ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.1 by root, Mon Nov 12 07:58:13 2007 UTC vs.
Revision 1.202 by root, Fri Oct 24 08:30:01 2008 UTC

…		…
2		2
3	libev - a high performance full-featured event loop written in C	3	libev - a high performance full-featured event loop written in C
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	#include <ev.h>	7	#include <ev.h>
		8
		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_ 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_ 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	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	}
8		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://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		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 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.
18		78
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).	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>).
31		96
		97	It also is quite fast (see this
		98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
		99	for example).
		100
32	=head1 CONVENTIONS	101	=head2 CONVENTIONS
33		102
34	Libev is very configurable. In this manual the default configuration	103	Libev is very configurable. In this manual the default (and most common)
35	will be described, which supports multiple event loops. For more info	104	configuration will be described, which supports multiple event loops. For
36	about various configuraiton options please have a look at the file	105	more info about various configuration options please have a look at
37	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
38	support for multiple event loops, then all functions taking an initial	107	for multiple event loops, then all functions taking an initial argument of
39	argument of name C<loop> (which is always of type C<struct ev_loop *>)	108	name C<loop> (which is always of type C<ev_loop *>) will not have
40	will not have this argument.	109	this argument.
41		110
42	=head1 TIME AND OTHER GLOBAL FUNCTIONS	111	=head2 TIME REPRESENTATION
43		112
44	Libev represents time as a single floating point number. This type is	113	Libev represents time as a single floating point number, representing the
		114	(fractional) number of seconds since the (POSIX) epoch (somewhere near
		115	the beginning of 1970, details are complicated, don't ask). This type is
45	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
46	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 floating point value. Unlike the name
		119	component C<stamp> might indicate, it is also used for time differences
		120	throughout libev.
		121
		122	=head1 ERROR HANDLING
		123
		124	Libev knows three classes of errors: operating system errors, usage errors
		125	and internal errors (bugs).
		126
		127	When libev catches an operating system error it cannot handle (for example
		128	a system call indicating a condition libev cannot fix), it calls the callback
		129	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		130	abort. The default is to print a diagnostic message and to call C<abort
		131	()>.
		132
		133	When libev detects a usage error such as a negative timer interval, then
		134	it will print a diagnostic message and abort (via the C<assert> mechanism,
		135	so C<NDEBUG> will disable this checking): these are programming errors in
		136	the libev caller and need to be fixed there.
		137
		138	Libev also has a few internal error-checking C<assert>ions, and also has
		139	extensive consistency checking code. These do not trigger under normal
		140	circumstances, as they indicate either a bug in libev or worse.
		141
		142
		143	=head1 GLOBAL FUNCTIONS
		144
		145	These functions can be called anytime, even before initialising the
		146	library in any way.
47		147
48	=over 4	148	=over 4
49		149
50	=item ev_tstamp ev_time ()	150	=item ev_tstamp ev_time ()
51		151
52	Returns the current time as libev would use it.	152	Returns the current time as libev would use it. Please note that the
		153	C<ev_now> function is usually faster and also often returns the timestamp
		154	you actually want to know.
		155
		156	=item ev_sleep (ev_tstamp interval)
		157
		158	Sleep for the given interval: The current thread will be blocked until
		159	either it is interrupted or the given time interval has passed. Basically
		160	this is a sub-second-resolution C<sleep ()>.
53		161
54	=item int ev_version_major ()	162	=item int ev_version_major ()
55		163
56	=item int ev_version_minor ()	164	=item int ev_version_minor ()
57		165
58	You can find out the major and minor version numbers of the library	166	You can find out the major and minor ABI version numbers of the library
59	you linked against by calling the functions C<ev_version_major> and	167	you linked against by calling the functions C<ev_version_major> and
60	C<ev_version_minor>. If you want, you can compare against the global	168	C<ev_version_minor>. If you want, you can compare against the global
61	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	169	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
62	version of the library your program was compiled against.	170	version of the library your program was compiled against.
63		171
		172	These version numbers refer to the ABI version of the library, not the
		173	release version.
		174
64	Usually, its a good idea to terminate if the major versions mismatch,	175	Usually, it's a good idea to terminate if the major versions mismatch,
65	as this indicates an incompatible change. Minor versions are usually	176	as this indicates an incompatible change. Minor versions are usually
66	compatible to older versions, so a larger minor version alone is usually	177	compatible to older versions, so a larger minor version alone is usually
67	not a problem.	178	not a problem.
68		179
		180	Example: Make sure we haven't accidentally been linked against the wrong
		181	version.
		182
		183	assert (("libev version mismatch",
		184	ev_version_major () == EV_VERSION_MAJOR
		185	&& ev_version_minor () >= EV_VERSION_MINOR));
		186
		187	=item unsigned int ev_supported_backends ()
		188
		189	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		190	value) compiled into this binary of libev (independent of their
		191	availability on the system you are running on). See C<ev_default_loop> for
		192	a description of the set values.
		193
		194	Example: make sure we have the epoll method, because yeah this is cool and
		195	a must have and can we have a torrent of it please!!!11
		196
		197	assert (("sorry, no epoll, no sex",
		198	ev_supported_backends () & EVBACKEND_EPOLL));
		199
		200	=item unsigned int ev_recommended_backends ()
		201
		202	Return the set of all backends compiled into this binary of libev and also
		203	recommended for this platform. This set is often smaller than the one
		204	returned by C<ev_supported_backends>, as for example kqueue is broken on
		205	most BSDs and will not be auto-detected unless you explicitly request it
		206	(assuming you know what you are doing). This is the set of backends that
		207	libev will probe for if you specify no backends explicitly.
		208
		209	=item unsigned int ev_embeddable_backends ()
		210
		211	Returns the set of backends that are embeddable in other event loops. This
		212	is the theoretical, all-platform, value. To find which backends
		213	might be supported on the current system, you would need to look at
		214	C<ev_embeddable_backends () & ev_supported_backends ()>, likewise for
		215	recommended ones.
		216
		217	See the description of C<ev_embed> watchers for more info.
		218
69	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	219	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
70		220
71	Sets the allocation function to use (the prototype is similar to the	221	Sets the allocation function to use (the prototype is similar - the
72	realloc function). It is used to allocate and free memory (no surprises	222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
73	here). If it returns zero when memory needs to be allocated, the library	223	used to allocate and free memory (no surprises here). If it returns zero
		224	when memory needs to be allocated (C<size != 0>), the library might abort
74	might abort or take some potentially destructive action. The default is	225	or take some potentially destructive action.
75	your system realloc function.	226
		227	Since some systems (at least OpenBSD and Darwin) fail to implement
		228	correct C<realloc> semantics, libev will use a wrapper around the system
		229	C<realloc> and C<free> functions by default.
76		230
77	You could override this function in high-availability programs to, say,	231	You could override this function in high-availability programs to, say,
78	free some memory if it cannot allocate memory, to use a special allocator,	232	free some memory if it cannot allocate memory, to use a special allocator,
79	or even to sleep a while and retry until some memory is available.	233	or even to sleep a while and retry until some memory is available.
80		234
		235	Example: Replace the libev allocator with one that waits a bit and then
		236	retries (example requires a standards-compliant C<realloc>).
		237
		238	static void *
		239	persistent_realloc (void *ptr, size_t size)
		240	{
		241	for (;;)
		242	{
		243	void *newptr = realloc (ptr, size);
		244
		245	if (newptr)
		246	return newptr;
		247
		248	sleep (60);
		249	}
		250	}
		251
		252	...
		253	ev_set_allocator (persistent_realloc);
		254
81	=item ev_set_syserr_cb (void (*cb)(const char *msg));	255	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
82		256
83	Set the callback function to call on a retryable syscall error (such	257	Set the callback function to call on a retryable system call error (such
84	as failed select, poll, epoll_wait). The message is a printable string	258	as failed select, poll, epoll_wait). The message is a printable string
85	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
86	callback is set, then libev will expect it to remedy the sitution, no	260	callback is set, then libev will expect it to remedy the situation, no
87	matter what, when it returns. That is, libev will geenrally retry the	261	matter what, when it returns. That is, libev will generally retry the
88	requested operation, or, if the condition doesn't go away, do bad stuff	262	requested operation, or, if the condition doesn't go away, do bad stuff
89	(such as abort).	263	(such as abort).
90		264
		265	Example: This is basically the same thing that libev does internally, too.
		266
		267	static void
		268	fatal_error (const char *msg)
		269	{
		270	perror (msg);
		271	abort ();
		272	}
		273
		274	...
		275	ev_set_syserr_cb (fatal_error);
		276
91	=back	277	=back
92		278
93	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
94		280
95	An event loop is described by a C<struct ev_loop *>. The library knows two	281	An event loop is described by a C<struct ev_loop *> (the C<struct>
96	types of such loops, the I<default> loop, which supports signals and child	282	is I<not> optional in this case, as there is also an C<ev_loop>
97	events, and dynamically created loops which do not.	283	I<function>).
98		284
99	If you use threads, a common model is to run the default event loop	285	The library knows two types of such loops, the I<default> loop, which
100	in your main thread (or in a separate thrad) and for each thread you	286	supports signals and child events, and dynamically created loops which do
101	create, you also create another event loop. Libev itself does no lockign	287	not.
102	whatsoever, so if you mix calls to different event loops, make sure you
103	lock (this is usually a bad idea, though, even if done right).
104		288
105	=over 4	289	=over 4
106		290
107	=item struct ev_loop *ev_default_loop (unsigned int flags)	291	=item struct ev_loop *ev_default_loop (unsigned int flags)
108		292
109	This will initialise the default event loop if it hasn't been initialised	293	This will initialise the default event loop if it hasn't been initialised
110	yet and return it. If the default loop could not be initialised, returns	294	yet and return it. If the default loop could not be initialised, returns
111	false. If it already was initialised it simply returns it (and ignores the	295	false. If it already was initialised it simply returns it (and ignores the
112	flags).	296	flags. If that is troubling you, check C<ev_backend ()> afterwards).
113		297
114	If you don't know what event loop to use, use the one returned from this	298	If you don't know what event loop to use, use the one returned from this
115	function.	299	function.
116		300
		301	Note that this function is I<not> thread-safe, so if you want to use it
		302	from multiple threads, you have to lock (note also that this is unlikely,
		303	as loops cannot bes hared easily between threads anyway).
		304
		305	The default loop is the only loop that can handle C<ev_signal> and
		306	C<ev_child> watchers, and to do this, it always registers a handler
		307	for C<SIGCHLD>. If this is a problem for your application you can either
		308	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
		309	can simply overwrite the C<SIGCHLD> signal handler I<after> calling
		310	C<ev_default_init>.
		311
117	The flags argument can be used to specify special behaviour or specific	312	The flags argument can be used to specify special behaviour or specific
118	backends to use, and is usually specified as 0 (or EVFLAG_AUTO)	313	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
119		314
120	It supports the following flags:	315	The following flags are supported:
121		316
122	=over 4	317	=over 4
123		318
124	=item EVFLAG_AUTO	319	=item C<EVFLAG_AUTO>
125		320
126	The default flags value. Use this if you have no clue (its the right	321	The default flags value. Use this if you have no clue (it's the right
127	thing, believe me).	322	thing, believe me).
128		323
129	=item EVFLAG_NOENV	324	=item C<EVFLAG_NOENV>
130		325
131	If this flag bit is ored into the flag value then libev will I<not> look	326	If this flag bit is or'ed into the flag value (or the program runs setuid
132	at the environment variable C<LIBEV_FLAGS>. Otherwise (the default), this	327	or setgid) then libev will I<not> look at the environment variable
133	environment variable will override the flags completely. This is useful	328	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
		329	override the flags completely if it is found in the environment. This is
134	to try out specific backends to tets their performance, or to work around	330	useful to try out specific backends to test their performance, or to work
135	bugs.	331	around bugs.
136		332
		333	=item C<EVFLAG_FORKCHECK>
		334
		335	Instead of calling C<ev_default_fork> or C<ev_loop_fork> manually after
		336	a fork, you can also make libev check for a fork in each iteration by
		337	enabling this flag.
		338
		339	This works by calling C<getpid ()> on every iteration of the loop,
		340	and thus this might slow down your event loop if you do a lot of loop
		341	iterations and little real work, but is usually not noticeable (on my
		342	GNU/Linux system for example, C<getpid> is actually a simple 5-insn sequence
		343	without a system call and thus I<very> fast, but my GNU/Linux system also has
		344	C<pthread_atfork> which is even faster).
		345
		346	The big advantage of this flag is that you can forget about fork (and
		347	forget about forgetting to tell libev about forking) when you use this
		348	flag.
		349
		350	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		351	environment variable.
		352
137	=item EVMETHOD_SELECT portable select backend	353	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
138		354
139	=item EVMETHOD_POLL poll backend (everywhere except windows)	355	This is your standard select(2) backend. Not I<completely> standard, as
		356	libev tries to roll its own fd_set with no limits on the number of fds,
		357	but if that fails, expect a fairly low limit on the number of fds when
		358	using this backend. It doesn't scale too well (O(highest_fd)), but its
		359	usually the fastest backend for a low number of (low-numbered :) fds.
140		360
141	=item EVMETHOD_EPOLL linux only	361	To get good performance out of this backend you need a high amount of
		362	parallelism (most of the file descriptors should be busy). If you are
		363	writing a server, you should C<accept ()> in a loop to accept as many
		364	connections as possible during one iteration. You might also want to have
		365	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		366	readiness notifications you get per iteration.
142		367
143	=item EVMETHOD_KQUEUE some bsds only	368	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		369	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		370	C<exceptfds> set on that platform).
144		371
145	=item EVMETHOD_DEVPOLL solaris 8 only	372	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		373
147	=item EVMETHOD_PORT solaris 10 only	374	And this is your standard poll(2) backend. It's more complicated
		375	than select, but handles sparse fds better and has no artificial
		376	limit on the number of fds you can use (except it will slow down
		377	considerably with a lot of inactive fds). It scales similarly to select,
		378	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		379	performance tips.
