset value, otherwise this is the
		1385	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
		1386
		1387	Can be modified any time, but changes only take effect when the periodic
		1388	timer fires or C<ev_periodic_again> is being called.
		1389
		1390	=item ev_tstamp interval [read-write]
		1391
		1392	The current interval value. Can be modified any time, but changes only
		1393	take effect when the periodic timer fires or C<ev_periodic_again> is being
		1394	called.
		1395
		1396	=item ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now) [read-write]
		1397
		1398	The current reschedule callback, or C<0>, if this functionality is
		1399	switched off. Can be changed any time, but changes only take effect when
		1400	the periodic timer fires or C<ev_periodic_again> is being called.
		1401
		1402	=item ev_tstamp at [read-only]
		1403
		1404	When active, contains the absolute time that the watcher is supposed to
		1405	trigger next.
		1406
597	=back	1407	=back
598		1408
		1409	=head3 Examples
		1410
		1411	Example: Call a callback every hour, or, more precisely, whenever the
		1412	system clock is divisible by 3600. The callback invocation times have
		1413	potentially a lot of jittering, but good long-term stability.
		1414
		1415	static void
		1416	clock_cb (struct ev_loop *loop, struct ev_io *w, int revents)
		1417	{
		1418	... its now a full hour (UTC, or TAI or whatever your clock follows)
		1419	}
		1420
		1421	struct ev_periodic hourly_tick;
		1422	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		1423	ev_periodic_start (loop, &hourly_tick);
		1424
		1425	Example: The same as above, but use a reschedule callback to do it:
		1426
		1427	#include <math.h>
		1428
		1429	static ev_tstamp
		1430	my_scheduler_cb (struct ev_periodic *w, ev_tstamp now)
		1431	{
		1432	return fmod (now, 3600.) + 3600.;
		1433	}
		1434
		1435	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		1436
		1437	Example: Call a callback every hour, starting now:
		1438
		1439	struct ev_periodic hourly_tick;
		1440	ev_periodic_init (&hourly_tick, clock_cb,
		1441	fmod (ev_now (loop), 3600.), 3600., 0);
		1442	ev_periodic_start (loop, &hourly_tick);
		1443
		1444
599	=head2 C<ev_signal> - signal me when a signal gets signalled	1445	=head2 C<ev_signal> - signal me when a signal gets signalled!
600		1446
601	Signal watchers will trigger an event when the process receives a specific	1447	Signal watchers will trigger an event when the process receives a specific
602	signal one or more times. Even though signals are very asynchronous, libev	1448	signal one or more times. Even though signals are very asynchronous, libev
603	will try it's best to deliver signals synchronously, i.e. as part of the	1449	will try it's best to deliver signals synchronously, i.e. as part of the
604	normal event processing, like any other event.	1450	normal event processing, like any other event.
…		…
608	with the kernel (thus it coexists with your own signal handlers as long	1454	with the kernel (thus it coexists with your own signal handlers as long
609	as you don't register any with libev). Similarly, when the last signal	1455	as you don't register any with libev). Similarly, when the last signal
610	watcher for a signal is stopped libev will reset the signal handler to	1456	watcher for a signal is stopped libev will reset the signal handler to
611	SIG_DFL (regardless of what it was set to before).	1457	SIG_DFL (regardless of what it was set to before).
612		1458
		1459	If possible and supported, libev will install its handlers with
		1460	C<SA_RESTART> behaviour enabled, so syscalls should not be unduly
		1461	interrupted. If you have a problem with syscalls getting interrupted by
		1462	signals you can block all signals in an C<ev_check> watcher and unblock
		1463	them in an C<ev_prepare> watcher.
		1464
		1465	=head3 Watcher-Specific Functions and Data Members
		1466
613	=over 4	1467	=over 4
614		1468
615	=item ev_signal_init (ev_signal *, callback, int signum)	1469	=item ev_signal_init (ev_signal *, callback, int signum)
616		1470
617	=item ev_signal_set (ev_signal *, int signum)	1471	=item ev_signal_set (ev_signal *, int signum)
618		1472
619	Configures the watcher to trigger on the given signal number (usually one	1473	Configures the watcher to trigger on the given signal number (usually one
620	of the C<SIGxxx> constants).	1474	of the C<SIGxxx> constants).
621		1475
		1476	=item int signum [read-only]
		1477
		1478	The signal the watcher watches out for.
		1479
622	=back	1480	=back
623		1481
		1482	=head3 Examples
		1483
		1484	Example: Try to exit cleanly on SIGINT and SIGTERM.
		1485
		1486	static void
		1487	sigint_cb (struct ev_loop *loop, struct ev_signal *w, int revents)
		1488	{
		1489	ev_unloop (loop, EVUNLOOP_ALL);
		1490	}
		1491
		1492	struct ev_signal signal_watcher;
		1493	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1494	ev_signal_start (loop, &sigint_cb);
		1495
		1496
624	=head2 C<ev_child> - wait for pid status changes	1497	=head2 C<ev_child> - watch out for process status changes
625		1498
626	Child watchers trigger when your process receives a SIGCHLD in response to	1499	Child watchers trigger when your process receives a SIGCHLD in response to
627	some child status changes (most typically when a child of yours dies).	1500	some child status changes (most typically when a child of yours dies). It
		1501	is permissible to install a child watcher I<after> the child has been
		1502	forked (which implies it might have already exited), as long as the event
		1503	loop isn't entered (or is continued from a watcher).
		1504
		1505	Only the default event loop is capable of handling signals, and therefore
		1506	you can only rgeister child watchers in the default event loop.
		1507
		1508	=head3 Process Interaction
		1509
		1510	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1511	initialised. This is necessary to guarantee proper behaviour even if
		1512	the first child watcher is started after the child exits. The occurance
		1513	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1514	synchronously as part of the event loop processing. Libev always reaps all
		1515	children, even ones not watched.
		1516
		1517	=head3 Overriding the Built-In Processing
		1518
		1519	Libev offers no special support for overriding the built-in child
		1520	processing, but if your application collides with libev's default child
		1521	handler, you can override it easily by installing your own handler for
		1522	C<SIGCHLD> after initialising the default loop, and making sure the
		1523	default loop never gets destroyed. You are encouraged, however, to use an
		1524	event-based approach to child reaping and thus use libev's support for
		1525	that, so other libev users can use C<ev_child> watchers freely.
		1526
		1527	=head3 Watcher-Specific Functions and Data Members
628		1528
629	=over 4	1529	=over 4
630		1530
631	=item ev_child_init (ev_child *, callback, int pid)	1531	=item ev_child_init (ev_child *, callback, int pid, int trace)
632		1532
633	=item ev_child_set (ev_child *, int pid)	1533	=item ev_child_set (ev_child *, int pid, int trace)
634		1534
635	Configures the watcher to wait for status changes of process C<pid> (or	1535	Configures the watcher to wait for status changes of process C<pid> (or
636	I<any> process if C<pid> is specified as C<0>). The callback can look	1536	I<any> process if C<pid> is specified as C<0>). The callback can look
637	at the C<rstatus> member of the C<ev_child> watcher structure to see	1537	at the C<rstatus> member of the C<ev_child> watcher structure to see
638	the status word (use the macros from C<sys/wait.h> and see your systems	1538	the status word (use the macros from C<sys/wait.h> and see your systems
639	C<waitpid> documentation). The C<rpid> member contains the pid of the	1539	C<waitpid> documentation). The C<rpid> member contains the pid of the
640	process causing the status change.	1540	process causing the status change. C<trace> must be either C<0> (only
		1541	activate the watcher when the process terminates) or C<1> (additionally
		1542	activate the watcher when the process is stopped or continued).
		1543
		1544	=item int pid [read-only]
		1545
		1546	The process id this watcher watches out for, or C<0>, meaning any process id.
		1547
		1548	=item int rpid [read-write]
		1549
		1550	The process id that detected a status change.
		1551
		1552	=item int rstatus [read-write]
		1553
		1554	The process exit/trace status caused by C<rpid> (see your systems
		1555	C<waitpid> and C<sys/wait.h> documentation for details).
641		1556
642	=back	1557	=back
643		1558
		1559	=head3 Examples
		1560
		1561	Example: C<fork()> a new process and install a child handler to wait for
		1562	its completion.
		1563
		1564	ev_child cw;
		1565
		1566	static void
		1567	child_cb (EV_P_ struct ev_child *w, int revents)
		1568	{
		1569	ev_child_stop (EV_A_ w);
		1570	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		1571	}
		1572
		1573	pid_t pid = fork ();
		1574
		1575	if (pid < 0)
		1576	// error
		1577	else if (pid == 0)
		1578	{
		1579	// the forked child executes here
		1580	exit (1);
		1581	}
		1582	else
		1583	{
		1584	ev_child_init (&cw, child_cb, pid, 0);
		1585	ev_child_start (EV_DEFAULT_ &cw);
		1586	}
		1587
		1588
		1589	=head2 C<ev_stat> - did the file attributes just change?