148		380
149	If one or more of these are ored into the flags value, then only these	381	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
150	backends will be tried (in the reverse order as given here). If one are	382	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
151	specified, any backend will do.	383
		384	=item C<EVBACKEND_EPOLL> (value 4, Linux)
		385
		386	For few fds, this backend is a bit little slower than poll and select,
		387	but it scales phenomenally better. While poll and select usually scale
		388	like O(total_fds) where n is the total number of fds (or the highest fd),
		389	epoll scales either O(1) or O(active_fds). The epoll design has a number
		390	of shortcomings, such as silently dropping events in some hard-to-detect
		391	cases and requiring a system call per fd change, no fork support and bad
		392	support for dup.
		393
		394	While stopping, setting and starting an I/O watcher in the same iteration
		395	will result in some caching, there is still a system call per such incident
		396	(because the fd could point to a different file description now), so its
		397	best to avoid that. Also, C<dup ()>'ed file descriptors might not work
		398	very well if you register events for both fds.
		399
		400	Please note that epoll sometimes generates spurious notifications, so you
		401	need to use non-blocking I/O or other means to avoid blocking when no data
		402	(or space) is available.
		403
		404	Best performance from this backend is achieved by not unregistering all
		405	watchers for a file descriptor until it has been closed, if possible,
		406	i.e. keep at least one watcher active per fd at all times. Stopping and
		407	starting a watcher (without re-setting it) also usually doesn't cause
		408	extra overhead.
		409
		410	While nominally embeddable in other event loops, this feature is broken in
		411	all kernel versions tested so far.
		412
		413	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		414	C<EVBACKEND_POLL>.
		415
		416	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		417
		418	Kqueue deserves special mention, as at the time of this writing, it was
		419	broken on all BSDs except NetBSD (usually it doesn't work reliably with
		420	anything but sockets and pipes, except on Darwin, where of course it's
		421	completely useless). For this reason it's not being "auto-detected" unless
		422	you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or
		423	libev was compiled on a known-to-be-good (-enough) system like NetBSD.
		424
		425	You still can embed kqueue into a normal poll or select backend and use it
		426	only for sockets (after having made sure that sockets work with kqueue on
		427	the target platform). See C<ev_embed> watchers for more info.
		428
		429	It scales in the same way as the epoll backend, but the interface to the
		430	kernel is more efficient (which says nothing about its actual speed, of
		431	course). While stopping, setting and starting an I/O watcher does never
		432	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		433	two event changes per incident. Support for C<fork ()> is very bad and it
		434	drops fds silently in similarly hard-to-detect cases.
		435
		436	This backend usually performs well under most conditions.
		437
		438	While nominally embeddable in other event loops, this doesn't work
		439	everywhere, so you might need to test for this. And since it is broken
		440	almost everywhere, you should only use it when you have a lot of sockets
		441	(for which it usually works), by embedding it into another event loop
		442	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
		443	using it only for sockets.
		444
		445	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		446	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		447	C<NOTE_EOF>.
		448
		449	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		450
		451	This is not implemented yet (and might never be, unless you send me an
		452	implementation). According to reports, C</dev/poll> only supports sockets
		453	and is not embeddable, which would limit the usefulness of this backend
		454	immensely.
		455
		456	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		457
		458	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		459	it's really slow, but it still scales very well (O(active_fds)).
		460
		461	Please note that Solaris event ports can deliver a lot of spurious
		462	notifications, so you need to use non-blocking I/O or other means to avoid
		463	blocking when no data (or space) is available.
		464
		465	While this backend scales well, it requires one system call per active
		466	file descriptor per loop iteration. For small and medium numbers of file
		467	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		468	might perform better.
		469
		470	On the positive side, with the exception of the spurious readiness
		471	notifications, this backend actually performed fully to specification
		472	in all tests and is fully embeddable, which is a rare feat among the
		473	OS-specific backends.
		474
		475	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		476	C<EVBACKEND_POLL>.
		477
		478	=item C<EVBACKEND_ALL>
		479
		480	Try all backends (even potentially broken ones that wouldn't be tried
		481	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		482	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		483
		484	It is definitely not recommended to use this flag.
152		485
153	=back	486	=back
		487
		488	If one or more of these are or'ed into the flags value, then only these
		489	backends will be tried (in the reverse order as listed here). If none are
		490	specified, all backends in C<ev_recommended_backends ()> will be tried.
		491
		492	Example: This is the most typical usage.
		493
		494	if (!ev_default_loop (0))
		495	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		496
		497	Example: Restrict libev to the select and poll backends, and do not allow
		498	environment settings to be taken into account:
		499
		500	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		501
		502	Example: Use whatever libev has to offer, but make sure that kqueue is
		503	used if available (warning, breaks stuff, best use only with your own
		504	private event loop and only if you know the OS supports your types of
		505	fds):
		506
		507	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
154		508
155	=item struct ev_loop *ev_loop_new (unsigned int flags)	509	=item struct ev_loop *ev_loop_new (unsigned int flags)
156		510
157	Similar to C<ev_default_loop>, but always creates a new event loop that is	511	Similar to C<ev_default_loop>, but always creates a new event loop that is
158	always distinct from the default loop. Unlike the default loop, it cannot	512	always distinct from the default loop. Unlike the default loop, it cannot
159	handle signal and child watchers, and attempts to do so will be greeted by	513	handle signal and child watchers, and attempts to do so will be greeted by
160	undefined behaviour (or a failed assertion if assertions are enabled).	514	undefined behaviour (or a failed assertion if assertions are enabled).
161		515
		516	Note that this function I<is> thread-safe, and the recommended way to use
		517	libev with threads is indeed to create one loop per thread, and using the
		518	default loop in the "main" or "initial" thread.
		519
		520	Example: Try to create a event loop that uses epoll and nothing else.
		521
		522	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		523	if (!epoller)
		524	fatal ("no epoll found here, maybe it hides under your chair");
		525
162	=item ev_default_destroy ()	526	=item ev_default_destroy ()
163		527
164	Destroys the default loop again (frees all memory and kernel state	528	Destroys the default loop again (frees all memory and kernel state
165	etc.). This stops all registered event watchers (by not touching them in	529	etc.). None of the active event watchers will be stopped in the normal
166	any way whatsoever, although you cnanot rely on this :).	530	sense, so e.g. C<ev_is_active> might still return true. It is your
		531	responsibility to either stop all watchers cleanly yourself I<before>
		532	calling this function, or cope with the fact afterwards (which is usually
		533	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		534	for example).
		535
		536	Note that certain global state, such as signal state, will not be freed by
		537	this function, and related watchers (such as signal and child watchers)
		538	would need to be stopped manually.
		539
		540	In general it is not advisable to call this function except in the
		541	rare occasion where you really need to free e.g. the signal handling
		542	pipe fds. If you need dynamically allocated loops it is better to use
		543	C<ev_loop_new> and C<ev_loop_destroy>).
167		544
168	=item ev_loop_destroy (loop)	545	=item ev_loop_destroy (loop)
169		546
170	Like C<ev_default_destroy>, but destroys an event loop created by an	547	Like C<ev_default_destroy>, but destroys an event loop created by an
171	earlier call to C<ev_loop_new>.	548	earlier call to C<ev_loop_new>.
172		549
173	=item ev_default_fork ()	550	=item ev_default_fork ()
174		551
		552	This function sets a flag that causes subsequent C<ev_loop> iterations
175	This function reinitialises the kernel state for backends that have	553	to reinitialise the kernel state for backends that have one. Despite the
176	one. Despite the name, you can call it anytime, but it makes most sense	554	name, you can call it anytime, but it makes most sense after forking, in
177	after forking, in either the parent or child process (or both, but that	555	the child process (or both child and parent, but that again makes little
178	again makes little sense).	556	sense). You I<must> call it in the child before using any of the libev
		557	functions, and it will only take effect at the next C<ev_loop> iteration.
179		558
180	You I<must> call this function after forking if and only if you want to	559	On the other hand, you only need to call this function in the child
181	use the event library in both processes. If you just fork+exec, you don't	560	process if and only if you want to use the event library in the child. If
182	have to call it.	561	you just fork+exec, you don't have to call it at all.
183		562
184	The function itself is quite fast and its usually not a problem to call	563	The function itself is quite fast and it's usually not a problem to call
185	it just in case after a fork. To make this easy, the function will fit in	564	it just in case after a fork. To make this easy, the function will fit in
186	quite nicely into a call to C<pthread_atfork>:	565	quite nicely into a call to C<pthread_atfork>:
187		566
188	pthread_atfork (0, 0, ev_default_fork);	567	pthread_atfork (0, 0, ev_default_fork);
189		568
190	=item ev_loop_fork (loop)	569	=item ev_loop_fork (loop)
191		570
192	Like C<ev_default_fork>, but acts on an event loop created by	571	Like C<ev_default_fork>, but acts on an event loop created by
193	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	572	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
194	after fork, and how you do this is entirely your own problem.	573	after fork that you want to re-use in the child, and how you do this is
		574	entirely your own problem.
195		575
		576	=item int ev_is_default_loop (loop)
		577
		578	Returns true when the given loop is, in fact, the default loop, and false
		579	otherwise.
		580
		581	=item unsigned int ev_loop_count (loop)
		582
		583	Returns the count of loop iterations for the loop, which is identical to
		584	the number of times libev did poll for new events. It starts at C<0> and
		585	happily wraps around with enough iterations.
		586
		587	This value can sometimes be useful as a generation counter of sorts (it
		588	"ticks" the number of loop iterations), as it roughly corresponds with
		589	C<ev_prepare> and C<ev_check> calls.
		590
196	=item unsigned int ev_method (loop)	591	=item unsigned int ev_backend (loop)
197		592
198	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	593	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
199	use.	594	use.
200		595
201	=item ev_tstamp = ev_now (loop)	596	=item ev_tstamp ev_now (loop)
202		597
203	Returns the current "event loop time", which is the time the event loop	598	Returns the current "event loop time", which is the time the event loop
204	got events and started processing them. This timestamp does not change	599	received events and started processing them. This timestamp does not
205	as long as callbacks are being processed, and this is also the base time	600	change as long as callbacks are being processed, and this is also the base
206	used for relative timers. You can treat it as the timestamp of the event	601	time used for relative timers. You can treat it as the timestamp of the
207	occuring (or more correctly, the mainloop finding out about it).	602	event occurring (or more correctly, libev finding out about it).
		603
		604	=item ev_now_update (loop)
		605
		606	Establishes the current time by querying the kernel, updating the time
		607	returned by C<ev_now ()> in the progress. This is a costly operation and
		608	is usually done automatically within C<ev_loop ()>.
		609
		610	This function is rarely useful, but when some event callback runs for a
		611	very long time without entering the event loop, updating libev's idea of
		612	the current time is a good idea.
		613
		614	See also "The special problem of time updates" in the C<ev_timer> section.
208		615
209	=item ev_loop (loop, int flags)	616	=item ev_loop (loop, int flags)
210		617
211	Finally, this is it, the event handler. This function usually is called	618	Finally, this is it, the event handler. This function usually is called
212	after you initialised all your watchers and you want to start handling	619	after you initialised all your watchers and you want to start handling
213	events.	620	events.
214		621
215	If the flags argument is specified as 0, it will not return until either	622	If the flags argument is specified as C<0>, it will not return until
216	no event watchers are active anymore or C<ev_unloop> was called.	623	either no event watchers are active anymore or C<ev_unloop> was called.
		624
		625	Please note that an explicit C<ev_unloop> is usually better than
		626	relying on all watchers to be stopped when deciding when a program has
		627	finished (especially in interactive programs), but having a program
		628	that automatically loops as long as it has to and no longer by virtue
		629	of relying on its watchers stopping correctly, that is truly a thing of
		630	beauty.
217		631
218	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	632	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
219	those events and any outstanding ones, but will not block your process in	633	those events and any already outstanding ones, but will not block your
220	case there are no events.	634	process in case there are no events and will return after one iteration of
		635	the loop.
221		636
222	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	637	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
223	neccessary) and will handle those and any outstanding ones. It will block	638	necessary) and will handle those and any already outstanding ones. It
224	your process until at least one new event arrives.	639	will block your process until at least one new event arrives (which could
		640	be an event internal to libev itself, so there is no guarentee that a
		641	user-registered callback will be called), and will return after one
		642	iteration of the loop.
225		643
226	This flags value could be used to implement alternative looping	644	This is useful if you are waiting for some external event in conjunction
227	constructs, but the C<prepare> and C<check> watchers provide a better and	645	with something not expressible using other libev watchers (i.e. "roll your
228	more generic mechanism.	646	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		647	usually a better approach for this kind of thing.
		648
		649	Here are the gory details of what C<ev_loop> does:
		650
		651	- Before the first iteration, call any pending watchers.
		652	* If EVFLAG_FORKCHECK was used, check for a fork.
		653	- If a fork was detected (by any means), queue and call all fork watchers.
		654	- Queue and call all prepare watchers.
		655	- If we have been forked, detach and recreate the kernel state
		656	as to not disturb the other process.
		657	- Update the kernel state with all outstanding changes.
		658	- Update the "event loop time" (ev_now ()).
		659	- Calculate for how long to sleep or block, if at all
		660	(active idle watchers, EVLOOP_NONBLOCK or not having
		661	any active watchers at all will result in not sleeping).
		662	- Sleep if the I/O and timer collect interval say so.
		663	- Block the process, waiting for any events.
		664	- Queue all outstanding I/O (fd) events.
		665	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		666	- Queue all expired timers.
		667	- Queue all expired periodics.
		668	- Unless any events are pending now, queue all idle watchers.
		669	- Queue all check watchers.
		670	- Call all queued watchers in reverse order (i.e. check watchers first).
		671	Signals and child watchers are implemented as I/O watchers, and will
		672	be handled here by queueing them when their watcher gets executed.
		673	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		674	were used, or there are no active watchers, return, otherwise
		675	continue with step *.
		676
		677	Example: Queue some jobs and then loop until no events are outstanding
		678	anymore.
		679
		680	... queue jobs here, make sure they register event watchers as long
		681	... as they still have work to do (even an idle watcher will do..)
		682	ev_loop (my_loop, 0);
		683	... jobs done or somebody called unloop. yeah!
229		684
230	=item ev_unloop (loop, how)	685	=item ev_unloop (loop, how)
231		686
232	Can be used to make a call to C<ev_loop> return early. The C<how> argument	687	Can be used to make a call to C<ev_loop> return early (but only after it
		688	has processed all outstanding events). The C<how> argument must be either
233	must be either C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop>	689	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
234	call return, or C<EVUNLOOP_ALL>, which will make all nested C<ev_loop>	690	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
235	calls return.	691
		692	This "unloop state" will be cleared when entering C<ev_loop> again.
		693
		694	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
236		695
237	=item ev_ref (loop)	696	=item ev_ref (loop)
238		697
239	=item ev_unref (loop)	698	=item ev_unref (loop)
240		699
241	Ref/unref can be used to add or remove a refcount on the event loop: Every	700	Ref/unref can be used to add or remove a reference count on the event
242	watcher keeps one reference. If you have a long-runing watcher you never	701	loop: Every watcher keeps one reference, and as long as the reference
243	unregister that should not keep ev_loop from running, ev_unref() after	702	count is nonzero, C<ev_loop> will not return on its own.
244	starting, and ev_ref() before stopping it. Libev itself uses this for	703
245	example for its internal signal pipe: It is not visible to you as a user	704	If you have a watcher you never unregister that should not keep C<ev_loop>
246	and should not keep C<ev_loop> from exiting if the work is done. It is	705	from returning, call ev_unref() after starting, and ev_ref() before
247	also an excellent way to do this for generic recurring timers or from	706	stopping it.
248	within third-party libraries. Just remember to unref after start and ref	707
249	before stop.	708	As an example, libev itself uses this for its internal signal pipe: It is
		709	not visible to the libev user and should not keep C<ev_loop> from exiting
		710	if no event watchers registered by it are active. It is also an excellent
		711	way to do this for generic recurring timers or from within third-party
		712	libraries. Just remember to I<unref after start> and I<ref before stop>
		713	(but only if the watcher wasn't active before, or was active before,
		714	respectively).
		715
		716	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		717	running when nothing else is active.
		718
		719	ev_signal exitsig;
		720	ev_signal_init (&exitsig, sig_cb, SIGINT);
		721	ev_signal_start (loop, &exitsig);
		722	evf_unref (loop);
		723
		724	Example: For some weird reason, unregister the above signal handler again.
		725
		726	ev_ref (loop);
		727	ev_signal_stop (loop, &exitsig);
		728
		729	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		730
		731	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		732
		733	These advanced functions influence the time that libev will spend waiting
		734	for events. Both time intervals are by default C<0>, meaning that libev
		735	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		736	latency.
		737
		738	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		739	allows libev to delay invocation of I/O and timer/periodic callbacks
		740	to increase efficiency of loop iterations (or to increase power-saving
		741	opportunities).
		742
		743	The idea is that sometimes your program runs just fast enough to handle
		744	one (or very few) event(s) per loop iteration. While this makes the
		745	program responsive, it also wastes a lot of CPU time to poll for new
		746	events, especially with backends like C<select ()> which have a high
		747	overhead for the actual polling but can deliver many events at once.
		748
		749	By setting a higher I<io collect interval> you allow libev to spend more
		750	time collecting I/O events, so you can handle more events per iteration,
		751	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		752	C<ev_timer>) will be not affected. Setting this to a non-null value will
		753	introduce an additional C<ev_sleep ()> call into most loop iterations.
		754
		755	Likewise, by setting a higher I<timeout collect interval> you allow libev
		756	to spend more time collecting timeouts, at the expense of increased
		757	latency/jitter/inexactness (the watcher callback will be called
		758	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		759	value will not introduce any overhead in libev.
		760
		761	Many (busy) programs can usually benefit by setting the I/O collect
		762	interval to a value near C<0.1> or so, which is often enough for
		763	interactive servers (of course not for games), likewise for timeouts. It
		764	usually doesn't make much sense to set it to a lower value than C<0.01>,
		765	as this approaches the timing granularity of most systems.
		766
		767	Setting the I<timeout collect interval> can improve the opportunity for
		768	saving power, as the program will "bundle" timer callback invocations that
		769	are "near" in time together, by delaying some, thus reducing the number of
		770	times the process sleeps and wakes up again. Another useful technique to
		771	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		772	they fire on, say, one-second boundaries only.
		773
		774	=item ev_loop_verify (loop)
		775
		776	This function only does something when C<EV_VERIFY> support has been
		777	compiled in, which is the default for non-minimal builds. It tries to go
		778	through all internal structures and checks them for validity. If anything
		779	is found to be inconsistent, it will print an error message to standard
		780	error and call C<abort ()>.
		781
		782	This can be used to catch bugs inside libev itself: under normal
		783	circumstances, this function will never abort as of course libev keeps its
		784	data structures consistent.
250		785
251	=back	786	=back
252		787
		788
253	=head1 ANATOMY OF A WATCHER	789	=head1 ANATOMY OF A WATCHER
		790
		791	In the following description, uppercase C<TYPE> in names stands for the
		792	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		793	watchers and C<ev_io_start> for I/O watchers.
254		794
255	A watcher is a structure that you create and register to record your	795	A watcher is a structure that you create and register to record your
256	interest in some event. For instance, if you want to wait for STDIN to	796	interest in some event. For instance, if you want to wait for STDIN to
257	become readable, you would create an ev_io watcher for that:	797	become readable, you would create an C<ev_io> watcher for that:
258		798
259	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	799	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
260	{	800	{
261	ev_io_stop (w);	801	ev_io_stop (w);
262	ev_unloop (loop, EVUNLOOP_ALL);	802	ev_unloop (loop, EVUNLOOP_ALL);
263	}	803	}
264		804
265	struct ev_loop *loop = ev_default_loop (0);	805	struct ev_loop *loop = ev_default_loop (0);
		806
266	struct ev_io stdin_watcher;	807	ev_io stdin_watcher;
		808
267	ev_init (&stdin_watcher, my_cb);	809	ev_init (&stdin_watcher, my_cb);
268	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	810	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
269	ev_io_start (loop, &stdin_watcher);	811	ev_io_start (loop, &stdin_watcher);
		812
270	ev_loop (loop, 0);	813	ev_loop (loop, 0);
271		814
272	As you can see, you are responsible for allocating the memory for your	815	As you can see, you are responsible for allocating the memory for your
273	watcher structures (and it is usually a bad idea to do this on the stack,	816	watcher structures (and it is I<usually> a bad idea to do this on the
274	although this can sometimes be quite valid).	817	stack).
		818
		819	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		820	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
275		821
276	Each watcher structure must be initialised by a call to C<ev_init	822	Each watcher structure must be initialised by a call to C<ev_init
277	(watcher *, callback)>, which expects a callback to be provided. This	823	(watcher *, callback)>, which expects a callback to be provided. This
278	callback gets invoked each time the event occurs (or, in the case of io	824	callback gets invoked each time the event occurs (or, in the case of I/O
279	watchers, each time the event loop detects that the file descriptor given	825	watchers, each time the event loop detects that the file descriptor given
280	is readable and/or writable).	826	is readable and/or writable).
281		827
282	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	828	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
283	with arguments specific to this watcher type. There is also a macro	829	macro to configure it, with arguments specific to the watcher type. There
284	to combine initialisation and setting in one call: C<< ev_<type>_init	830	is also a macro to combine initialisation and setting in one call: C<<
285	(watcher *, callback, ...) >>.	831	ev_TYPE_init (watcher *, callback, ...) >>.
286		832
287	To make the watcher actually watch out for events, you have to start it	833	To make the watcher actually watch out for events, you have to start it
288	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	834	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
289	*) >>), and you can stop watching for events at any time by calling the	835	*) >>), and you can stop watching for events at any time by calling the
290	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	836	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
291		837
292	As long as your watcher is active (has been started but not stopped) you	838	As long as your watcher is active (has been started but not stopped) you
293	must not touch the values stored in it. Most specifically you must never	839	must not touch the values stored in it. Most specifically you must never
294	reinitialise it or call its set method.	840	reinitialise it or call its C<ev_TYPE_set> macro.
295
296	You cna check wether an event is active by calling the C<ev_is_active
297	(watcher *)> macro. To see wether an event is outstanding (but the
298	callback for it has not been called yet) you cna use the C<ev_is_pending
299	(watcher *)> macro.
300		841
301	Each and every callback receives the event loop pointer as first, the	842	Each and every callback receives the event loop pointer as first, the
302	registered watcher structure as second, and a bitset of received events as	843	registered watcher structure as second, and a bitset of received events as
303	third argument.	844	third argument.
304		845
305	The rceeived events usually include a single bit per event type received	846	The received events usually include a single bit per event type received
306	(you can receive multiple events at the same time). The possible bit masks	847	(you can receive multiple events at the same time). The possible bit masks
307	are:	848	are:
308		849
309	=over 4	850	=over 4
310		851
311	=item EV_READ	852	=item C<EV_READ>
312		853
313	=item EV_WRITE	854	=item C<EV_WRITE>
314		855
315	The file descriptor in the ev_io watcher has become readable and/or	856	The file descriptor in the C<ev_io> watcher has become readable and/or
316	writable.	857	writable.
317		858
318	=item EV_TIMEOUT	859	=item C<EV_TIMEOUT>
319		860
320	The ev_timer watcher has timed out.	861	The C<ev_timer> watcher has timed out.
321		862
322	=item EV_PERIODIC	863	=item C<EV_PERIODIC>
323		864
324	The ev_periodic watcher has timed out.	865	The C<ev_periodic> watcher has timed out.
325		866
326	=item EV_SIGNAL	867	=item C<EV_SIGNAL>
327		868
328	The signal specified in the ev_signal watcher has been received by a thread.	869	The signal specified in the C<ev_signal> watcher has been received by a thread.
329		870
330	=item EV_CHILD	871	=item C<EV_CHILD>
331		872
332	The pid specified in the ev_child watcher has received a status change.	873	The pid specified in the C<ev_child> watcher has received a status change.
333		874
		875	=item C<EV_STAT>
		876
		877	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		878
334	=item EV_IDLE	879	=item C<EV_IDLE>
335		880
336	The ev_idle watcher has determined that you have nothing better to do.	881	The C<ev_idle> watcher has determined that you have nothing better to do.
337		882
338	=item EV_PREPARE	883	=item C<EV_PREPARE>
339		884
340	=item EV_CHECK	885	=item C<EV_CHECK>
341		886
342	All ev_prepare watchers are invoked just I<before> C<ev_loop> starts	887	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts
343	to gather new events, and all ev_check watchers are invoked just after	888	to gather new events, and all C<ev_check> watchers are invoked just after
344	C<ev_loop> has gathered them, but before it invokes any callbacks for any	889	C<ev_loop> has gathered them, but before it invokes any callbacks for any
345	received events. Callbacks of both watcher types can start and stop as	890	received events. Callbacks of both watcher types can start and stop as
346	many watchers as they want, and all of them will be taken into account	891	many watchers as they want, and all of them will be taken into account
347	(for example, a ev_prepare watcher might start an idle watcher to keep	892	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
348	C<ev_loop> from blocking).	893	C<ev_loop> from blocking).
349		894
		895	=item C<EV_EMBED>
		896
		897	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		898
		899	=item C<EV_FORK>
		900
		901	The event loop has been resumed in the child process after fork (see
		902	C<ev_fork>).
		903
		904	=item C<EV_ASYNC>
		905
		906	The given async watcher has been asynchronously notified (see C<ev_async>).
		907
350	=item EV_ERROR	908	=item C<EV_ERROR>
351		909
352	An unspecified error has occured, the watcher has been stopped. This might	910	An unspecified error has occurred, the watcher has been stopped. This might
353	happen because the watcher could not be properly started because libev	911	happen because the watcher could not be properly started because libev
354	ran out of memory, a file descriptor was found to be closed or any other	912	ran out of memory, a file descriptor was found to be closed or any other
		913	problem. Libev considers these application bugs.
		914
355	problem. You best act on it by reporting the problem and somehow coping	915	You best act on it by reporting the problem and somehow coping with the
356	with the watcher being stopped.	916	watcher being stopped. Note that well-written programs should not receive
		917	an error ever, so when your watcher receives it, this usually indicates a
		918	bug in your program.
357		919
358	Libev will usually signal a few "dummy" events together with an error,	920	Libev will usually signal a few "dummy" events together with an error, for
359	for example it might indicate that a fd is readable or writable, and if	921	example it might indicate that a fd is readable or writable, and if your
360	your callbacks is well-written it can just attempt the operation and cope	922	callbacks is well-written it can just attempt the operation and cope with
361	with the error from read() or write(). This will not work in multithreaded	923	the error from read() or write(). This will not work in multi-threaded
362	programs, though, so beware.	924	programs, though, as the fd could already be closed and reused for another
		925	thing, so beware.
363		926
364	=back	927	=back
365		928
		929	=head2 GENERIC WATCHER FUNCTIONS
		930
		931	=over 4
		932
		933	=item C<ev_init> (ev_TYPE *watcher, callback)
		934
		935	This macro initialises the generic portion of a watcher. The contents
		936	of the watcher object can be arbitrary (so C<malloc> will do). Only
		937	the generic parts of the watcher are initialised, you I<need> to call
		938	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		939	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		940	which rolls both calls into one.
		941
		942	You can reinitialise a watcher at any time as long as it has been stopped
		943	(or never started) and there are no pending events outstanding.
		944
		945	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		946	int revents)>.
		947
		948	Example: Initialise an C<ev_io> watcher in two steps.
		949
		950	ev_io w;
		951	ev_init (&w, my_cb);
		952	ev_io_set (&w, STDIN_FILENO, EV_READ);
		953
		954	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		955
		956	This macro initialises the type-specific parts of a watcher. You need to
		957	call C<ev_init> at least once before you call this macro, but you can
		958	call C<ev_TYPE_set> any number of times. You must not, however, call this
		959	macro on a watcher that is active (it can be pending, however, which is a
		960	difference to the C<ev_init> macro).