		1590
		1591	This watches a filesystem path for attribute changes. That is, it calls
		1592	C<stat> regularly (or when the OS says it changed) and sees if it changed
		1593	compared to the last time, invoking the callback if it did.
		1594
		1595	The path does not need to exist: changing from "path exists" to "path does
		1596	not exist" is a status change like any other. The condition "path does
		1597	not exist" is signified by the C<st_nlink> field being zero (which is
		1598	otherwise always forced to be at least one) and all the other fields of
		1599	the stat buffer having unspecified contents.
		1600
		1601	The path I<should> be absolute and I<must not> end in a slash. If it is
		1602	relative and your working directory changes, the behaviour is undefined.
		1603
		1604	Since there is no standard to do this, the portable implementation simply
		1605	calls C<stat (2)> regularly on the path to see if it changed somehow. You
		1606	can specify a recommended polling interval for this case. If you specify
		1607	a polling interval of C<0> (highly recommended!) then a I<suitable,
		1608	unspecified default> value will be used (which you can expect to be around
		1609	five seconds, although this might change dynamically). Libev will also
		1610	impose a minimum interval which is currently around C<0.1>, but thats
		1611	usually overkill.
		1612
		1613	This watcher type is not meant for massive numbers of stat watchers,
		1614	as even with OS-supported change notifications, this can be
		1615	resource-intensive.
		1616
		1617	At the time of this writing, only the Linux inotify interface is
		1618	implemented (implementing kqueue support is left as an exercise for the
		1619	reader). Inotify will be used to give hints only and should not change the
		1620	semantics of C<ev_stat> watchers, which means that libev sometimes needs
		1621	to fall back to regular polling again even with inotify, but changes are
		1622	usually detected immediately, and if the file exists there will be no
		1623	polling.
		1624
		1625	=head3 ABI Issues (Largefile Support)
		1626
		1627	Libev by default (unless the user overrides this) uses the default
		1628	compilation environment, which means that on systems with optionally
		1629	disabled large file support, you get the 32 bit version of the stat
		1630	structure. When using the library from programs that change the ABI to
		1631	use 64 bit file offsets the programs will fail. In that case you have to
		1632	compile libev with the same flags to get binary compatibility. This is
		1633	obviously the case with any flags that change the ABI, but the problem is
		1634	most noticably with ev_stat and largefile support.
		1635
		1636	=head3 Inotify
		1637
		1638	When C<inotify (7)> support has been compiled into libev (generally only
		1639	available on Linux) and present at runtime, it will be used to speed up
		1640	change detection where possible. The inotify descriptor will be created lazily
		1641	when the first C<ev_stat> watcher is being started.
		1642
		1643	Inotify presense does not change the semantics of C<ev_stat> watchers
		1644	except that changes might be detected earlier, and in some cases, to avoid
		1645	making regular C<stat> calls. Even in the presense of inotify support
		1646	there are many cases where libev has to resort to regular C<stat> polling.
		1647
		1648	(There is no support for kqueue, as apparently it cannot be used to
		1649	implement this functionality, due to the requirement of having a file
		1650	descriptor open on the object at all times).
		1651
		1652	=head3 The special problem of stat time resolution
		1653
		1654	The C<stat ()> syscall only supports full-second resolution portably, and
		1655	even on systems where the resolution is higher, many filesystems still
		1656	only support whole seconds.
		1657
		1658	That means that, if the time is the only thing that changes, you might
		1659	miss updates: on the first update, C<ev_stat> detects a change and calls
		1660	your callback, which does something. When there is another update within
		1661	the same second, C<ev_stat> will be unable to detect it.
		1662
		1663	The solution to this is to delay acting on a change for a second (or till
		1664	the next second boundary), using a roughly one-second delay C<ev_timer>
		1665	(C<ev_timer_set (w, 0., 1.01); ev_timer_again (loop, w)>). The C<.01>
		1666	is added to work around small timing inconsistencies of some operating
		1667	systems.
		1668
		1669	=head3 Watcher-Specific Functions and Data Members
		1670
		1671	=over 4
		1672
		1673	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		1674
		1675	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		1676
		1677	Configures the watcher to wait for status changes of the given
		1678	C<path>. The C<interval> is a hint on how quickly a change is expected to
		1679	be detected and should normally be specified as C<0> to let libev choose
		1680	a suitable value. The memory pointed to by C<path> must point to the same
		1681	path for as long as the watcher is active.
		1682
		1683	The callback will be receive C<EV_STAT> when a change was detected,
		1684	relative to the attributes at the time the watcher was started (or the
		1685	last change was detected).
		1686
		1687	=item ev_stat_stat (loop, ev_stat *)
		1688
		1689	Updates the stat buffer immediately with new values. If you change the
		1690	watched path in your callback, you could call this fucntion to avoid
		1691	detecting this change (while introducing a race condition). Can also be
		1692	useful simply to find out the new values.
		1693
		1694	=item ev_statdata attr [read-only]
		1695
		1696	The most-recently detected attributes of the file. Although the type is of
		1697	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		1698	suitable for your system. If the C<st_nlink> member is C<0>, then there
		1699	was some error while C<stat>ing the file.
		1700
		1701	=item ev_statdata prev [read-only]
		1702
		1703	The previous attributes of the file. The callback gets invoked whenever
		1704	C<prev> != C<attr>.
		1705
		1706	=item ev_tstamp interval [read-only]
		1707
		1708	The specified interval.
		1709
		1710	=item const char *path [read-only]
		1711
		1712	The filesystem path that is being watched.
		1713
		1714	=back
		1715
		1716	=head3 Examples
		1717
		1718	Example: Watch C</etc/passwd> for attribute changes.
		1719
		1720	static void
		1721	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		1722	{
		1723	/* /etc/passwd changed in some way */
		1724	if (w->attr.st_nlink)
		1725	{
		1726	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		1727	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		1728	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		1729	}
		1730	else
		1731	/* you shalt not abuse printf for puts */
		1732	puts ("wow, /etc/passwd is not there, expect problems. "
		1733	"if this is windows, they already arrived\n");
		1734	}
		1735
		1736	...
		1737	ev_stat passwd;
		1738
		1739	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		1740	ev_stat_start (loop, &passwd);
		1741
		1742	Example: Like above, but additionally use a one-second delay so we do not
		1743	miss updates (however, frequent updates will delay processing, too, so
		1744	one might do the work both on C<ev_stat> callback invocation I<and> on
		1745	C<ev_timer> callback invocation).
		1746
		1747	static ev_stat passwd;
		1748	static ev_timer timer;
		1749
		1750	static void
		1751	timer_cb (EV_P_ ev_timer *w, int revents)
		1752	{
		1753	ev_timer_stop (EV_A_ w);
		1754
		1755	/* now it's one second after the most recent passwd change */
		1756	}
		1757
		1758	static void
		1759	stat_cb (EV_P_ ev_stat *w, int revents)
		1760	{
		1761	/* reset the one-second timer */
		1762	ev_timer_again (EV_A_ &timer);
		1763	}
		1764
		1765	...
		1766	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		1767	ev_stat_start (loop, &passwd);
		1768	ev_timer_init (&timer, timer_cb, 0., 1.01);
		1769
		1770
644	=head2 C<ev_idle> - when you've got nothing better to do	1771	=head2 C<ev_idle> - when you've got nothing better to do...
645		1772
646	Idle watchers trigger events when there are no other events are pending	1773	Idle watchers trigger events when no other events of the same or higher
647	(prepare, check and other idle watchers do not count). That is, as long	1774	priority are pending (prepare, check and other idle watchers do not
648	as your process is busy handling sockets or timeouts (or even signals,	1775	count).
649	imagine) it will not be triggered. But when your process is idle all idle	1776
650	watchers are being called again and again, once per event loop iteration -	1777	That is, as long as your process is busy handling sockets or timeouts
		1778	(or even signals, imagine) of the same or higher priority it will not be
		1779	triggered. But when your process is idle (or only lower-priority watchers
		1780	are pending), the idle watchers are being called once per event loop
651	until stopped, that is, or your process receives more events and becomes	1781	iteration - until stopped, that is, or your process receives more events
652	busy.	1782	and becomes busy again with higher priority stuff.
653		1783
654	The most noteworthy effect is that as long as any idle watchers are	1784	The most noteworthy effect is that as long as any idle watchers are
655	active, the process will not block when waiting for new events.	1785	active, the process will not block when waiting for new events.
656		1786
657	Apart from keeping your process non-blocking (which is a useful	1787	Apart from keeping your process non-blocking (which is a useful
658	effect on its own sometimes), idle watchers are a good place to do	1788	effect on its own sometimes), idle watchers are a good place to do
659	"pseudo-background processing", or delay processing stuff to after the	1789	"pseudo-background processing", or delay processing stuff to after the
660	event loop has handled all outstanding events.	1790	event loop has handled all outstanding events.
661		1791
		1792	=head3 Watcher-Specific Functions and Data Members
		1793
662	=over 4	1794	=over 4
663		1795
664	=item ev_idle_init (ev_signal *, callback)	1796	=item ev_idle_init (ev_signal *, callback)
665		1797
666	Initialises and configures the idle watcher - it has no parameters of any	1798	Initialises and configures the idle watcher - it has no parameters of any
667	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	1799	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
668	believe me.	1800	believe me.
669		1801
670	=back	1802	=back
671		1803
		1804	=head3 Examples
		1805
		1806	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		1807	callback, free it. Also, use no error checking, as usual.
		1808
		1809	static void
		1810	idle_cb (struct ev_loop *loop, struct ev_idle *w, int revents)
		1811	{
		1812	free (w);
		1813	// now do something you wanted to do when the program has
		1814	// no longer anything immediate to do.
		1815	}
		1816
		1817	struct ev_idle *idle_watcher = malloc (sizeof (struct ev_idle));
		1818	ev_idle_init (idle_watcher, idle_cb);
		1819	ev_idle_start (loop, idle_cb);
		1820
		1821
672	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	1822	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
673		1823
674	Prepare and check watchers are usually (but not always) used in tandem:	1824	Prepare and check watchers are usually (but not always) used in tandem:
675	Prepare watchers get invoked before the process blocks and check watchers	1825	prepare watchers get invoked before the process blocks and check watchers
676	afterwards.	1826	afterwards.
677		1827
		1828	You I<must not> call C<ev_loop> or similar functions that enter
		1829	the current event loop from either C<ev_prepare> or C<ev_check>
		1830	watchers. Other loops than the current one are fine, however. The
		1831	rationale behind this is that you do not need to check for recursion in
		1832	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		1833	C<ev_check> so if you have one watcher of each kind they will always be
		1834	called in pairs bracketing the blocking call.
		1835
678	Their main purpose is to integrate other event mechanisms into libev. This	1836	Their main purpose is to integrate other event mechanisms into libev and
679	could be used, for example, to track variable changes, implement your own	1837	their use is somewhat advanced. This could be used, for example, to track
680	watchers, integrate net-snmp or a coroutine library and lots more.	1838	variable changes, implement your own watchers, integrate net-snmp or a
		1839	coroutine library and lots more. They are also occasionally useful if
		1840	you cache some data and want to flush it before blocking (for example,
		1841	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		1842	watcher).
681		1843
682	This is done by examining in each prepare call which file descriptors need	1844	This is done by examining in each prepare call which file descriptors need
683	to be watched by the other library, registering C<ev_io> watchers for	1845	to be watched by the other library, registering C<ev_io> watchers for
684	them and starting an C<ev_timer> watcher for any timeouts (many libraries	1846	them and starting an C<ev_timer> watcher for any timeouts (many libraries
685	provide just this functionality). Then, in the check watcher you check for	1847	provide just this functionality). Then, in the check watcher you check for
686	any events that occured (by checking the pending status of all watchers	1848	any events that occured (by checking the pending status of all watchers
687	and stopping them) and call back into the library. The I/O and timer	1849	and stopping them) and call back into the library. The I/O and timer
688	callbacks will never actually be called (but must be valid neverthelles,	1850	callbacks will never actually be called (but must be valid nevertheless,
689	because you never know, you know?).	1851	because you never know, you know?).
690		1852
691	As another example, the Perl Coro module uses these hooks to integrate	1853	As another example, the Perl Coro module uses these hooks to integrate
692	coroutines into libev programs, by yielding to other active coroutines	1854	coroutines into libev programs, by yielding to other active coroutines
693	during each prepare and only letting the process block if no coroutines	1855	during each prepare and only letting the process block if no coroutines
694	are ready to run (its actually more complicated, it only runs coroutines	1856	are ready to run (it's actually more complicated: it only runs coroutines
695	with priority higher than the event loop and one lower priority once,	1857	with priority higher than or equal to the event loop and one coroutine
696	using idle watchers to keep the event loop from blocking if lower-priority	1858	of lower priority, but only once, using idle watchers to keep the event
697	coroutines exist, thus mapping low-priority coroutines to idle/background	1859	loop from blocking if lower-priority coroutines are active, thus mapping
698	tasks).	1860	low-priority coroutines to idle/background tasks).
		1861
		1862	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		1863	priority, to ensure that they are being run before any other watchers
		1864	after the poll. Also, C<ev_check> watchers (and C<ev_prepare> watchers,
		1865	too) should not activate ("feed") events into libev. While libev fully
		1866	supports this, they will be called before other C<ev_check> watchers
		1867	did their job. As C<ev_check> watchers are often used to embed other
		1868	(non-libev) event loops those other event loops might be in an unusable
		1869	state until their C<ev_check> watcher ran (always remind yourself to
		1870	coexist peacefully with others).
		1871
		1872	=head3 Watcher-Specific Functions and Data Members
699		1873
700	=over 4	1874	=over 4
701		1875
702	=item ev_prepare_init (ev_prepare *, callback)	1876	=item ev_prepare_init (ev_prepare *, callback)
703		1877
…		…
707	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	1881	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
708	macros, but using them is utterly, utterly and completely pointless.	1882	macros, but using them is utterly, utterly and completely pointless.
709		1883
710	=back	1884	=back
711		1885
		1886	=head3 Examples
		1887
		1888	There are a number of principal ways to embed other event loops or modules
		1889	into libev. Here are some ideas on how to include libadns into libev
		1890	(there is a Perl module named C<EV::ADNS> that does this, which you could
		1891	use for an actually working example. Another Perl module named C<EV::Glib>
		1892	embeds a Glib main context into libev, and finally, C<Glib::EV> embeds EV
		1893	into the Glib event loop).
		1894
		1895	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		1896	and in a check watcher, destroy them and call into libadns. What follows
		1897	is pseudo-code only of course. This requires you to either use a low
		1898	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		1899	the callbacks for the IO/timeout watchers might not have been called yet.
		1900
		1901	static ev_io iow [nfd];
		1902	static ev_timer tw;
		1903
		1904	static void
		1905	io_cb (ev_loop *loop, ev_io *w, int revents)
		1906	{
		1907	}
		1908
		1909	// create io watchers for each fd and a timer before blocking
		1910	static void
		1911	adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
		1912	{
		1913	int timeout = 3600000;
		1914	struct pollfd fds [nfd];
		1915	// actual code will need to loop here and realloc etc.
		1916	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		1917
		1918	/* the callback is illegal, but won't be called as we stop during check */
		1919	ev_timer_init (&tw, 0, timeout * 1e-3);
		1920	ev_timer_start (loop, &tw);
		1921
		1922	// create one ev_io per pollfd
		1923	for (int i = 0; i < nfd; ++i)
		1924	{
		1925	ev_io_init (iow + i, io_cb, fds [i].fd,
		1926	((fds [i].events & POLLIN ? EV_READ : 0)
		1927	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		1928
		1929	fds [i].revents = 0;
		1930	ev_io_start (loop, iow + i);
		1931	}
		1932	}
		1933
		1934	// stop all watchers after blocking
		1935	static void
		1936	adns_check_cb (ev_loop *loop, ev_check *w, int revents)
		1937	{
		1938	ev_timer_stop (loop, &tw);
		1939
		1940	for (int i = 0; i < nfd; ++i)
		1941	{
		1942	// set the relevant poll flags
		1943	// could also call adns_processreadable etc. here
		1944	struct pollfd *fd = fds + i;
		1945	int revents = ev_clear_pending (iow + i);
		1946	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		1947	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		1948
		1949	// now stop the watcher
		1950	ev_io_stop (loop, iow + i);
		1951	}
		1952
		1953	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		1954	}
		1955
		1956	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		1957	in the prepare watcher and would dispose of the check watcher.
		1958
		1959	Method 3: If the module to be embedded supports explicit event
		1960	notification (adns does), you can also make use of the actual watcher
		1961	callbacks, and only destroy/create the watchers in the prepare watcher.
		1962
		1963	static void
		1964	timer_cb (EV_P_ ev_timer *w, int revents)
		1965	{
		1966	adns_state ads = (adns_state)w->data;
		1967	update_now (EV_A);
		1968
		1969	adns_processtimeouts (ads, &tv_now);
		1970	}
		1971
		1972	static void
		1973	io_cb (EV_P_ ev_io *w, int revents)
		1974	{
		1975	adns_state ads = (adns_state)w->data;
		1976	update_now (EV_A);
		1977
		1978	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		1979	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		1980	}
		1981
		1982	// do not ever call adns_afterpoll
		1983
		1984	Method 4: Do not use a prepare or check watcher because the module you
		1985	want to embed is too inflexible to support it. Instead, youc na override
		1986	their poll function. The drawback with this solution is that the main
		1987	loop is now no longer controllable by EV. The C<Glib::EV> module does
		1988	this.