		961
		962	Although some watcher types do not have type-specific arguments
		963	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		964
		965	See C<ev_init>, above, for an example.
		966
		967	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		968
		969	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		970	calls into a single call. This is the most convenient method to initialise
		971	a watcher. The same limitations apply, of course.
		972
		973	Example: Initialise and set an C<ev_io> watcher in one step.
		974
		975	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		976
		977	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		978
		979	Starts (activates) the given watcher. Only active watchers will receive
		980	events. If the watcher is already active nothing will happen.
		981
		982	Example: Start the C<ev_io> watcher that is being abused as example in this
		983	whole section.
		984
		985	ev_io_start (EV_DEFAULT_UC, &w);
		986
		987	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		988
		989	Stops the given watcher if active, and clears the pending status (whether
		990	the watcher was active or not).
		991
		992	It is possible that stopped watchers are pending - for example,
		993	non-repeating timers are being stopped when they become pending - but
		994	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		995	pending. If you want to free or reuse the memory used by the watcher it is
		996	therefore a good idea to always call its C<ev_TYPE_stop> function.
		997
		998	=item bool ev_is_active (ev_TYPE *watcher)
		999
		1000	Returns a true value iff the watcher is active (i.e. it has been started
		1001	and not yet been stopped). As long as a watcher is active you must not modify
		1002	it.
		1003
		1004	=item bool ev_is_pending (ev_TYPE *watcher)
		1005
		1006	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1007	events but its callback has not yet been invoked). As long as a watcher
		1008	is pending (but not active) you must not call an init function on it (but
		1009	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1010	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1011	it).
		1012
		1013	=item callback ev_cb (ev_TYPE *watcher)
		1014
		1015	Returns the callback currently set on the watcher.
		1016
		1017	=item ev_cb_set (ev_TYPE *watcher, callback)
		1018
		1019	Change the callback. You can change the callback at virtually any time
		1020	(modulo threads).
		1021
		1022	=item ev_set_priority (ev_TYPE *watcher, priority)
		1023
		1024	=item int ev_priority (ev_TYPE *watcher)
		1025
		1026	Set and query the priority of the watcher. The priority is a small
		1027	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1028	(default: C<-2>). Pending watchers with higher priority will be invoked
		1029	before watchers with lower priority, but priority will not keep watchers
		1030	from being executed (except for C<ev_idle> watchers).
		1031
		1032	This means that priorities are I<only> used for ordering callback
		1033	invocation after new events have been received. This is useful, for
		1034	example, to reduce latency after idling, or more often, to bind two
		1035	watchers on the same event and make sure one is called first.
		1036
		1037	If you need to suppress invocation when higher priority events are pending
		1038	you need to look at C<ev_idle> watchers, which provide this functionality.
		1039
		1040	You I<must not> change the priority of a watcher as long as it is active or
		1041	pending.
		1042
		1043	The default priority used by watchers when no priority has been set is
		1044	always C<0>, which is supposed to not be too high and not be too low :).
		1045
		1046	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1047	fine, as long as you do not mind that the priority value you query might
		1048	or might not have been clamped to the valid range.
		1049
		1050	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1051
		1052	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1053	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1054	can deal with that fact, as both are simply passed through to the
		1055	callback.
		1056
		1057	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1058
		1059	If the watcher is pending, this function clears its pending status and
		1060	returns its C<revents> bitset (as if its callback was invoked). If the
		1061	watcher isn't pending it does nothing and returns C<0>.
		1062
		1063	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1064	callback to be invoked, which can be accomplished with this function.
		1065
		1066	=back
		1067
		1068
366	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1069	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
367		1070
368	Each watcher has, by default, a member C<void *data> that you can change	1071	Each watcher has, by default, a member C<void *data> that you can change
369	and read at any time, libev will completely ignore it. This cna be used	1072	and read at any time: libev will completely ignore it. This can be used
370	to associate arbitrary data with your watcher. If you need more data and	1073	to associate arbitrary data with your watcher. If you need more data and
371	don't want to allocate memory and store a pointer to it in that data	1074	don't want to allocate memory and store a pointer to it in that data
372	member, you can also "subclass" the watcher type and provide your own	1075	member, you can also "subclass" the watcher type and provide your own
373	data:	1076	data:
374		1077
375	struct my_io	1078	struct my_io
376	{	1079	{
377	struct ev_io io;	1080	ev_io io;
378	int otherfd;	1081	int otherfd;
379	void *somedata;	1082	void *somedata;
380	struct whatever *mostinteresting;	1083	struct whatever *mostinteresting;
381	}	1084	};
		1085
		1086	...
		1087	struct my_io w;
		1088	ev_io_init (&w.io, my_cb, fd, EV_READ);
382		1089
383	And since your callback will be called with a pointer to the watcher, you	1090	And since your callback will be called with a pointer to the watcher, you
384	can cast it back to your own type:	1091	can cast it back to your own type:
385		1092
386	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1093	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
387	{	1094	{
388	struct my_io *w = (struct my_io *)w_;	1095	struct my_io *w = (struct my_io *)w_;
389	...	1096	...
390	}	1097	}
391		1098
392	More interesting and less C-conformant ways of catsing your callback type	1099	More interesting and less C-conformant ways of casting your callback type
393	have been omitted....	1100	instead have been omitted.
		1101
		1102	Another common scenario is to use some data structure with multiple
		1103	embedded watchers:
		1104
		1105	struct my_biggy
		1106	{
		1107	int some_data;
		1108	ev_timer t1;
		1109	ev_timer t2;
		1110	}
		1111
		1112	In this case getting the pointer to C<my_biggy> is a bit more
		1113	complicated: Either you store the address of your C<my_biggy> struct
		1114	in the C<data> member of the watcher (for woozies), or you need to use
		1115	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1116	programmers):
		1117
		1118	#include <stddef.h>
		1119
		1120	static void
		1121	t1_cb (EV_P_ ev_timer *w, int revents)
		1122	{
		1123	struct my_biggy big = (struct my_biggy *
		1124	(((char *)w) - offsetof (struct my_biggy, t1));
		1125	}
		1126
		1127	static void
		1128	t2_cb (EV_P_ ev_timer *w, int revents)
		1129	{
		1130	struct my_biggy big = (struct my_biggy *
		1131	(((char *)w) - offsetof (struct my_biggy, t2));
		1132	}
394		1133
395		1134
396	=head1 WATCHER TYPES	1135	=head1 WATCHER TYPES
397		1136
398	This section describes each watcher in detail, but will not repeat	1137	This section describes each watcher in detail, but will not repeat
399	information given in the last section.	1138	information given in the last section. Any initialisation/set macros,
		1139	functions and members specific to the watcher type are explained.
400		1140
		1141	Members are additionally marked with either I<[read-only]>, meaning that,
		1142	while the watcher is active, you can look at the member and expect some
		1143	sensible content, but you must not modify it (you can modify it while the
		1144	watcher is stopped to your hearts content), or I<[read-write]>, which
		1145	means you can expect it to have some sensible content while the watcher
		1146	is active, but you can also modify it. Modifying it may not do something
		1147	sensible or take immediate effect (or do anything at all), but libev will
		1148	not crash or malfunction in any way.
		1149
		1150
401	=head2 struct ev_io - is my file descriptor readable or writable	1151	=head2 C<ev_io> - is this file descriptor readable or writable?
402		1152
403	I/O watchers check wether a file descriptor is readable or writable	1153	I/O watchers check whether a file descriptor is readable or writable
404	in each iteration of the event loop (This behaviour is called	1154	in each iteration of the event loop, or, more precisely, when reading
405	level-triggering because you keep receiving events as long as the	1155	would not block the process and writing would at least be able to write
406	condition persists. Remember you cna stop the watcher if you don't want to	1156	some data. This behaviour is called level-triggering because you keep
407	act on the event and neither want to receive future events).	1157	receiving events as long as the condition persists. Remember you can stop
		1158	the watcher if you don't want to act on the event and neither want to
		1159	receive future events.
		1160
		1161	In general you can register as many read and/or write event watchers per
		1162	fd as you want (as long as you don't confuse yourself). Setting all file
		1163	descriptors to non-blocking mode is also usually a good idea (but not
		1164	required if you know what you are doing).
		1165
		1166	If you cannot use non-blocking mode, then force the use of a
		1167	known-to-be-good backend (at the time of this writing, this includes only
		1168	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
		1169
		1170	Another thing you have to watch out for is that it is quite easy to
		1171	receive "spurious" readiness notifications, that is your callback might
		1172	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1173	because there is no data. Not only are some backends known to create a
		1174	lot of those (for example Solaris ports), it is very easy to get into
		1175	this situation even with a relatively standard program structure. Thus
		1176	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1177	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1178
		1179	If you cannot run the fd in non-blocking mode (for example you should
		1180	not play around with an Xlib connection), then you have to separately
		1181	re-test whether a file descriptor is really ready with a known-to-be good
		1182	interface such as poll (fortunately in our Xlib example, Xlib already
		1183	does this on its own, so its quite safe to use). Some people additionally
		1184	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1185	indefinitely.
		1186
		1187	But really, best use non-blocking mode.
		1188
		1189	=head3 The special problem of disappearing file descriptors
		1190
		1191	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1192	descriptor (either due to calling C<close> explicitly or any other means,
		1193	such as C<dup2>). The reason is that you register interest in some file
		1194	descriptor, but when it goes away, the operating system will silently drop
		1195	this interest. If another file descriptor with the same number then is
		1196	registered with libev, there is no efficient way to see that this is, in
		1197	fact, a different file descriptor.
		1198
		1199	To avoid having to explicitly tell libev about such cases, libev follows
		1200	the following policy: Each time C<ev_io_set> is being called, libev
		1201	will assume that this is potentially a new file descriptor, otherwise
		1202	it is assumed that the file descriptor stays the same. That means that
		1203	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1204	descriptor even if the file descriptor number itself did not change.
		1205
		1206	This is how one would do it normally anyway, the important point is that
		1207	the libev application should not optimise around libev but should leave
		1208	optimisations to libev.
		1209
		1210	=head3 The special problem of dup'ed file descriptors
		1211
		1212	Some backends (e.g. epoll), cannot register events for file descriptors,
		1213	but only events for the underlying file descriptions. That means when you
		1214	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1215	events for them, only one file descriptor might actually receive events.
		1216
		1217	There is no workaround possible except not registering events
		1218	for potentially C<dup ()>'ed file descriptors, or to resort to
		1219	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1220
		1221	=head3 The special problem of fork
		1222
		1223	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1224	useless behaviour. Libev fully supports fork, but needs to be told about
		1225	it in the child.
		1226
		1227	To support fork in your programs, you either have to call
		1228	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1229	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1230	C<EVBACKEND_POLL>.
		1231
		1232	=head3 The special problem of SIGPIPE
		1233
		1234	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1235	when writing to a pipe whose other end has been closed, your program gets
		1236	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1237	this is sensible behaviour, for daemons, this is usually undesirable.
		1238
		1239	So when you encounter spurious, unexplained daemon exits, make sure you
		1240	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1241	somewhere, as that would have given you a big clue).
		1242
		1243
		1244	=head3 Watcher-Specific Functions
408		1245
409	=over 4	1246	=over 4
410		1247
411	=item ev_io_init (ev_io *, callback, int fd, int events)	1248	=item ev_io_init (ev_io *, callback, int fd, int events)
412		1249
413	=item ev_io_set (ev_io *, int fd, int events)	1250	=item ev_io_set (ev_io *, int fd, int events)
414		1251
415	Configures an ev_io watcher. The fd is the file descriptor to rceeive	1252	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
416	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1253	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
417	EV_WRITE> to receive the given events.	1254	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1255
		1256	=item int fd [read-only]
		1257
		1258	The file descriptor being watched.
		1259
		1260	=item int events [read-only]
		1261
		1262	The events being watched.
418		1263
419	=back	1264	=back
420		1265
		1266	=head3 Examples
		1267
		1268	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1269	readable, but only once. Since it is likely line-buffered, you could
		1270	attempt to read a whole line in the callback.
		1271
		1272	static void
		1273	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1274	{
		1275	ev_io_stop (loop, w);
		1276	.. read from stdin here (or from w->fd) and handle any I/O errors
		1277	}
		1278
		1279	...
		1280	struct ev_loop *loop = ev_default_init (0);
		1281	ev_io stdin_readable;
		1282	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1283	ev_io_start (loop, &stdin_readable);
		1284	ev_loop (loop, 0);
		1285
		1286
421	=head2 struct ev_timer - relative and optionally recurring timeouts	1287	=head2 C<ev_timer> - relative and optionally repeating timeouts
422		1288
423	Timer watchers are simple relative timers that generate an event after a	1289	Timer watchers are simple relative timers that generate an event after a
424	given time, and optionally repeating in regular intervals after that.	1290	given time, and optionally repeating in regular intervals after that.
425		1291
426	The timers are based on real time, that is, if you register an event that	1292	The timers are based on real time, that is, if you register an event that
427	times out after an hour and youreset your system clock to last years	1293	times out after an hour and you reset your system clock to January last
428	time, it will still time out after (roughly) and hour. "Roughly" because	1294	year, it will still time out after (roughly) one hour. "Roughly" because
429	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1295	detecting time jumps is hard, and some inaccuracies are unavoidable (the
430	monotonic clock option helps a lot here).	1296	monotonic clock option helps a lot here).
431		1297
		1298	The callback is guaranteed to be invoked only I<after> its timeout has
		1299	passed, but if multiple timers become ready during the same loop iteration
		1300	then order of execution is undefined.
		1301
		1302	=head3 Be smart about timeouts
		1303
		1304	Many real-world problems involve some kind of timeout, usually for error
		1305	recovery. A typical example is an HTTP request - if the other side hangs,
		1306	you want to raise some error after a while.
		1307
		1308	What follows are some ways to handle this problem, from obvious and
		1309	inefficient to smart and efficient.