		1989
		1990	static gint
		1991	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		1992	{
		1993	int got_events = 0;
		1994
		1995	for (n = 0; n < nfds; ++n)
		1996	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		1997
		1998	if (timeout >= 0)
		1999	// create/start timer
		2000
		2001	// poll
		2002	ev_loop (EV_A_ 0);
		2003
		2004	// stop timer again
		2005	if (timeout >= 0)
		2006	ev_timer_stop (EV_A_ &to);
		2007
		2008	// stop io watchers again - their callbacks should have set
		2009	for (n = 0; n < nfds; ++n)
		2010	ev_io_stop (EV_A_ iow [n]);
		2011
		2012	return got_events;
		2013	}
		2014
		2015
		2016	=head2 C<ev_embed> - when one backend isn't enough...
		2017
		2018	This is a rather advanced watcher type that lets you embed one event loop
		2019	into another (currently only C<ev_io> events are supported in the embedded
		2020	loop, other types of watchers might be handled in a delayed or incorrect
		2021	fashion and must not be used).
		2022
		2023	There are primarily two reasons you would want that: work around bugs and
		2024	prioritise I/O.
		2025
		2026	As an example for a bug workaround, the kqueue backend might only support
		2027	sockets on some platform, so it is unusable as generic backend, but you
		2028	still want to make use of it because you have many sockets and it scales
		2029	so nicely. In this case, you would create a kqueue-based loop and embed it
		2030	into your default loop (which might use e.g. poll). Overall operation will
		2031	be a bit slower because first libev has to poll and then call kevent, but
		2032	at least you can use both at what they are best.
		2033
		2034	As for prioritising I/O: rarely you have the case where some fds have
		2035	to be watched and handled very quickly (with low latency), and even
		2036	priorities and idle watchers might have too much overhead. In this case
		2037	you would put all the high priority stuff in one loop and all the rest in
		2038	a second one, and embed the second one in the first.
		2039
		2040	As long as the watcher is active, the callback will be invoked every time
		2041	there might be events pending in the embedded loop. The callback must then
		2042	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
		2043	their callbacks (you could also start an idle watcher to give the embedded
		2044	loop strictly lower priority for example). You can also set the callback
		2045	to C<0>, in which case the embed watcher will automatically execute the
		2046	embedded loop sweep.
		2047
		2048	As long as the watcher is started it will automatically handle events. The
		2049	callback will be invoked whenever some events have been handled. You can
		2050	set the callback to C<0> to avoid having to specify one if you are not
		2051	interested in that.
		2052
		2053	Also, there have not currently been made special provisions for forking:
		2054	when you fork, you not only have to call C<ev_loop_fork> on both loops,
		2055	but you will also have to stop and restart any C<ev_embed> watchers
		2056	yourself.
		2057
		2058	Unfortunately, not all backends are embeddable, only the ones returned by
		2059	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2060	portable one.
		2061
		2062	So when you want to use this feature you will always have to be prepared
		2063	that you cannot get an embeddable loop. The recommended way to get around
		2064	this is to have a separate variables for your embeddable loop, try to
		2065	create it, and if that fails, use the normal loop for everything.
		2066
		2067	=head3 Watcher-Specific Functions and Data Members
		2068
		2069	=over 4
		2070
		2071	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2072
		2073	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2074
		2075	Configures the watcher to embed the given loop, which must be
		2076	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2077	invoked automatically, otherwise it is the responsibility of the callback
		2078	to invoke it (it will continue to be called until the sweep has been done,
		2079	if you do not want thta, you need to temporarily stop the embed watcher).
		2080
		2081	=item ev_embed_sweep (loop, ev_embed *)
		2082
		2083	Make a single, non-blocking sweep over the embedded loop. This works
		2084	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2085	apropriate way for embedded loops.
		2086
		2087	=item struct ev_loop *other [read-only]
		2088
		2089	The embedded event loop.
		2090
		2091	=back
		2092
		2093	=head3 Examples
		2094
		2095	Example: Try to get an embeddable event loop and embed it into the default
		2096	event loop. If that is not possible, use the default loop. The default
		2097	loop is stored in C<loop_hi>, while the mebeddable loop is stored in
		2098	C<loop_lo> (which is C<loop_hi> in the acse no embeddable loop can be
		2099	used).
		2100
		2101	struct ev_loop *loop_hi = ev_default_init (0);
		2102	struct ev_loop *loop_lo = 0;
		2103	struct ev_embed embed;
		2104
		2105	// see if there is a chance of getting one that works
		2106	// (remember that a flags value of 0 means autodetection)
		2107	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2108	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2109	: 0;
		2110
		2111	// if we got one, then embed it, otherwise default to loop_hi
		2112	if (loop_lo)
		2113	{
		2114	ev_embed_init (&embed, 0, loop_lo);
		2115	ev_embed_start (loop_hi, &embed);
		2116	}
		2117	else
		2118	loop_lo = loop_hi;
		2119
		2120	Example: Check if kqueue is available but not recommended and create
		2121	a kqueue backend for use with sockets (which usually work with any
		2122	kqueue implementation). Store the kqueue/socket-only event loop in
		2123	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2124
		2125	struct ev_loop *loop = ev_default_init (0);
		2126	struct ev_loop *loop_socket = 0;
		2127	struct ev_embed embed;
		2128
		2129	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2130	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2131	{
		2132	ev_embed_init (&embed, 0, loop_socket);
		2133	ev_embed_start (loop, &embed);
		2134	}
		2135
		2136	if (!loop_socket)
		2137	loop_socket = loop;
		2138
		2139	// now use loop_socket for all sockets, and loop for everything else
		2140
		2141
		2142	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2143
		2144	Fork watchers are called when a C<fork ()> was detected (usually because
		2145	whoever is a good citizen cared to tell libev about it by calling
		2146	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2147	event loop blocks next and before C<ev_check> watchers are being called,
		2148	and only in the child after the fork. If whoever good citizen calling
		2149	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2150	handlers will be invoked, too, of course.
		2151
		2152	=head3 Watcher-Specific Functions and Data Members
		2153
		2154	=over 4
		2155
		2156	=item ev_fork_init (ev_signal *, callback)
		2157
		2158	Initialises and configures the fork watcher - it has no parameters of any
		2159	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2160	believe me.
		2161
		2162	=back
		2163
		2164
		2165	=head2 C<ev_async> - how to wake up another event loop
		2166
		2167	In general, you cannot use an C<ev_loop> from multiple threads or other
		2168	asynchronous sources such as signal handlers (as opposed to multiple event
		2169	loops - those are of course safe to use in different threads).
		2170
		2171	Sometimes, however, you need to wake up another event loop you do not
		2172	control, for example because it belongs to another thread. This is what
		2173	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2174	can signal it by calling C<ev_async_send>, which is thread- and signal
		2175	safe.
		2176
		2177	This functionality is very similar to C<ev_signal> watchers, as signals,
		2178	too, are asynchronous in nature, and signals, too, will be compressed
		2179	(i.e. the number of callback invocations may be less than the number of
		2180	C<ev_async_sent> calls).
		2181
		2182	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2183	just the default loop.
		2184
		2185	=head3 Queueing
		2186
		2187	C<ev_async> does not support queueing of data in any way. The reason
		2188	is that the author does not know of a simple (or any) algorithm for a
		2189	multiple-writer-single-reader queue that works in all cases and doesn't
		2190	need elaborate support such as pthreads.
		2191
		2192	That means that if you want to queue data, you have to provide your own
		2193	queue. But at least I can tell you would implement locking around your
		2194	queue:
		2195
		2196	=over 4
		2197
		2198	=item queueing from a signal handler context
		2199
		2200	To implement race-free queueing, you simply add to the queue in the signal
		2201	handler but you block the signal handler in the watcher callback. Here is an example that does that for
		2202	some fictitiuous SIGUSR1 handler:
		2203
		2204	static ev_async mysig;
		2205
		2206	static void
		2207	sigusr1_handler (void)
		2208	{
		2209	sometype data;
		2210
		2211	// no locking etc.
		2212	queue_put (data);
		2213	ev_async_send (EV_DEFAULT_ &mysig);
		2214	}
		2215
		2216	static void
		2217	mysig_cb (EV_P_ ev_async *w, int revents)
		2218	{
		2219	sometype data;
		2220	sigset_t block, prev;
		2221
		2222	sigemptyset (&block);
		2223	sigaddset (&block, SIGUSR1);
		2224	sigprocmask (SIG_BLOCK, &block, &prev);
		2225
		2226	while (queue_get (&data))
		2227	process (data);
		2228
		2229	if (sigismember (&prev, SIGUSR1)
		2230	sigprocmask (SIG_UNBLOCK, &block, 0);
		2231	}
		2232
		2233	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2234	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2235	either...).
		2236
		2237	=item queueing from a thread context
		2238
		2239	The strategy for threads is different, as you cannot (easily) block
		2240	threads but you can easily preempt them, so to queue safely you need to
		2241	employ a traditional mutex lock, such as in this pthread example:
		2242
		2243	static ev_async mysig;
		2244	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2245
		2246	static void
		2247	otherthread (void)
		2248	{
		2249	// only need to lock the actual queueing operation
		2250	pthread_mutex_lock (&mymutex);
		2251	queue_put (data);
		2252	pthread_mutex_unlock (&mymutex);
		2253
		2254	ev_async_send (EV_DEFAULT_ &mysig);
		2255	}
		2256
		2257	static void
		2258	mysig_cb (EV_P_ ev_async *w, int revents)
		2259	{
		2260	pthread_mutex_lock (&mymutex);
		2261
		2262	while (queue_get (&data))
		2263	process (data);
		2264
		2265	pthread_mutex_unlock (&mymutex);
		2266	}
		2267
		2268	=back
		2269
		2270
		2271	=head3 Watcher-Specific Functions and Data Members
		2272
		2273	=over 4
		2274
		2275	=item ev_async_init (ev_async *, callback)
		2276
		2277	Initialises and configures the async watcher - it has no parameters of any
		2278	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
		2279	believe me.
		2280
		2281	=item ev_async_send (loop, ev_async *)
		2282
		2283	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2284	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2285	C<ev_feed_event>, this call is safe to do in other threads, signal or
		2286	similar contexts (see the dicusssion of C<EV_ATOMIC_T> in the embedding
		2287	section below on what exactly this means).
		2288
		2289	This call incurs the overhead of a syscall only once per loop iteration,
		2290	so while the overhead might be noticable, it doesn't apply to repeated
		2291	calls to C<ev_async_send>.
		2292
		2293	=item bool = ev_async_pending (ev_async *)
		2294
		2295	Returns a non-zero value when C<ev_async_send> has been called on the
		2296	watcher but the event has not yet been processed (or even noted) by the
		2297	event loop.
		2298
		2299	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2300	the loop iterates next and checks for the watcher to have become active,
		2301	it will reset the flag again. C<ev_async_pending> can be used to very
		2302	quickly check wether invoking the loop might be a good idea.
		2303
		2304	Not that this does I<not> check wether the watcher itself is pending, only
		2305	wether it has been requested to make this watcher pending.
		2306
		2307	=back
		2308
		2309
712	=head1 OTHER FUNCTIONS	2310	=head1 OTHER FUNCTIONS
713		2311
714	There are some other functions of possible interest. Described. Here. Now.	2312	There are some other functions of possible interest. Described. Here. Now.
715		2313
716	=over 4	2314	=over 4
…		…
718	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2316	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
719		2317
720	This function combines a simple timer and an I/O watcher, calls your	2318	This function combines a simple timer and an I/O watcher, calls your
721	callback on whichever event happens first and automatically stop both	2319	callback on whichever event happens first and automatically stop both
722	watchers. This is useful if you want to wait for a single event on an fd	2320	watchers. This is useful if you want to wait for a single event on an fd
723	or timeout without havign to allocate/configure/start/stop/free one or	2321	or timeout without having to allocate/configure/start/stop/free one or
724	more watchers yourself.	2322	more watchers yourself.
725		2323
726	If C<fd> is less than 0, then no I/O watcher will be started and events	2324	If C<fd> is less than 0, then no I/O watcher will be started and events
727	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2325	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and
728	C<events> set will be craeted and started.	2326	C<events> set will be craeted and started.
…		…
731	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2329	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
732	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2330	repeat = 0) will be started. While C<0> is a valid timeout, it is of
733	dubious value.	2331	dubious value.
734		2332
735	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	2333	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
736	passed an events set like normal event callbacks (with a combination of	2334	passed an C<revents> set like normal event callbacks (a combination of
737	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2335	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
738	value passed to C<ev_once>:	2336	value passed to C<ev_once>:
739		2337
740	static void stdin_ready (int revents, void *arg)	2338	static void stdin_ready (int revents, void *arg)
741	{	2339	{
…		…
745	/* stdin might have data for us, joy! */;	2343	/* stdin might have data for us, joy! */;
746	}	2344	}
747		2345
748	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2346	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
749		2347
750	=item ev_feed_event (loop, watcher, int events)	2348	=item ev_feed_event (ev_loop *, watcher *, int revents)
751		2349
752	Feeds the given event set into the event loop, as if the specified event	2350	Feeds the given event set into the event loop, as if the specified event
753	had happened for the specified watcher (which must be a pointer to an	2351	had happened for the specified watcher (which must be a pointer to an
754	initialised but not necessarily started event watcher).	2352	initialised but not necessarily started event watcher).
755		2353
756	=item ev_feed_fd_event (loop, int fd, int revents)	2354	=item ev_feed_fd_event (ev_loop *, int fd, int revents)
757		2355
758	Feed an event on the given fd, as if a file descriptor backend detected	2356	Feed an event on the given fd, as if a file descriptor backend detected
759	the given events it.	2357	the given events it.
760		2358
761	=item ev_feed_signal_event (loop, int signum)	2359	=item ev_feed_signal_event (ev_loop *loop, int signum)
762		2360
763	Feed an event as if the given signal occured (loop must be the default loop!).	2361	Feed an event as if the given signal occured (C<loop> must be the default
		2362	loop!).
764		2363
765	=back	2364	=back
766		2365
		2366
		2367	=head1 LIBEVENT EMULATION
		2368
		2369	Libev offers a compatibility emulation layer for libevent. It cannot
		2370	emulate the internals of libevent, so here are some usage hints:
		2371
		2372	=over 4
		2373
		2374	=item * Use it by including <event.h>, as usual.
		2375
		2376	=item * The following members are fully supported: ev_base, ev_callback,
		2377	ev_arg, ev_fd, ev_res, ev_events.
		2378
		2379	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		2380	maintained by libev, it does not work exactly the same way as in libevent (consider
		2381	it a private API).
		2382
		2383	=item * Priorities are not currently supported. Initialising priorities
		2384	will fail and all watchers will have the same priority, even though there
		2385	is an ev_pri field.
		2386
		2387	=item * Other members are not supported.
		2388
		2389	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		2390	to use the libev header file and library.
		2391
		2392	=back
		2393
		2394	=head1 C++ SUPPORT
		2395
		2396	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		2397	you to use some convinience methods to start/stop watchers and also change
		2398	the callback model to a model using method callbacks on objects.
		2399
		2400	To use it,
		2401
		2402	#include <ev++.h>
		2403
		2404	This automatically includes F<ev.h> and puts all of its definitions (many
		2405	of them macros) into the global namespace. All C++ specific things are
		2406	put into the C<ev> namespace. It should support all the same embedding
		2407	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		2408
		2409	Care has been taken to keep the overhead low. The only data member the C++
		2410	classes add (compared to plain C-style watchers) is the event loop pointer
		2411	that the watcher is associated with (or no additional members at all if
		2412	you disable C<EV_MULTIPLICITY> when embedding libev).
		2413
		2414	Currently, functions, and static and non-static member functions can be
		2415	used as callbacks. Other types should be easy to add as long as they only
		2416	need one additional pointer for context. If you need support for other
		2417	types of functors please contact the author (preferably after implementing
		2418	it).
		2419
		2420	Here is a list of things available in the C<ev> namespace:
		2421
		2422	=over 4
		2423
		2424	=item C<ev::READ>, C<ev::WRITE> etc.
		2425
		2426	These are just enum values with the same values as the C<EV_READ> etc.
		2427	macros from F<ev.h>.
		2428
		2429	=item C<ev::tstamp>, C<ev::now>
		2430
		2431	Aliases to the same types/functions as with the C<ev_> prefix.
		2432
		2433	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		2434
		2435	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		2436	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		2437	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		2438	defines by many implementations.
		2439
		2440	All of those classes have these methods:
		2441
		2442	=over 4
		2443
		2444	=item ev::TYPE::TYPE ()
		2445
		2446	=item ev::TYPE::TYPE (struct ev_loop *)
		2447
		2448	=item ev::TYPE::~TYPE
		2449
		2450	The constructor (optionally) takes an event loop to associate the watcher
		2451	with. If it is omitted, it will use C<EV_DEFAULT>.
		2452
		2453	The constructor calls C<ev_init> for you, which means you have to call the
		2454	C<set> method before starting it.
		2455
		2456	It will not set a callback, however: You have to call the templated C<set>
		2457	method to set a callback before you can start the watcher.
		2458
		2459	(The reason why you have to use a method is a limitation in C++ which does
		2460	not allow explicit template arguments for constructors).
		2461
		2462	The destructor automatically stops the watcher if it is active.
		2463
		2464	=item w->set<class, &class::method> (object *)
		2465
		2466	This method sets the callback method to call. The method has to have a
		2467	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		2468	first argument and the C<revents> as second. The object must be given as
		2469	parameter and is stored in the C<data> member of the watcher.
		2470
		2471	This method synthesizes efficient thunking code to call your method from
		2472	the C callback that libev requires. If your compiler can inline your
		2473	callback (i.e. it is visible to it at the place of the C<set> call and
		2474	your compiler is good :), then the method will be fully inlined into the
		2475	thunking function, making it as fast as a direct C callback.