		1310
		1311	In the following, a 60 second activity timeout is assumed - a timeout that
		1312	gets reset to 60 seconds each time there is activity (e.g. each time some
		1313	data or other life sign was received).
		1314
432	=over 4	1315	=over 4
433		1316
		1317	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1318
		1319	This is the most obvious, but not the most simple way: In the beginning,
		1320	start the watcher:
		1321
		1322	ev_timer_init (timer, callback, 60., 0.);
		1323	ev_timer_start (loop, timer);
		1324
		1325	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1326	and start it again:
		1327
		1328	ev_timer_stop (loop, timer);
		1329	ev_timer_set (timer, 60., 0.);
		1330	ev_timer_start (loop, timer);
		1331
		1332	This is relatively simple to implement, but means that each time there is
		1333	some activity, libev will first have to remove the timer from its internal
		1334	data structure and then add it again. Libev tries to be fast, but it's
		1335	still not a constant-time operation.
		1336
		1337	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1338
		1339	This is the easiest way, and involves using C<ev_timer_again> instead of
		1340	C<ev_timer_start>.
		1341
		1342	To implement this, configure an C<ev_timer> with a C<repeat> value
		1343	of C<60> and then call C<ev_timer_again> at start and each time you
		1344	successfully read or write some data. If you go into an idle state where
		1345	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1346	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1347
		1348	That means you can ignore both the C<ev_timer_start> function and the
		1349	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1350	member and C<ev_timer_again>.
		1351
		1352	At start:
		1353
		1354	ev_timer_init (timer, callback);
		1355	timer->repeat = 60.;
		1356	ev_timer_again (loop, timer);
		1357
		1358	Each time there is some activity:
		1359
		1360	ev_timer_again (loop, timer);
		1361
		1362	It is even possible to change the time-out on the fly, regardless of
		1363	whether the watcher is active or not:
		1364
		1365	timer->repeat = 30.;
		1366	ev_timer_again (loop, timer);
		1367
		1368	This is slightly more efficient then stopping/starting the timer each time
		1369	you want to modify its timeout value, as libev does not have to completely
		1370	remove and re-insert the timer from/into its internal data structure.
		1371
		1372	It is, however, even simpler than the "obvious" way to do it.
		1373
		1374	=item 3. Let the timer time out, but then re-arm it as required.
		1375
		1376	This method is more tricky, but usually most efficient: Most timeouts are
		1377	relatively long compared to the intervals between other activity - in
		1378	our example, within 60 seconds, there are usually many I/O events with
		1379	associated activity resets.
		1380
		1381	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1382	but remember the time of last activity, and check for a real timeout only
		1383	within the callback:
		1384
		1385	ev_tstamp last_activity; // time of last activity
		1386
		1387	static void
		1388	callback (EV_P_ ev_timer *w, int revents)
		1389	{
		1390	ev_tstamp now = ev_now (EV_A);
		1391	ev_tstamp timeout = last_activity + 60.;
		1392
		1393	// if last_activity + 60. is older than now, we did time out
		1394	if (timeout < now)
		1395	{
		1396	// timeout occured, take action
		1397	}
		1398	else
		1399	{
		1400	// callback was invoked, but there was some activity, re-arm
		1401	// the watcher to fire in last_activity + 60, which is
		1402	// guaranteed to be in the future, so "again" is positive:
		1403	w->again = timeout - now;
		1404	ev_timer_again (EV_A_ w);
		1405	}
		1406	}
		1407
		1408	To summarise the callback: first calculate the real timeout (defined
		1409	as "60 seconds after the last activity"), then check if that time has
		1410	been reached, which means something I<did>, in fact, time out. Otherwise
		1411	the callback was invoked too early (C<timeout> is in the future), so
		1412	re-schedule the timer to fire at that future time, to see if maybe we have
		1413	a timeout then.
		1414
		1415	Note how C<ev_timer_again> is used, taking advantage of the
		1416	C<ev_timer_again> optimisation when the timer is already running.
		1417
		1418	This scheme causes more callback invocations (about one every 60 seconds
		1419	minus half the average time between activity), but virtually no calls to
		1420	libev to change the timeout.
		1421
		1422	To start the timer, simply initialise the watcher and set C<last_activity>
		1423	to the current time (meaning we just have some activity :), then call the
		1424	callback, which will "do the right thing" and start the timer:
		1425
		1426	ev_timer_init (timer, callback);
		1427	last_activity = ev_now (loop);
		1428	callback (loop, timer, EV_TIMEOUT);
		1429
		1430	And when there is some activity, simply store the current time in
		1431	C<last_activity>, no libev calls at all:
		1432
		1433	last_actiivty = ev_now (loop);
		1434
		1435	This technique is slightly more complex, but in most cases where the
		1436	time-out is unlikely to be triggered, much more efficient.
		1437
		1438	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1439	callback :) - just change the timeout and invoke the callback, which will
		1440	fix things for you.
		1441
		1442	=item 4. Wee, just use a double-linked list for your timeouts.
		1443
		1444	If there is not one request, but many thousands (millions...), all
		1445	employing some kind of timeout with the same timeout value, then one can
		1446	do even better:
		1447
		1448	When starting the timeout, calculate the timeout value and put the timeout
		1449	at the I<end> of the list.
		1450
		1451	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1452	the list is expected to fire (for example, using the technique #3).
		1453
		1454	When there is some activity, remove the timer from the list, recalculate
		1455	the timeout, append it to the end of the list again, and make sure to
		1456	update the C<ev_timer> if it was taken from the beginning of the list.
		1457
		1458	This way, one can manage an unlimited number of timeouts in O(1) time for
		1459	starting, stopping and updating the timers, at the expense of a major
		1460	complication, and having to use a constant timeout. The constant timeout
		1461	ensures that the list stays sorted.
		1462
		1463	=back
		1464
		1465	So which method the best?
		1466
		1467	Method #2 is a simple no-brain-required solution that is adequate in most
		1468	situations. Method #3 requires a bit more thinking, but handles many cases
		1469	better, and isn't very complicated either. In most case, choosing either
		1470	one is fine, with #3 being better in typical situations.
		1471
		1472	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1473	rather complicated, but extremely efficient, something that really pays
		1474	off after the first million or so of active timers, i.e. it's usually
		1475	overkill :)
		1476
		1477	=head3 The special problem of time updates
		1478
		1479	Establishing the current time is a costly operation (it usually takes at
		1480	least two system calls): EV therefore updates its idea of the current
		1481	time only before and after C<ev_loop> collects new events, which causes a
		1482	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1483	lots of events in one iteration.
		1484
		1485	The relative timeouts are calculated relative to the C<ev_now ()>
		1486	time. This is usually the right thing as this timestamp refers to the time
		1487	of the event triggering whatever timeout you are modifying/starting. If
		1488	you suspect event processing to be delayed and you I<need> to base the
		1489	timeout on the current time, use something like this to adjust for this:
		1490
		1491	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
		1492
		1493	If the event loop is suspended for a long time, you can also force an
		1494	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1495	()>.
		1496
		1497	=head3 Watcher-Specific Functions and Data Members
		1498
		1499	=over 4
		1500
434	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1501	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
435		1502
436	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1503	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
437		1504
438	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1505	Configure the timer to trigger after C<after> seconds. If C<repeat>
439	C<0.>, then it will automatically be stopped. If it is positive, then the	1506	is C<0.>, then it will automatically be stopped once the timeout is
440	timer will automatically be configured to trigger again C<repeat> seconds	1507	reached. If it is positive, then the timer will automatically be
441	later, again, and again, until stopped manually.	1508	configured to trigger again C<repeat> seconds later, again, and again,
		1509	until stopped manually.
442		1510
443	The timer itself will do a best-effort at avoiding drift, that is, if you	1511	The timer itself will do a best-effort at avoiding drift, that is, if
444	configure a timer to trigger every 10 seconds, then it will trigger at	1512	you configure a timer to trigger every 10 seconds, then it will normally
445	exactly 10 second intervals. If, however, your program cannot keep up with	1513	trigger at exactly 10 second intervals. If, however, your program cannot
446	the timer (ecause it takes longer than those 10 seconds to do stuff) the	1514	keep up with the timer (because it takes longer than those 10 seconds to
447	timer will not fire more than once per event loop iteration.	1515	do stuff) the timer will not fire more than once per event loop iteration.
448		1516
449	=item ev_timer_again (loop)	1517	=item ev_timer_again (loop, ev_timer *)
450		1518
451	This will act as if the timer timed out and restart it again if it is	1519	This will act as if the timer timed out and restart it again if it is
452	repeating. The exact semantics are:	1520	repeating. The exact semantics are:
453		1521
		1522	If the timer is pending, its pending status is cleared.
		1523
454	If the timer is started but nonrepeating, stop it.	1524	If the timer is started but non-repeating, stop it (as if it timed out).
455		1525
456	If the timer is repeating, either start it if necessary (with the repeat	1526	If the timer is repeating, either start it if necessary (with the
457	value), or reset the running timer to the repeat value.	1527	C<repeat> value), or reset the running timer to the C<repeat> value.
458		1528
459	This sounds a bit complicated, but here is a useful and typical	1529	This sounds a bit complicated, see "Be smart about timeouts", above, for a
460	example: Imagine you have a tcp connection and you want a so-called idle	1530	usage example.
461	timeout, that is, you want to be called when there have been, say, 60	1531
462	seconds of inactivity on the socket. The easiest way to do this is to	1532	=item ev_tstamp repeat [read-write]
463	configure an ev_timer with after=repeat=60 and calling ev_timer_again each	1533
464	time you successfully read or write some data. If you go into an idle	1534	The current C<repeat> value. Will be used each time the watcher times out
465	state where you do not expect data to travel on the socket, you can stop	1535	or C<ev_timer_again> is called, and determines the next timeout (if any),
466	the timer, and again will automatically restart it if need be.	1536	which is also when any modifications are taken into account.
467		1537
468	=back	1538	=back
469		1539
470	=head2 ev_periodic	1540	=head3 Examples
		1541
		1542	Example: Create a timer that fires after 60 seconds.
		1543
		1544	static void
		1545	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1546	{
		1547	.. one minute over, w is actually stopped right here
		1548	}
		1549
		1550	ev_timer mytimer;
		1551	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1552	ev_timer_start (loop, &mytimer);
		1553
		1554	Example: Create a timeout timer that times out after 10 seconds of
		1555	inactivity.
		1556
		1557	static void
		1558	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1559	{
		1560	.. ten seconds without any activity
		1561	}
		1562
		1563	ev_timer mytimer;
		1564	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1565	ev_timer_again (&mytimer); /* start timer */
		1566	ev_loop (loop, 0);
		1567
		1568	// and in some piece of code that gets executed on any "activity":
		1569	// reset the timeout to start ticking again at 10 seconds
		1570	ev_timer_again (&mytimer);
		1571
		1572
		1573	=head2 C<ev_periodic> - to cron or not to cron?
471		1574
472	Periodic watchers are also timers of a kind, but they are very versatile	1575	Periodic watchers are also timers of a kind, but they are very versatile
473	(and unfortunately a bit complex).	1576	(and unfortunately a bit complex).
474		1577
475	Unlike ev_timer's, they are not based on real time (or relative time)	1578	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
476	but on wallclock time (absolute time). You can tell a periodic watcher	1579	but on wall clock time (absolute time). You can tell a periodic watcher
477	to trigger "at" some specific point in time. For example, if you tell a	1580	to trigger after some specific point in time. For example, if you tell a
478	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1581	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()
479	+ 10.>) and then reset your system clock to the last year, then it will	1582	+ 10.>, that is, an absolute time not a delay) and then reset your system
		1583	clock to January of the previous year, then it will take more than year
480	take a year to trigger the event (unlike an ev_timer, which would trigger	1584	to trigger the event (unlike an C<ev_timer>, which would still trigger
481	roughly 10 seconds later and of course not if you reset your system time	1585	roughly 10 seconds later as it uses a relative timeout).
482	again).
483		1586
484	They can also be used to implement vastly more complex timers, such as	1587	C<ev_periodic>s can also be used to implement vastly more complex timers,
485	triggering an event on eahc midnight, local time.	1588	such as triggering an event on each "midnight, local time", or other
		1589	complicated rules.
		1590
		1591	As with timers, the callback is guaranteed to be invoked only when the
		1592	time (C<at>) has passed, but if multiple periodic timers become ready
		1593	during the same loop iteration, then order of execution is undefined.
		1594
		1595	=head3 Watcher-Specific Functions and Data Members
486		1596
487	=over 4	1597	=over 4
488		1598
489	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1599	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
490		1600
491	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1601	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
492		1602
493	Lots of arguments, lets sort it out... There are basically three modes of	1603	Lots of arguments, lets sort it out... There are basically three modes of
494	operation, and we will explain them from simplest to complex:	1604	operation, and we will explain them from simplest to most complex:
495
496		1605
497	=over 4	1606	=over 4
498		1607
499	=item * absolute timer (interval = reschedule_cb = 0)	1608	=item * absolute timer (at = time, interval = reschedule_cb = 0)
500		1609
501	In this configuration the watcher triggers an event at the wallclock time	1610	In this configuration the watcher triggers an event after the wall clock
502	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1611	time C<at> has passed. It will not repeat and will not adjust when a time
503	that is, if it is to be run at January 1st 2011 then it will run when the	1612	jump occurs, that is, if it is to be run at January 1st 2011 then it will
504	system time reaches or surpasses this time.	1613	only run when the system clock reaches or surpasses this time.
505		1614
506	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1615	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
507		1616
508	In this mode the watcher will always be scheduled to time out at the next	1617	In this mode the watcher will always be scheduled to time out at the next
509	C<at + N * interval> time (for some integer N) and then repeat, regardless	1618	C<at + N * interval> time (for some integer N, which can also be negative)
510	of any time jumps.	1619	and then repeat, regardless of any time jumps.