		2476
		2477	Example: simple class declaration and watcher initialisation
		2478
		2479	struct myclass
		2480	{
		2481	void io_cb (ev::io &w, int revents) { }
		2482	}
		2483
		2484	myclass obj;
		2485	ev::io iow;
		2486	iow.set <myclass, &myclass::io_cb> (&obj);
		2487
		2488	=item w->set<function> (void *data = 0)
		2489
		2490	Also sets a callback, but uses a static method or plain function as
		2491	callback. The optional C<data> argument will be stored in the watcher's
		2492	C<data> member and is free for you to use.
		2493
		2494	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		2495
		2496	See the method-C<set> above for more details.
		2497
		2498	Example:
		2499
		2500	static void io_cb (ev::io &w, int revents) { }
		2501	iow.set <io_cb> ();
		2502
		2503	=item w->set (struct ev_loop *)
		2504
		2505	Associates a different C<struct ev_loop> with this watcher. You can only
		2506	do this when the watcher is inactive (and not pending either).
		2507
		2508	=item w->set ([args])
		2509
		2510	Basically the same as C<ev_TYPE_set>, with the same args. Must be
		2511	called at least once. Unlike the C counterpart, an active watcher gets
		2512	automatically stopped and restarted when reconfiguring it with this
		2513	method.
		2514
		2515	=item w->start ()
		2516
		2517	Starts the watcher. Note that there is no C<loop> argument, as the
		2518	constructor already stores the event loop.
		2519
		2520	=item w->stop ()
		2521
		2522	Stops the watcher if it is active. Again, no C<loop> argument.
		2523
		2524	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		2525
		2526	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		2527	C<ev_TYPE_again> function.
		2528
		2529	=item w->sweep () (C<ev::embed> only)
		2530
		2531	Invokes C<ev_embed_sweep>.
		2532
		2533	=item w->update () (C<ev::stat> only)
		2534
		2535	Invokes C<ev_stat_stat>.
		2536
		2537	=back
		2538
		2539	=back
		2540
		2541	Example: Define a class with an IO and idle watcher, start one of them in
		2542	the constructor.
		2543
		2544	class myclass
		2545	{
		2546	ev::io io; void io_cb (ev::io &w, int revents);
		2547	ev:idle idle void idle_cb (ev::idle &w, int revents);
		2548
		2549	myclass (int fd)
		2550	{
		2551	io .set <myclass, &myclass::io_cb > (this);
		2552	idle.set <myclass, &myclass::idle_cb> (this);
		2553
		2554	io.start (fd, ev::READ);
		2555	}
		2556	};
		2557
		2558
		2559	=head1 OTHER LANGUAGE BINDINGS
		2560
		2561	Libev does not offer other language bindings itself, but bindings for a
		2562	numbe rof languages exist in the form of third-party packages. If you know
		2563	any interesting language binding in addition to the ones listed here, drop
		2564	me a note.
		2565
		2566	=over 4
		2567
		2568	=item Perl
		2569
		2570	The EV module implements the full libev API and is actually used to test
		2571	libev. EV is developed together with libev. Apart from the EV core module,
		2572	there are additional modules that implement libev-compatible interfaces
		2573	to C<libadns> (C<EV::ADNS>), C<Net::SNMP> (C<Net::SNMP::EV>) and the
		2574	C<libglib> event core (C<Glib::EV> and C<EV::Glib>).
		2575
		2576	It can be found and installed via CPAN, its homepage is found at
		2577	L<http://software.schmorp.de/pkg/EV>.
		2578
		2579	=item Ruby
		2580
		2581	Tony Arcieri has written a ruby extension that offers access to a subset
		2582	of the libev API and adds filehandle abstractions, asynchronous DNS and
		2583	more on top of it. It can be found via gem servers. Its homepage is at
		2584	L<http://rev.rubyforge.org/>.
		2585
		2586	=item D
		2587
		2588	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		2589	be found at L<http://git.llucax.com.ar/?p=software/ev.d.git;a=summary>.
		2590
		2591	=back
		2592
		2593
		2594	=head1 MACRO MAGIC
		2595
		2596	Libev can be compiled with a variety of options, the most fundamantal
		2597	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		2598	functions and callbacks have an initial C<struct ev_loop *> argument.
		2599
		2600	To make it easier to write programs that cope with either variant, the
		2601	following macros are defined:
		2602
		2603	=over 4
		2604
		2605	=item C<EV_A>, C<EV_A_>
		2606
		2607	This provides the loop I<argument> for functions, if one is required ("ev
		2608	loop argument"). The C<EV_A> form is used when this is the sole argument,
		2609	C<EV_A_> is used when other arguments are following. Example:
		2610
		2611	ev_unref (EV_A);
		2612	ev_timer_add (EV_A_ watcher);
		2613	ev_loop (EV_A_ 0);
		2614
		2615	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		2616	which is often provided by the following macro.
		2617
		2618	=item C<EV_P>, C<EV_P_>
		2619
		2620	This provides the loop I<parameter> for functions, if one is required ("ev
		2621	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		2622	C<EV_P_> is used when other parameters are following. Example:
		2623
		2624	// this is how ev_unref is being declared
		2625	static void ev_unref (EV_P);
		2626
		2627	// this is how you can declare your typical callback
		2628	static void cb (EV_P_ ev_timer *w, int revents)
		2629
		2630	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		2631	suitable for use with C<EV_A>.
		2632
		2633	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		2634
		2635	Similar to the other two macros, this gives you the value of the default
		2636	loop, if multiple loops are supported ("ev loop default").
		2637
		2638	=back
		2639
		2640	Example: Declare and initialise a check watcher, utilising the above
		2641	macros so it will work regardless of whether multiple loops are supported
		2642	or not.
		2643
		2644	static void
		2645	check_cb (EV_P_ ev_timer *w, int revents)
		2646	{
		2647	ev_check_stop (EV_A_ w);
		2648	}
		2649
		2650	ev_check check;
		2651	ev_check_init (&check, check_cb);
		2652	ev_check_start (EV_DEFAULT_ &check);
		2653	ev_loop (EV_DEFAULT_ 0);
		2654
		2655	=head1 EMBEDDING
		2656
		2657	Libev can (and often is) directly embedded into host
		2658	applications. Examples of applications that embed it include the Deliantra
		2659	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		2660	and rxvt-unicode.
		2661
		2662	The goal is to enable you to just copy the necessary files into your
		2663	source directory without having to change even a single line in them, so
		2664	you can easily upgrade by simply copying (or having a checked-out copy of
		2665	libev somewhere in your source tree).
		2666
		2667	=head2 FILESETS
		2668
		2669	Depending on what features you need you need to include one or more sets of files
		2670	in your app.
		2671
		2672	=head3 CORE EVENT LOOP
		2673
		2674	To include only the libev core (all the C<ev_*> functions), with manual
		2675	configuration (no autoconf):
		2676
		2677	#define EV_STANDALONE 1
		2678	#include "ev.c"
		2679
		2680	This will automatically include F<ev.h>, too, and should be done in a
		2681	single C source file only to provide the function implementations. To use
		2682	it, do the same for F<ev.h> in all files wishing to use this API (best
		2683	done by writing a wrapper around F<ev.h> that you can include instead and
		2684	where you can put other configuration options):
		2685
		2686	#define EV_STANDALONE 1
		2687	#include "ev.h"
		2688
		2689	Both header files and implementation files can be compiled with a C++
		2690	compiler (at least, thats a stated goal, and breakage will be treated
		2691	as a bug).
		2692
		2693	You need the following files in your source tree, or in a directory
		2694	in your include path (e.g. in libev/ when using -Ilibev):
		2695
		2696	ev.h
		2697	ev.c
		2698	ev_vars.h
		2699	ev_wrap.h
		2700
		2701	ev_win32.c required on win32 platforms only
		2702
		2703	ev_select.c only when select backend is enabled (which is enabled by default)
		2704	ev_poll.c only when poll backend is enabled (disabled by default)
		2705	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		2706	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		2707	ev_port.c only when the solaris port backend is enabled (disabled by default)
		2708
		2709	F<ev.c> includes the backend files directly when enabled, so you only need
		2710	to compile this single file.
		2711
		2712	=head3 LIBEVENT COMPATIBILITY API
		2713
		2714	To include the libevent compatibility API, also include:
		2715
		2716	#include "event.c"
		2717
		2718	in the file including F<ev.c>, and:
		2719
		2720	#include "event.h"
		2721
		2722	in the files that want to use the libevent API. This also includes F<ev.h>.
		2723
		2724	You need the following additional files for this:
		2725
		2726	event.h
		2727	event.c
		2728
		2729	=head3 AUTOCONF SUPPORT
		2730
		2731	Instead of using C<EV_STANDALONE=1> and providing your config in
		2732	whatever way you want, you can also C<m4_include([libev.m4])> in your
		2733	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		2734	include F<config.h> and configure itself accordingly.
		2735
		2736	For this of course you need the m4 file:
		2737
		2738	libev.m4
		2739
		2740	=head2 PREPROCESSOR SYMBOLS/MACROS
		2741
		2742	Libev can be configured via a variety of preprocessor symbols you have to define
		2743	before including any of its files. The default is not to build for multiplicity
		2744	and only include the select backend.
		2745
		2746	=over 4
		2747
		2748	=item EV_STANDALONE
		2749
		2750	Must always be C<1> if you do not use autoconf configuration, which
		2751	keeps libev from including F<config.h>, and it also defines dummy
		2752	implementations for some libevent functions (such as logging, which is not
		2753	supported). It will also not define any of the structs usually found in
		2754	F<event.h> that are not directly supported by the libev core alone.
		2755
		2756	=item EV_USE_MONOTONIC
		2757
		2758	If defined to be C<1>, libev will try to detect the availability of the
		2759	monotonic clock option at both compiletime and runtime. Otherwise no use
		2760	of the monotonic clock option will be attempted. If you enable this, you
		2761	usually have to link against librt or something similar. Enabling it when
		2762	the functionality isn't available is safe, though, although you have
		2763	to make sure you link against any libraries where the C<clock_gettime>
		2764	function is hiding in (often F<-lrt>).
		2765
		2766	=item EV_USE_REALTIME
		2767
		2768	If defined to be C<1>, libev will try to detect the availability of the
		2769	realtime clock option at compiletime (and assume its availability at
		2770	runtime if successful). Otherwise no use of the realtime clock option will
		2771	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
		2772	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
		2773	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
		2774
		2775	=item EV_USE_NANOSLEEP
		2776
		2777	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		2778	and will use it for delays. Otherwise it will use C<select ()>.
		2779
		2780	=item EV_USE_SELECT
		2781
		2782	If undefined or defined to be C<1>, libev will compile in support for the
		2783	C<select>(2) backend. No attempt at autodetection will be done: if no
		2784	other method takes over, select will be it. Otherwise the select backend
		2785	will not be compiled in.
		2786
		2787	=item EV_SELECT_USE_FD_SET
		2788
		2789	If defined to C<1>, then the select backend will use the system C<fd_set>
		2790	structure. This is useful if libev doesn't compile due to a missing
		2791	C<NFDBITS> or C<fd_mask> definition or it misguesses the bitset layout on
		2792	exotic systems. This usually limits the range of file descriptors to some
		2793	low limit such as 1024 or might have other limitations (winsocket only
		2794	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
		2795	influence the size of the C<fd_set> used.
		2796
		2797	=item EV_SELECT_IS_WINSOCKET
		2798
		2799	When defined to C<1>, the select backend will assume that
		2800	select/socket/connect etc. don't understand file descriptors but
		2801	wants osf handles on win32 (this is the case when the select to
		2802	be used is the winsock select). This means that it will call
		2803	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		2804	it is assumed that all these functions actually work on fds, even
		2805	on win32. Should not be defined on non-win32 platforms.
		2806
		2807	=item EV_FD_TO_WIN32_HANDLE
		2808
		2809	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		2810	file descriptors to socket handles. When not defining this symbol (the
		2811	default), then libev will call C<_get_osfhandle>, which is usually
		2812	correct. In some cases, programs use their own file descriptor management,
		2813	in which case they can provide this function to map fds to socket handles.
		2814
		2815	=item EV_USE_POLL
		2816
		2817	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		2818	backend. Otherwise it will be enabled on non-win32 platforms. It
		2819	takes precedence over select.
		2820
		2821	=item EV_USE_EPOLL
		2822
		2823	If defined to be C<1>, libev will compile in support for the Linux
		2824	C<epoll>(7) backend. Its availability will be detected at runtime,
		2825	otherwise another method will be used as fallback. This is the
		2826	preferred backend for GNU/Linux systems.
		2827
		2828	=item EV_USE_KQUEUE
		2829
		2830	If defined to be C<1>, libev will compile in support for the BSD style
		2831	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		2832	otherwise another method will be used as fallback. This is the preferred
		2833	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		2834	supports some types of fds correctly (the only platform we found that
		2835	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		2836	not be used unless explicitly requested. The best way to use it is to find
		2837	out whether kqueue supports your type of fd properly and use an embedded
		2838	kqueue loop.
		2839
		2840	=item EV_USE_PORT
		2841
		2842	If defined to be C<1>, libev will compile in support for the Solaris
		2843	10 port style backend. Its availability will be detected at runtime,
		2844	otherwise another method will be used as fallback. This is the preferred
		2845	backend for Solaris 10 systems.
		2846
		2847	=item EV_USE_DEVPOLL
		2848
		2849	reserved for future expansion, works like the USE symbols above.
		2850
		2851	=item EV_USE_INOTIFY
		2852
		2853	If defined to be C<1>, libev will compile in support for the Linux inotify
		2854	interface to speed up C<ev_stat> watchers. Its actual availability will
		2855	be detected at runtime.
		2856
		2857	=item EV_ATOMIC_T
		2858
		2859	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		2860	access is atomic with respect to other threads or signal contexts. No such
		2861	type is easily found in the C language, so you can provide your own type
		2862	that you know is safe for your purposes. It is used both for signal handler "locking"
		2863	as well as for signal and thread safety in C<ev_async> watchers.
		2864
		2865	In the absense of this define, libev will use C<sig_atomic_t volatile>
		2866	(from F<signal.h>), which is usually good enough on most platforms.
		2867
		2868	=item EV_H
		2869
		2870	The name of the F<ev.h> header file used to include it. The default if
		2871	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		2872	used to virtually rename the F<ev.h> header file in case of conflicts.
		2873
		2874	=item EV_CONFIG_H
		2875
		2876	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		2877	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		2878	C<EV_H>, above.
		2879
		2880	=item EV_EVENT_H
		2881
		2882	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		2883	of how the F<event.h> header can be found, the default is C<"event.h">.
		2884
		2885	=item EV_PROTOTYPES
		2886
		2887	If defined to be C<0>, then F<ev.h> will not define any function
		2888	prototypes, but still define all the structs and other symbols. This is
		2889	occasionally useful if you want to provide your own wrapper functions
		2890	around libev functions.
		2891
		2892	=item EV_MULTIPLICITY
		2893
		2894	If undefined or defined to C<1>, then all event-loop-specific functions
		2895	will have the C<struct ev_loop *> as first argument, and you can create
		2896	additional independent event loops. Otherwise there will be no support
		2897	for multiple event loops and there is no first event loop pointer
		2898	argument. Instead, all functions act on the single default loop.
		2899
		2900	=item EV_MINPRI
		2901
		2902	=item EV_MAXPRI
		2903
		2904	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		2905	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		2906	provide for more priorities by overriding those symbols (usually defined
		2907	to be C<-2> and C<2>, respectively).
		2908
		2909	When doing priority-based operations, libev usually has to linearly search
		2910	all the priorities, so having many of them (hundreds) uses a lot of space
		2911	and time, so using the defaults of five priorities (-2 .. +2) is usually
		2912	fine.
		2913
		2914	If your embedding app does not need any priorities, defining these both to
		2915	C<0> will save some memory and cpu.
		2916
		2917	=item EV_PERIODIC_ENABLE
		2918
		2919	If undefined or defined to be C<1>, then periodic timers are supported. If
		2920	defined to be C<0>, then they are not. Disabling them saves a few kB of
		2921	code.
		2922
		2923	=item EV_IDLE_ENABLE
		2924
		2925	If undefined or defined to be C<1>, then idle watchers are supported. If
		2926	defined to be C<0>, then they are not. Disabling them saves a few kB of
		2927	code.
		2928
		2929	=item EV_EMBED_ENABLE
		2930
		2931	If undefined or defined to be C<1>, then embed watchers are supported. If
		2932	defined to be C<0>, then they are not.
		2933
		2934	=item EV_STAT_ENABLE
		2935
		2936	If undefined or defined to be C<1>, then stat watchers are supported. If
		2937	defined to be C<0>, then they are not.
		2938
		2939	=item EV_FORK_ENABLE
		2940
		2941	If undefined or defined to be C<1>, then fork watchers are supported. If
		2942	defined to be C<0>, then they are not.
		2943
		2944	=item EV_ASYNC_ENABLE
		2945
		2946	If undefined or defined to be C<1>, then async watchers are supported. If
		2947	defined to be C<0>, then they are not.
		2948
		2949	=item EV_MINIMAL
		2950
		2951	If you need to shave off some kilobytes of code at the expense of some
		2952	speed, define this symbol to C<1>. Currently only used for gcc to override
		2953	some inlining decisions, saves roughly 30% codesize of amd64.
		2954
		2955	=item EV_PID_HASHSIZE
		2956
		2957	C<ev_child> watchers use a small hash table to distribute workload by
		2958	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		2959	than enough. If you need to manage thousands of children you might want to
		2960	increase this value (I<must> be a power of two).