511		1620
512	This can be used to create timers that do not drift with respect to system	1621	This can be used to create timers that do not drift with respect to the
513	time:	1622	system clock, for example, here is a C<ev_periodic> that triggers each
		1623	hour, on the hour:
514		1624
515	ev_periodic_set (&periodic, 0., 3600., 0);	1625	ev_periodic_set (&periodic, 0., 3600., 0);
516		1626
517	This doesn't mean there will always be 3600 seconds in between triggers,	1627	This doesn't mean there will always be 3600 seconds in between triggers,
518	but only that the the callback will be called when the system time shows a	1628	but only that the callback will be called when the system time shows a
519	full hour (UTC), or more correct, when the system time is evenly divisible	1629	full hour (UTC), or more correctly, when the system time is evenly divisible
520	by 3600.	1630	by 3600.
521		1631
522	Another way to think about it (for the mathematically inclined) is that	1632	Another way to think about it (for the mathematically inclined) is that
523	ev_periodic will try to run the callback in this mode at the next possible	1633	C<ev_periodic> will try to run the callback in this mode at the next possible
524	time where C<time = at (mod interval)>, regardless of any time jumps.	1634	time where C<time = at (mod interval)>, regardless of any time jumps.
525		1635
		1636	For numerical stability it is preferable that the C<at> value is near
		1637	C<ev_now ()> (the current time), but there is no range requirement for
		1638	this value, and in fact is often specified as zero.
		1639
		1640	Note also that there is an upper limit to how often a timer can fire (CPU
		1641	speed for example), so if C<interval> is very small then timing stability
		1642	will of course deteriorate. Libev itself tries to be exact to be about one
		1643	millisecond (if the OS supports it and the machine is fast enough).
		1644
526	=item * manual reschedule mode (reschedule_cb = callback)	1645	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
527		1646
528	In this mode the values for C<interval> and C<at> are both being	1647	In this mode the values for C<interval> and C<at> are both being
529	ignored. Instead, each time the periodic watcher gets scheduled, the	1648	ignored. Instead, each time the periodic watcher gets scheduled, the
530	reschedule callback will be called with the watcher as first, and the	1649	reschedule callback will be called with the watcher as first, and the
531	current time as second argument.	1650	current time as second argument.
532		1651
533	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	1652	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
534	periodic watcher, ever, or make any event loop modificstions>. If you need	1653	ever, or make ANY event loop modifications whatsoever>.
535	to stop it, return 1e30 (or so, fudge fudge) and stop it afterwards.
536		1654
		1655	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1656	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1657	only event loop modification you are allowed to do).
		1658
537	Its prototype is c<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1659	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
538	ev_tstamp now)>, e.g.:	1660	*w, ev_tstamp now)>, e.g.:
539		1661
		1662	static ev_tstamp
540	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1663	my_rescheduler (ev_periodic *w, ev_tstamp now)
541	{	1664	{
542	return now + 60.;	1665	return now + 60.;
543	}	1666	}
544		1667
545	It must return the next time to trigger, based on the passed time value	1668	It must return the next time to trigger, based on the passed time value
546	(that is, the lowest time value larger than to the second argument). It	1669	(that is, the lowest time value larger than to the second argument). It
547	will usually be called just before the callback will be triggered, but	1670	will usually be called just before the callback will be triggered, but
548	might be called at other times, too.	1671	might be called at other times, too.
549		1672
		1673	NOTE: I<< This callback must always return a time that is higher than or
		1674	equal to the passed C<now> value >>.
		1675
550	This can be used to create very complex timers, such as a timer that	1676	This can be used to create very complex timers, such as a timer that
551	triggers on each midnight, local time. To do this, you would calculate the	1677	triggers on "next midnight, local time". To do this, you would calculate the
552	next midnight after C<now> and return the timestamp value for this. How you do this	1678	next midnight after C<now> and return the timestamp value for this. How
553	is, again, up to you (but it is not trivial).	1679	you do this is, again, up to you (but it is not trivial, which is the main
		1680	reason I omitted it as an example).
554		1681
555	=back	1682	=back
556		1683
557	=item ev_periodic_again (loop, ev_periodic *)	1684	=item ev_periodic_again (loop, ev_periodic *)
558		1685
559	Simply stops and restarts the periodic watcher again. This is only useful	1686	Simply stops and restarts the periodic watcher again. This is only useful
560	when you changed some parameters or the reschedule callback would return	1687	when you changed some parameters or the reschedule callback would return
561	a different time than the last time it was called (e.g. in a crond like	1688	a different time than the last time it was called (e.g. in a crond like
562	program when the crontabs have changed).	1689	program when the crontabs have changed).
563		1690
		1691	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1692
		1693	When active, returns the absolute time that the watcher is supposed to
		1694	trigger next.
		1695
		1696	=item ev_tstamp offset [read-write]
		1697
		1698	When repeating, this contains the offset value, otherwise this is the
		1699	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
		1700
		1701	Can be modified any time, but changes only take effect when the periodic
		1702	timer fires or C<ev_periodic_again> is being called.
		1703
		1704	=item ev_tstamp interval [read-write]
		1705
		1706	The current interval value. Can be modified any time, but changes only
		1707	take effect when the periodic timer fires or C<ev_periodic_again> is being
		1708	called.
		1709
		1710	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		1711
		1712	The current reschedule callback, or C<0>, if this functionality is
		1713	switched off. Can be changed any time, but changes only take effect when
		1714	the periodic timer fires or C<ev_periodic_again> is being called.
		1715
564	=back	1716	=back
565		1717
		1718	=head3 Examples
		1719
		1720	Example: Call a callback every hour, or, more precisely, whenever the
		1721	system time is divisible by 3600. The callback invocation times have
		1722	potentially a lot of jitter, but good long-term stability.
		1723
		1724	static void
		1725	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		1726	{
		1727	... its now a full hour (UTC, or TAI or whatever your clock follows)
		1728	}
		1729
		1730	ev_periodic hourly_tick;
		1731	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		1732	ev_periodic_start (loop, &hourly_tick);
		1733
		1734	Example: The same as above, but use a reschedule callback to do it:
		1735
		1736	#include <math.h>
		1737
		1738	static ev_tstamp
		1739	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		1740	{
		1741	return now + (3600. - fmod (now, 3600.));
		1742	}
		1743
		1744	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		1745
		1746	Example: Call a callback every hour, starting now:
		1747
		1748	ev_periodic hourly_tick;
		1749	ev_periodic_init (&hourly_tick, clock_cb,
		1750	fmod (ev_now (loop), 3600.), 3600., 0);
		1751	ev_periodic_start (loop, &hourly_tick);
		1752
		1753
566	=head2 ev_signal - signal me when a signal gets signalled	1754	=head2 C<ev_signal> - signal me when a signal gets signalled!
567		1755
568	Signal watchers will trigger an event when the process receives a specific	1756	Signal watchers will trigger an event when the process receives a specific
569	signal one or more times. Even though signals are very asynchronous, libev	1757	signal one or more times. Even though signals are very asynchronous, libev
570	will try its best to deliver signals synchronously, i.e. as part of the	1758	will try it's best to deliver signals synchronously, i.e. as part of the
571	normal event processing, like any other event.	1759	normal event processing, like any other event.
572		1760
		1761	If you want signals asynchronously, just use C<sigaction> as you would
		1762	do without libev and forget about sharing the signal. You can even use
		1763	C<ev_async> from a signal handler to synchronously wake up an event loop.
		1764
573	You cna configure as many watchers as you like per signal. Only when the	1765	You can configure as many watchers as you like per signal. Only when the
574	first watcher gets started will libev actually register a signal watcher	1766	first watcher gets started will libev actually register a signal handler
575	with the kernel (thus it coexists with your own signal handlers as long	1767	with the kernel (thus it coexists with your own signal handlers as long as
576	as you don't register any with libev). Similarly, when the last signal	1768	you don't register any with libev for the same signal). Similarly, when
577	watcher for a signal is stopped libev will reset the signal handler to	1769	the last signal watcher for a signal is stopped, libev will reset the
578	SIG_DFL (regardless of what it was set to before).	1770	signal handler to SIG_DFL (regardless of what it was set to before).
		1771
		1772	If possible and supported, libev will install its handlers with
		1773	C<SA_RESTART> behaviour enabled, so system calls should not be unduly
		1774	interrupted. If you have a problem with system calls getting interrupted by
		1775	signals you can block all signals in an C<ev_check> watcher and unblock
		1776	them in an C<ev_prepare> watcher.
		1777
		1778	=head3 Watcher-Specific Functions and Data Members
579		1779
580	=over 4	1780	=over 4
581		1781
582	=item ev_signal_init (ev_signal *, callback, int signum)	1782	=item ev_signal_init (ev_signal *, callback, int signum)
583		1783
584	=item ev_signal_set (ev_signal *, int signum)	1784	=item ev_signal_set (ev_signal *, int signum)
585		1785
586	Configures the watcher to trigger on the given signal number (usually one	1786	Configures the watcher to trigger on the given signal number (usually one
587	of the C<SIGxxx> constants).	1787	of the C<SIGxxx> constants).
588		1788
		1789	=item int signum [read-only]
		1790
		1791	The signal the watcher watches out for.
		1792
589	=back	1793	=back
590		1794
		1795	=head3 Examples
		1796
		1797	Example: Try to exit cleanly on SIGINT.
		1798
		1799	static void
		1800	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		1801	{
		1802	ev_unloop (loop, EVUNLOOP_ALL);
		1803	}
		1804
		1805	ev_signal signal_watcher;
		1806	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1807	ev_signal_start (loop, &signal_watcher);
		1808
		1809
591	=head2 ev_child - wait for pid status changes	1810	=head2 C<ev_child> - watch out for process status changes
592		1811
593	Child watchers trigger when your process receives a SIGCHLD in response to	1812	Child watchers trigger when your process receives a SIGCHLD in response to
594	some child status changes (most typically when a child of yours dies).	1813	some child status changes (most typically when a child of yours dies or
		1814	exits). It is permissible to install a child watcher I<after> the child
		1815	has been forked (which implies it might have already exited), as long
		1816	as the event loop isn't entered (or is continued from a watcher), i.e.,
		1817	forking and then immediately registering a watcher for the child is fine,
		1818	but forking and registering a watcher a few event loop iterations later is
		1819	not.
		1820
		1821	Only the default event loop is capable of handling signals, and therefore
		1822	you can only register child watchers in the default event loop.
		1823
		1824	=head3 Process Interaction
		1825
		1826	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1827	initialised. This is necessary to guarantee proper behaviour even if
		1828	the first child watcher is started after the child exits. The occurrence
		1829	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1830	synchronously as part of the event loop processing. Libev always reaps all
		1831	children, even ones not watched.
		1832
		1833	=head3 Overriding the Built-In Processing
		1834
		1835	Libev offers no special support for overriding the built-in child
		1836	processing, but if your application collides with libev's default child
		1837	handler, you can override it easily by installing your own handler for
		1838	C<SIGCHLD> after initialising the default loop, and making sure the
		1839	default loop never gets destroyed. You are encouraged, however, to use an
		1840	event-based approach to child reaping and thus use libev's support for
		1841	that, so other libev users can use C<ev_child> watchers freely.
		1842
		1843	=head3 Stopping the Child Watcher
		1844
		1845	Currently, the child watcher never gets stopped, even when the
		1846	child terminates, so normally one needs to stop the watcher in the
		1847	callback. Future versions of libev might stop the watcher automatically
		1848	when a child exit is detected.
		1849
		1850	=head3 Watcher-Specific Functions and Data Members
595		1851
596	=over 4	1852	=over 4
597		1853
598	=item ev_child_init (ev_child *, callback, int pid)	1854	=item ev_child_init (ev_child *, callback, int pid, int trace)
599		1855
600	=item ev_child_set (ev_child *, int pid)	1856	=item ev_child_set (ev_child *, int pid, int trace)
601		1857
602	Configures the watcher to wait for status changes of process C<pid> (or	1858	Configures the watcher to wait for status changes of process C<pid> (or
603	I<any> process if C<pid> is specified as C<0>). The callback can look	1859	I<any> process if C<pid> is specified as C<0>). The callback can look
604	at the C<rstatus> member of the C<ev_child> watcher structure to see	1860	at the C<rstatus> member of the C<ev_child> watcher structure to see
605	the status word (use the macros from C<sys/wait.h>). The C<rpid> member	1861	the status word (use the macros from C<sys/wait.h> and see your systems
606	contains the pid of the process causing the status change.	1862	C<waitpid> documentation). The C<rpid> member contains the pid of the
		1863	process causing the status change. C<trace> must be either C<0> (only
		1864	activate the watcher when the process terminates) or C<1> (additionally
		1865	activate the watcher when the process is stopped or continued).
		1866
		1867	=item int pid [read-only]
		1868
		1869	The process id this watcher watches out for, or C<0>, meaning any process id.
		1870
		1871	=item int rpid [read-write]
		1872
		1873	The process id that detected a status change.
		1874
		1875	=item int rstatus [read-write]
		1876
		1877	The process exit/trace status caused by C<rpid> (see your systems
		1878	C<waitpid> and C<sys/wait.h> documentation for details).
607		1879
608	=back	1880	=back
609		1881
		1882	=head3 Examples
		1883
		1884	Example: C<fork()> a new process and install a child handler to wait for
		1885	its completion.
		1886
		1887	ev_child cw;
		1888
		1889	static void
		1890	child_cb (EV_P_ ev_child *w, int revents)
		1891	{
		1892	ev_child_stop (EV_A_ w);
		1893	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		1894	}
		1895
		1896	pid_t pid = fork ();
		1897
		1898	if (pid < 0)
		1899	// error
		1900	else if (pid == 0)
		1901	{
		1902	// the forked child executes here
		1903	exit (1);
		1904	}
		1905	else
		1906	{
		1907	ev_child_init (&cw, child_cb, pid, 0);
		1908	ev_child_start (EV_DEFAULT_ &cw);
		1909	}
		1910
		1911
		1912	=head2 C<ev_stat> - did the file attributes just change?
		1913
		1914	This watches a file system path for attribute changes. That is, it calls
		1915	C<stat> regularly (or when the OS says it changed) and sees if it changed
		1916	compared to the last time, invoking the callback if it did.