		2961
		2962	=item EV_INOTIFY_HASHSIZE
		2963
		2964	C<ev_stat> watchers use a small hash table to distribute workload by
		2965	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		2966	usually more than enough. If you need to manage thousands of C<ev_stat>
		2967	watchers you might want to increase this value (I<must> be a power of
		2968	two).
		2969
		2970	=item EV_COMMON
		2971
		2972	By default, all watchers have a C<void *data> member. By redefining
		2973	this macro to a something else you can include more and other types of
		2974	members. You have to define it each time you include one of the files,
		2975	though, and it must be identical each time.
		2976
		2977	For example, the perl EV module uses something like this:
		2978
		2979	#define EV_COMMON \
		2980	SV *self; /* contains this struct */ \
		2981	SV *cb_sv, *fh /* note no trailing ";" */
		2982
		2983	=item EV_CB_DECLARE (type)
		2984
		2985	=item EV_CB_INVOKE (watcher, revents)
		2986
		2987	=item ev_set_cb (ev, cb)
		2988
		2989	Can be used to change the callback member declaration in each watcher,
		2990	and the way callbacks are invoked and set. Must expand to a struct member
		2991	definition and a statement, respectively. See the F<ev.h> header file for
		2992	their default definitions. One possible use for overriding these is to
		2993	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		2994	method calls instead of plain function calls in C++.
		2995
		2996	=head2 EXPORTED API SYMBOLS
		2997
		2998	If you need to re-export the API (e.g. via a dll) and you need a list of
		2999	exported symbols, you can use the provided F<Symbol.*> files which list
		3000	all public symbols, one per line:
		3001
		3002	Symbols.ev for libev proper
		3003	Symbols.event for the libevent emulation
		3004
		3005	This can also be used to rename all public symbols to avoid clashes with
		3006	multiple versions of libev linked together (which is obviously bad in
		3007	itself, but sometimes it is inconvinient to avoid this).
		3008
		3009	A sed command like this will create wrapper C<#define>'s that you need to
		3010	include before including F<ev.h>:
		3011
		3012	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3013
		3014	This would create a file F<wrap.h> which essentially looks like this:
		3015
		3016	#define ev_backend myprefix_ev_backend
		3017	#define ev_check_start myprefix_ev_check_start
		3018	#define ev_check_stop myprefix_ev_check_stop
		3019	...
		3020
		3021	=head2 EXAMPLES
		3022
		3023	For a real-world example of a program the includes libev
		3024	verbatim, you can have a look at the EV perl module
		3025	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3026	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3027	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3028	will be compiled. It is pretty complex because it provides its own header
		3029	file.
		3030
		3031	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3032	that everybody includes and which overrides some configure choices:
		3033
		3034	#define EV_MINIMAL 1
		3035	#define EV_USE_POLL 0
		3036	#define EV_MULTIPLICITY 0
		3037	#define EV_PERIODIC_ENABLE 0
		3038	#define EV_STAT_ENABLE 0
		3039	#define EV_FORK_ENABLE 0
		3040	#define EV_CONFIG_H <config.h>
		3041	#define EV_MINPRI 0
		3042	#define EV_MAXPRI 0
		3043
		3044	#include "ev++.h"
		3045
		3046	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3047
		3048	#include "ev_cpp.h"
		3049	#include "ev.c"
		3050
		3051
		3052	=head1 COMPLEXITIES
		3053
		3054	In this section the complexities of (many of) the algorithms used inside
		3055	libev will be explained. For complexity discussions about backends see the
		3056	documentation for C<ev_default_init>.
		3057
		3058	All of the following are about amortised time: If an array needs to be
		3059	extended, libev needs to realloc and move the whole array, but this
		3060	happens asymptotically never with higher number of elements, so O(1) might
		3061	mean it might do a lengthy realloc operation in rare cases, but on average
		3062	it is much faster and asymptotically approaches constant time.
		3063
		3064	=over 4
		3065
		3066	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		3067
		3068	This means that, when you have a watcher that triggers in one hour and
		3069	there are 100 watchers that would trigger before that then inserting will
		3070	have to skip roughly seven (C<ld 100>) of these watchers.
		3071
		3072	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		3073
		3074	That means that changing a timer costs less than removing/adding them
		3075	as only the relative motion in the event queue has to be paid for.
		3076
		3077	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		3078
		3079	These just add the watcher into an array or at the head of a list.
		3080
		3081	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		3082
		3083	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		3084
		3085	These watchers are stored in lists then need to be walked to find the
		3086	correct watcher to remove. The lists are usually short (you don't usually
		3087	have many watchers waiting for the same fd or signal).
		3088
		3089	=item Finding the next timer in each loop iteration: O(1)
		3090
		3091	By virtue of using a binary heap, the next timer is always found at the
		3092	beginning of the storage array.
		3093
		3094	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3095
		3096	A change means an I/O watcher gets started or stopped, which requires
		3097	libev to recalculate its status (and possibly tell the kernel, depending
		3098	on backend and wether C<ev_io_set> was used).
		3099
		3100	=item Activating one watcher (putting it into the pending state): O(1)
		3101
		3102	=item Priority handling: O(number_of_priorities)
		3103
		3104	Priorities are implemented by allocating some space for each
		3105	priority. When doing priority-based operations, libev usually has to
		3106	linearly search all the priorities, but starting/stopping and activating
		3107	watchers becomes O(1) w.r.t. priority handling.
		3108
		3109	=item Sending an ev_async: O(1)
		3110
		3111	=item Processing ev_async_send: O(number_of_async_watchers)
		3112
		3113	=item Processing signals: O(max_signal_number)
		3114
		3115	Sending involves a syscall I<iff> there were no other C<ev_async_send>
		3116	calls in the current loop iteration. Checking for async and signal events
		3117	involves iterating over all running async watchers or all signal numbers.
		3118
		3119	=back
		3120
		3121
		3122	=head1 Win32 platform limitations and workarounds
		3123
		3124	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3125	requires, and its I/O model is fundamentally incompatible with the POSIX
		3126	model. Libev still offers limited functionality on this platform in
		3127	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3128	descriptors. This only applies when using Win32 natively, not when using
		3129	e.g. cygwin.
		3130
		3131	There is no supported compilation method available on windows except
		3132	embedding it into other applications.
		3133
		3134	Due to the many, low, and arbitrary limits on the win32 platform and the
		3135	abysmal performance of winsockets, using a large number of sockets is not
		3136	recommended (and not reasonable). If your program needs to use more than
		3137	a hundred or so sockets, then likely it needs to use a totally different
		3138	implementation for windows, as libev offers the POSIX model, which cannot
		3139	be implemented efficiently on windows (microsoft monopoly games).
		3140
		3141	=over 4
		3142
		3143	=item The winsocket select function
		3144
		3145	The winsocket C<select> function doesn't follow POSIX in that it requires
		3146	socket I<handles> and not socket I<file descriptors>. This makes select
		3147	very inefficient, and also requires a mapping from file descriptors
		3148	to socket handles. See the discussion of the C<EV_SELECT_USE_FD_SET>,
		3149	C<EV_SELECT_IS_WINSOCKET> and C<EV_FD_TO_WIN32_HANDLE> preprocessor
		3150	symbols for more info.
		3151
		3152	The configuration for a "naked" win32 using the microsoft runtime
		3153	libraries and raw winsocket select is:
		3154
		3155	#define EV_USE_SELECT 1
		3156	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3157
		3158	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3159	complexity in the O(n²) range when using win32.
		3160
		3161	=item Limited number of file descriptors
		3162
		3163	Windows has numerous arbitrary (and low) limits on things. Early versions
		3164	of winsocket's select only supported waiting for a max. of C<64> handles
		3165	(probably owning to the fact that all windows kernels can only wait for
		3166	C<64> things at the same time internally; microsoft recommends spawning a
		3167	chain of threads and wait for 63 handles and the previous thread in each).
		3168
		3169	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3170	to some high number (e.g. C<2048>) before compiling the winsocket select
		3171	call (which might be in libev or elsewhere, for example, perl does its own
		3172	select emulation on windows).
		3173
		3174	Another limit is the number of file descriptors in the microsoft runtime
		3175	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3176	or something like this inside microsoft). You can increase this by calling
		3177	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3178	arbitrary limit), but is broken in many versions of the microsoft runtime
		3179	libraries.
		3180
		3181	This might get you to about C<512> or C<2048> sockets (depending on
		3182	windows version and/or the phase of the moon). To get more, you need to
		3183	wrap all I/O functions and provide your own fd management, but the cost of
		3184	calling select (O(n²)) will likely make this unworkable.
		3185
		3186	=back
		3187
		3188
767	=head1 AUTHOR	3189	=head1 AUTHOR
768		3190
769	Marc Lehmann <libev@schmorp.de>.	3191	Marc Lehmann <libev@schmorp.de>.
770		3192

Diff Legend

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