		1917
		1918	The path does not need to exist: changing from "path exists" to "path does
		1919	not exist" is a status change like any other. The condition "path does
		1920	not exist" is signified by the C<st_nlink> field being zero (which is
		1921	otherwise always forced to be at least one) and all the other fields of
		1922	the stat buffer having unspecified contents.
		1923
		1924	The path I<should> be absolute and I<must not> end in a slash. If it is
		1925	relative and your working directory changes, the behaviour is undefined.
		1926
		1927	Since there is no standard kernel interface to do this, the portable
		1928	implementation simply calls C<stat (2)> regularly on the path to see if
		1929	it changed somehow. You can specify a recommended polling interval for
		1930	this case. If you specify a polling interval of C<0> (highly recommended!)
		1931	then a I<suitable, unspecified default> value will be used (which
		1932	you can expect to be around five seconds, although this might change
		1933	dynamically). Libev will also impose a minimum interval which is currently
		1934	around C<0.1>, but thats usually overkill.
		1935
		1936	This watcher type is not meant for massive numbers of stat watchers,
		1937	as even with OS-supported change notifications, this can be
		1938	resource-intensive.
		1939
		1940	At the time of this writing, the only OS-specific interface implemented
		1941	is the Linux inotify interface (implementing kqueue support is left as
		1942	an exercise for the reader. Note, however, that the author sees no way
		1943	of implementing C<ev_stat> semantics with kqueue).
		1944
		1945	=head3 ABI Issues (Largefile Support)
		1946
		1947	Libev by default (unless the user overrides this) uses the default
		1948	compilation environment, which means that on systems with large file
		1949	support disabled by default, you get the 32 bit version of the stat
		1950	structure. When using the library from programs that change the ABI to
		1951	use 64 bit file offsets the programs will fail. In that case you have to
		1952	compile libev with the same flags to get binary compatibility. This is
		1953	obviously the case with any flags that change the ABI, but the problem is
		1954	most noticeably disabled with ev_stat and large file support.
		1955
		1956	The solution for this is to lobby your distribution maker to make large
		1957	file interfaces available by default (as e.g. FreeBSD does) and not
		1958	optional. Libev cannot simply switch on large file support because it has
		1959	to exchange stat structures with application programs compiled using the
		1960	default compilation environment.
		1961
		1962	=head3 Inotify and Kqueue
		1963
		1964	When C<inotify (7)> support has been compiled into libev (generally
		1965	only available with Linux 2.6.25 or above due to bugs in earlier
		1966	implementations) and present at runtime, it will be used to speed up
		1967	change detection where possible. The inotify descriptor will be created
		1968	lazily when the first C<ev_stat> watcher is being started.
		1969
		1970	Inotify presence does not change the semantics of C<ev_stat> watchers
		1971	except that changes might be detected earlier, and in some cases, to avoid
		1972	making regular C<stat> calls. Even in the presence of inotify support
		1973	there are many cases where libev has to resort to regular C<stat> polling,
		1974	but as long as the path exists, libev usually gets away without polling.
		1975
		1976	There is no support for kqueue, as apparently it cannot be used to
		1977	implement this functionality, due to the requirement of having a file
		1978	descriptor open on the object at all times, and detecting renames, unlinks
		1979	etc. is difficult.
		1980
		1981	=head3 The special problem of stat time resolution
		1982
		1983	The C<stat ()> system call only supports full-second resolution portably, and
		1984	even on systems where the resolution is higher, most file systems still
		1985	only support whole seconds.
		1986
		1987	That means that, if the time is the only thing that changes, you can
		1988	easily miss updates: on the first update, C<ev_stat> detects a change and
		1989	calls your callback, which does something. When there is another update
		1990	within the same second, C<ev_stat> will be unable to detect unless the
		1991	stat data does change in other ways (e.g. file size).
		1992
		1993	The solution to this is to delay acting on a change for slightly more
		1994	than a second (or till slightly after the next full second boundary), using
		1995	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		1996	ev_timer_again (loop, w)>).
		1997
		1998	The C<.02> offset is added to work around small timing inconsistencies
		1999	of some operating systems (where the second counter of the current time
		2000	might be be delayed. One such system is the Linux kernel, where a call to
		2001	C<gettimeofday> might return a timestamp with a full second later than
		2002	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2003	update file times then there will be a small window where the kernel uses
		2004	the previous second to update file times but libev might already execute
		2005	the timer callback).
		2006
		2007	=head3 Watcher-Specific Functions and Data Members
		2008
		2009	=over 4
		2010
		2011	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2012
		2013	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2014
		2015	Configures the watcher to wait for status changes of the given
		2016	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2017	be detected and should normally be specified as C<0> to let libev choose
		2018	a suitable value. The memory pointed to by C<path> must point to the same
		2019	path for as long as the watcher is active.
		2020
		2021	The callback will receive an C<EV_STAT> event when a change was detected,
		2022	relative to the attributes at the time the watcher was started (or the
		2023	last change was detected).
		2024
		2025	=item ev_stat_stat (loop, ev_stat *)
		2026
		2027	Updates the stat buffer immediately with new values. If you change the
		2028	watched path in your callback, you could call this function to avoid
		2029	detecting this change (while introducing a race condition if you are not
		2030	the only one changing the path). Can also be useful simply to find out the
		2031	new values.
		2032
		2033	=item ev_statdata attr [read-only]
		2034
		2035	The most-recently detected attributes of the file. Although the type is
		2036	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2037	suitable for your system, but you can only rely on the POSIX-standardised
		2038	members to be present. If the C<st_nlink> member is C<0>, then there was
		2039	some error while C<stat>ing the file.
		2040
		2041	=item ev_statdata prev [read-only]
		2042
		2043	The previous attributes of the file. The callback gets invoked whenever
		2044	C<prev> != C<attr>, or, more precisely, one or more of these members
		2045	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2046	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2047
		2048	=item ev_tstamp interval [read-only]
		2049
		2050	The specified interval.
		2051
		2052	=item const char *path [read-only]
		2053
		2054	The file system path that is being watched.
		2055
		2056	=back
		2057
		2058	=head3 Examples
		2059
		2060	Example: Watch C</etc/passwd> for attribute changes.
		2061
		2062	static void
		2063	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2064	{
		2065	/* /etc/passwd changed in some way */
		2066	if (w->attr.st_nlink)
		2067	{
		2068	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2069	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2070	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2071	}
		2072	else
		2073	/* you shalt not abuse printf for puts */
		2074	puts ("wow, /etc/passwd is not there, expect problems. "
		2075	"if this is windows, they already arrived\n");
		2076	}
		2077
		2078	...
		2079	ev_stat passwd;
		2080
		2081	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2082	ev_stat_start (loop, &passwd);
		2083
		2084	Example: Like above, but additionally use a one-second delay so we do not
		2085	miss updates (however, frequent updates will delay processing, too, so
		2086	one might do the work both on C<ev_stat> callback invocation I<and> on
		2087	C<ev_timer> callback invocation).
		2088
		2089	static ev_stat passwd;
		2090	static ev_timer timer;
		2091
		2092	static void
		2093	timer_cb (EV_P_ ev_timer *w, int revents)
		2094	{
		2095	ev_timer_stop (EV_A_ w);
		2096
		2097	/* now it's one second after the most recent passwd change */
		2098	}
		2099
		2100	static void
		2101	stat_cb (EV_P_ ev_stat *w, int revents)
		2102	{
		2103	/* reset the one-second timer */
		2104	ev_timer_again (EV_A_ &timer);
		2105	}
		2106
		2107	...
		2108	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2109	ev_stat_start (loop, &passwd);
		2110	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2111
		2112
610	=head2 ev_idle - when you've got nothing better to do	2113	=head2 C<ev_idle> - when you've got nothing better to do...
611		2114
612	Idle watchers trigger events when there are no other I/O or timer (or	2115	Idle watchers trigger events when no other events of the same or higher
613	periodic) events pending. That is, as long as your process is busy	2116	priority are pending (prepare, check and other idle watchers do not count
614	handling sockets or timeouts it will not be called. But when your process	2117	as receiving "events").
615	is idle all idle watchers are being called again and again - until	2118
		2119	That is, as long as your process is busy handling sockets or timeouts
		2120	(or even signals, imagine) of the same or higher priority it will not be
		2121	triggered. But when your process is idle (or only lower-priority watchers
		2122	are pending), the idle watchers are being called once per event loop
616	stopped, that is, or your process receives more events.	2123	iteration - until stopped, that is, or your process receives more events
		2124	and becomes busy again with higher priority stuff.
617		2125
618	The most noteworthy effect is that as long as any idle watchers are	2126	The most noteworthy effect is that as long as any idle watchers are
619	active, the process will not block when waiting for new events.	2127	active, the process will not block when waiting for new events.
620		2128
621	Apart from keeping your process non-blocking (which is a useful	2129	Apart from keeping your process non-blocking (which is a useful
622	effect on its own sometimes), idle watchers are a good place to do	2130	effect on its own sometimes), idle watchers are a good place to do
623	"pseudo-background processing", or delay processing stuff to after the	2131	"pseudo-background processing", or delay processing stuff to after the
624	event loop has handled all outstanding events.	2132	event loop has handled all outstanding events.
625		2133
		2134	=head3 Watcher-Specific Functions and Data Members
		2135
626	=over 4	2136	=over 4
627		2137
628	=item ev_idle_init (ev_signal *, callback)	2138	=item ev_idle_init (ev_signal *, callback)
629		2139
630	Initialises and configures the idle watcher - it has no parameters of any	2140	Initialises and configures the idle watcher - it has no parameters of any
631	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2141	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
632	believe me.	2142	believe me.
633		2143
634	=back	2144	=back
635		2145
636	=head2 prepare and check - your hooks into the event loop	2146	=head3 Examples
637		2147
		2148	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2149	callback, free it. Also, use no error checking, as usual.
		2150
		2151	static void
		2152	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2153	{
		2154	free (w);
		2155	// now do something you wanted to do when the program has
		2156	// no longer anything immediate to do.
		2157	}
		2158
		2159	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2160	ev_idle_init (idle_watcher, idle_cb);
		2161	ev_idle_start (loop, idle_cb);
		2162
		2163
		2164	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
		2165
638	Prepare and check watchers usually (but not always) are used in	2166	Prepare and check watchers are usually (but not always) used in pairs:
639	tandom. Prepare watchers get invoked before the process blocks and check	2167	prepare watchers get invoked before the process blocks and check watchers
640	watchers afterwards.	2168	afterwards.
641		2169
		2170	You I<must not> call C<ev_loop> or similar functions that enter
		2171	the current event loop from either C<ev_prepare> or C<ev_check>
		2172	watchers. Other loops than the current one are fine, however. The
		2173	rationale behind this is that you do not need to check for recursion in
		2174	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2175	C<ev_check> so if you have one watcher of each kind they will always be
		2176	called in pairs bracketing the blocking call.
		2177
642	Their main purpose is to integrate other event mechanisms into libev. This	2178	Their main purpose is to integrate other event mechanisms into libev and
643	could be used, for example, to track variable changes, implement your own	2179	their use is somewhat advanced. They could be used, for example, to track
644	watchers, integrate net-snmp or a coroutine library and lots more.	2180	variable changes, implement your own watchers, integrate net-snmp or a
		2181	coroutine library and lots more. They are also occasionally useful if
		2182	you cache some data and want to flush it before blocking (for example,
		2183	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2184	watcher).
645		2185
646	This is done by examining in each prepare call which file descriptors need	2186	This is done by examining in each prepare call which file descriptors
647	to be watched by the other library, registering ev_io watchers for them	2187	need to be watched by the other library, registering C<ev_io> watchers
648	and starting an ev_timer watcher for any timeouts (many libraries provide	2188	for them and starting an C<ev_timer> watcher for any timeouts (many
649	just this functionality). Then, in the check watcher you check for any	2189	libraries provide exactly this functionality). Then, in the check watcher,
650	events that occured (by making your callbacks set soem flags for example)	2190	you check for any events that occurred (by checking the pending status
651	and call back into the library.	2191	of all watchers and stopping them) and call back into the library. The
		2192	I/O and timer callbacks will never actually be called (but must be valid
		2193	nevertheless, because you never know, you know?).
652		2194
653	As another example, the perl Coro module uses these hooks to integrate	2195	As another example, the Perl Coro module uses these hooks to integrate
654	coroutines into libev programs, by yielding to other active coroutines	2196	coroutines into libev programs, by yielding to other active coroutines
655	during each prepare and only letting the process block if no coroutines	2197	during each prepare and only letting the process block if no coroutines
656	are ready to run.	2198	are ready to run (it's actually more complicated: it only runs coroutines
		2199	with priority higher than or equal to the event loop and one coroutine
		2200	of lower priority, but only once, using idle watchers to keep the event
		2201	loop from blocking if lower-priority coroutines are active, thus mapping
		2202	low-priority coroutines to idle/background tasks).
		2203
		2204	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2205	priority, to ensure that they are being run before any other watchers
		2206	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2207
		2208	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2209	activate ("feed") events into libev. While libev fully supports this, they
		2210	might get executed before other C<ev_check> watchers did their job. As
		2211	C<ev_check> watchers are often used to embed other (non-libev) event
		2212	loops those other event loops might be in an unusable state until their
		2213	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2214	others).
		2215
		2216	=head3 Watcher-Specific Functions and Data Members
657		2217
658	=over 4	2218	=over 4
659		2219
660	=item ev_prepare_init (ev_prepare *, callback)	2220	=item ev_prepare_init (ev_prepare *, callback)
661		2221
662	=item ev_check_init (ev_check *, callback)	2222	=item ev_check_init (ev_check *, callback)
663		2223
664	Initialises and configures the prepare or check watcher - they have no	2224	Initialises and configures the prepare or check watcher - they have no
665	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2225	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
666	macros, but using them is utterly, utterly pointless.	2226	macros, but using them is utterly, utterly, utterly and completely
		2227	pointless.
667		2228
668	=back	2229	=back
669		2230
		2231	=head3 Examples
		2232
		2233	There are a number of principal ways to embed other event loops or modules
		2234	into libev. Here are some ideas on how to include libadns into libev
		2235	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2236	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2237	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2238	Glib event loop).
		2239
		2240	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2241	and in a check watcher, destroy them and call into libadns. What follows
		2242	is pseudo-code only of course. This requires you to either use a low
		2243	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2244	the callbacks for the IO/timeout watchers might not have been called yet.
		2245
		2246	static ev_io iow [nfd];
		2247	static ev_timer tw;
		2248
		2249	static void
		2250	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2251	{
		2252	}
		2253
		2254	// create io watchers for each fd and a timer before blocking
		2255	static void
		2256	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2257	{
		2258	int timeout = 3600000;
		2259	struct pollfd fds [nfd];
		2260	// actual code will need to loop here and realloc etc.
		2261	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2262
		2263	/* the callback is illegal, but won't be called as we stop during check */
		2264	ev_timer_init (&tw, 0, timeout * 1e-3);
		2265	ev_timer_start (loop, &tw);
		2266
		2267	// create one ev_io per pollfd
		2268	for (int i = 0; i < nfd; ++i)
		2269	{
		2270	ev_io_init (iow + i, io_cb, fds [i].fd,
		2271	((fds [i].events & POLLIN ? EV_READ : 0)
		2272	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2273
		2274	fds [i].revents = 0;
		2275	ev_io_start (loop, iow + i);
		2276	}
		2277	}
		2278
		2279	// stop all watchers after blocking
		2280	static void
		2281	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2282	{
		2283	ev_timer_stop (loop, &tw);
		2284
		2285	for (int i = 0; i < nfd; ++i)
		2286	{
		2287	// set the relevant poll flags
		2288	// could also call adns_processreadable etc. here
		2289	struct pollfd *fd = fds + i;
		2290	int revents = ev_clear_pending (iow + i);
		2291	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2292	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2293
		2294	// now stop the watcher
		2295	ev_io_stop (loop, iow + i);
		2296	}
		2297
		2298	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2299	}
		2300
		2301	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2302	in the prepare watcher and would dispose of the check watcher.
		2303
		2304	Method 3: If the module to be embedded supports explicit event
		2305	notification (libadns does), you can also make use of the actual watcher
		2306	callbacks, and only destroy/create the watchers in the prepare watcher.
		2307
		2308	static void
		2309	timer_cb (EV_P_ ev_timer *w, int revents)
		2310	{
		2311	adns_state ads = (adns_state)w->data;
		2312	update_now (EV_A);
		2313
		2314	adns_processtimeouts (ads, &tv_now);
		2315	}
		2316
		2317	static void
		2318	io_cb (EV_P_ ev_io *w, int revents)
		2319	{
		2320	adns_state ads = (adns_state)w->data;
		2321	update_now (EV_A);
		2322
		2323	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2324	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2325	}
		2326
		2327	// do not ever call adns_afterpoll
		2328
		2329	Method 4: Do not use a prepare or check watcher because the module you
		2330	want to embed is not flexible enough to support it. Instead, you can
		2331	override their poll function. The drawback with this solution is that the
		2332	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2333	this approach, effectively embedding EV as a client into the horrible
		2334	libglib event loop.
		2335
		2336	static gint
		2337	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2338	{
		2339	int got_events = 0;
		2340
		2341	for (n = 0; n < nfds; ++n)
		2342	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2343
		2344	if (timeout >= 0)
		2345	// create/start timer
		2346
		2347	// poll
		2348	ev_loop (EV_A_ 0);
		2349
		2350	// stop timer again
		2351	if (timeout >= 0)
		2352	ev_timer_stop (EV_A_ &to);
		2353
		2354	// stop io watchers again - their callbacks should have set
		2355	for (n = 0; n < nfds; ++n)
		2356	ev_io_stop (EV_A_ iow [n]);
		2357
		2358	return got_events;
		2359	}
		2360
		2361
		2362	=head2 C<ev_embed> - when one backend isn't enough...
		2363
		2364	This is a rather advanced watcher type that lets you embed one event loop
		2365	into another (currently only C<ev_io> events are supported in the embedded
		2366	loop, other types of watchers might be handled in a delayed or incorrect
		2367	fashion and must not be used).
		2368
		2369	There are primarily two reasons you would want that: work around bugs and
		2370	prioritise I/O.
		2371
		2372	As an example for a bug workaround, the kqueue backend might only support
		2373	sockets on some platform, so it is unusable as generic backend, but you
		2374	still want to make use of it because you have many sockets and it scales
		2375	so nicely. In this case, you would create a kqueue-based loop and embed
		2376	it into your default loop (which might use e.g. poll). Overall operation
		2377	will be a bit slower because first libev has to call C<poll> and then
		2378	C<kevent>, but at least you can use both mechanisms for what they are
		2379	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2380
		2381	As for prioritising I/O: under rare circumstances you have the case where
		2382	some fds have to be watched and handled very quickly (with low latency),
		2383	and even priorities and idle watchers might have too much overhead. In
		2384	this case you would put all the high priority stuff in one loop and all
		2385	the rest in a second one, and embed the second one in the first.
		2386
		2387	As long as the watcher is active, the callback will be invoked every time
		2388	there might be events pending in the embedded loop. The callback must then
		2389	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
		2390	their callbacks (you could also start an idle watcher to give the embedded
		2391	loop strictly lower priority for example). You can also set the callback
		2392	to C<0>, in which case the embed watcher will automatically execute the
		2393	embedded loop sweep.
		2394
		2395	As long as the watcher is started it will automatically handle events. The
		2396	callback will be invoked whenever some events have been handled. You can
		2397	set the callback to C<0> to avoid having to specify one if you are not
		2398	interested in that.
		2399
		2400	Also, there have not currently been made special provisions for forking:
		2401	when you fork, you not only have to call C<ev_loop_fork> on both loops,
		2402	but you will also have to stop and restart any C<ev_embed> watchers
		2403	yourself - but you can use a fork watcher to handle this automatically,
		2404	and future versions of libev might do just that.
		2405
		2406	Unfortunately, not all backends are embeddable: only the ones returned by
		2407	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2408	portable one.
		2409
		2410	So when you want to use this feature you will always have to be prepared
		2411	that you cannot get an embeddable loop. The recommended way to get around
		2412	this is to have a separate variables for your embeddable loop, try to
		2413	create it, and if that fails, use the normal loop for everything.
		2414
		2415	=head3 C<ev_embed> and fork
		2416
		2417	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2418	automatically be applied to the embedded loop as well, so no special
		2419	fork handling is required in that case. When the watcher is not running,
		2420	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2421	as applicable.
		2422
		2423	=head3 Watcher-Specific Functions and Data Members
		2424
		2425	=over 4
		2426
		2427	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2428
		2429	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2430
		2431	Configures the watcher to embed the given loop, which must be
		2432	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2433	invoked automatically, otherwise it is the responsibility of the callback
		2434	to invoke it (it will continue to be called until the sweep has been done,
		2435	if you do not want that, you need to temporarily stop the embed watcher).
		2436
		2437	=item ev_embed_sweep (loop, ev_embed *)
		2438
		2439	Make a single, non-blocking sweep over the embedded loop. This works
		2440	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2441	appropriate way for embedded loops.
		2442
		2443	=item struct ev_loop *other [read-only]
		2444
		2445	The embedded event loop.
		2446
		2447	=back
		2448
		2449	=head3 Examples
		2450
		2451	Example: Try to get an embeddable event loop and embed it into the default
		2452	event loop. If that is not possible, use the default loop. The default
		2453	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2454	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2455	used).
		2456
		2457	struct ev_loop *loop_hi = ev_default_init (0);
		2458	struct ev_loop *loop_lo = 0;
		2459	ev_embed embed;
		2460
		2461	// see if there is a chance of getting one that works
		2462	// (remember that a flags value of 0 means autodetection)
		2463	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2464	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2465	: 0;
		2466
		2467	// if we got one, then embed it, otherwise default to loop_hi
		2468	if (loop_lo)
		2469	{
		2470	ev_embed_init (&embed, 0, loop_lo);
		2471	ev_embed_start (loop_hi, &embed);
		2472	}
		2473	else
		2474	loop_lo = loop_hi;
		2475
		2476	Example: Check if kqueue is available but not recommended and create
		2477	a kqueue backend for use with sockets (which usually work with any
		2478	kqueue implementation). Store the kqueue/socket-only event loop in
		2479	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2480
		2481	struct ev_loop *loop = ev_default_init (0);
		2482	struct ev_loop *loop_socket = 0;
		2483	ev_embed embed;
		2484
		2485	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2486	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2487	{
		2488	ev_embed_init (&embed, 0, loop_socket);
		2489	ev_embed_start (loop, &embed);
		2490	}
		2491
		2492	if (!loop_socket)
		2493	loop_socket = loop;
		2494
		2495	// now use loop_socket for all sockets, and loop for everything else
		2496
		2497
		2498	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2499
		2500	Fork watchers are called when a C<fork ()> was detected (usually because
		2501	whoever is a good citizen cared to tell libev about it by calling
		2502	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2503	event loop blocks next and before C<ev_check> watchers are being called,
		2504	and only in the child after the fork. If whoever good citizen calling
		2505	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2506	handlers will be invoked, too, of course.
		2507
		2508	=head3 Watcher-Specific Functions and Data Members
		2509
		2510	=over 4
		2511
		2512	=item ev_fork_init (ev_signal *, callback)
		2513
		2514	Initialises and configures the fork watcher - it has no parameters of any
		2515	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2516	believe me.
		2517
		2518	=back
		2519
		2520
		2521	=head2 C<ev_async> - how to wake up another event loop
		2522
		2523	In general, you cannot use an C<ev_loop> from multiple threads or other
		2524	asynchronous sources such as signal handlers (as opposed to multiple event
		2525	loops - those are of course safe to use in different threads).
		2526
		2527	Sometimes, however, you need to wake up another event loop you do not
		2528	control, for example because it belongs to another thread. This is what
		2529	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2530	can signal it by calling C<ev_async_send>, which is thread- and signal
		2531	safe.
		2532
		2533	This functionality is very similar to C<ev_signal> watchers, as signals,
		2534	too, are asynchronous in nature, and signals, too, will be compressed
		2535	(i.e. the number of callback invocations may be less than the number of
		2536	C<ev_async_sent> calls).
		2537
		2538	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2539	just the default loop.
		2540
		2541	=head3 Queueing
		2542
		2543	C<ev_async> does not support queueing of data in any way. The reason
		2544	is that the author does not know of a simple (or any) algorithm for a
		2545	multiple-writer-single-reader queue that works in all cases and doesn't
		2546	need elaborate support such as pthreads.
		2547
		2548	That means that if you want to queue data, you have to provide your own
		2549	queue. But at least I can tell you how to implement locking around your
		2550	queue:
		2551
		2552	=over 4
		2553
		2554	=item queueing from a signal handler context
		2555
		2556	To implement race-free queueing, you simply add to the queue in the signal
		2557	handler but you block the signal handler in the watcher callback. Here is
		2558	an example that does that for some fictitious SIGUSR1 handler:
		2559
		2560	static ev_async mysig;
		2561
		2562	static void
		2563	sigusr1_handler (void)
		2564	{
		2565	sometype data;
		2566
		2567	// no locking etc.
		2568	queue_put (data);
		2569	ev_async_send (EV_DEFAULT_ &mysig);
		2570	}
		2571
		2572	static void
		2573	mysig_cb (EV_P_ ev_async *w, int revents)
		2574	{
		2575	sometype data;
		2576	sigset_t block, prev;
		2577
		2578	sigemptyset (&block);
		2579	sigaddset (&block, SIGUSR1);
		2580	sigprocmask (SIG_BLOCK, &block, &prev);
		2581
		2582	while (queue_get (&data))
		2583	process (data);
		2584
		2585	if (sigismember (&prev, SIGUSR1)
		2586	sigprocmask (SIG_UNBLOCK, &block, 0);
		2587	}
		2588
		2589	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2590	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2591	either...).
		2592
		2593	=item queueing from a thread context
		2594
		2595	The strategy for threads is different, as you cannot (easily) block
		2596	threads but you can easily preempt them, so to queue safely you need to
		2597	employ a traditional mutex lock, such as in this pthread example:
		2598
		2599	static ev_async mysig;
		2600	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2601
		2602	static void
		2603	otherthread (void)
		2604	{
		2605	// only need to lock the actual queueing operation
		2606	pthread_mutex_lock (&mymutex);
		2607	queue_put (data);
		2608	pthread_mutex_unlock (&mymutex);
		2609
		2610	ev_async_send (EV_DEFAULT_ &mysig);
		2611	}
		2612
		2613	static void
		2614	mysig_cb (EV_P_ ev_async *w, int revents)
		2615	{
		2616	pthread_mutex_lock (&mymutex);
		2617
		2618	while (queue_get (&data))
		2619	process (data);
		2620
		2621	pthread_mutex_unlock (&mymutex);
		2622	}
		2623
		2624	=back
		2625
		2626
		2627	=head3 Watcher-Specific Functions and Data Members
		2628
		2629	=over 4
		2630
		2631	=item ev_async_init (ev_async *, callback)
		2632
		2633	Initialises and configures the async watcher - it has no parameters of any
		2634	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
		2635	trust me.
		2636
		2637	=item ev_async_send (loop, ev_async *)
		2638
		2639	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2640	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2641	C<ev_feed_event>, this call is safe to do from other threads, signal or
		2642	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		2643	section below on what exactly this means).
		2644
		2645	This call incurs the overhead of a system call only once per loop iteration,
		2646	so while the overhead might be noticeable, it doesn't apply to repeated
		2647	calls to C<ev_async_send>.
		2648
		2649	=item bool = ev_async_pending (ev_async *)
		2650
		2651	Returns a non-zero value when C<ev_async_send> has been called on the
		2652	watcher but the event has not yet been processed (or even noted) by the
		2653	event loop.
		2654
		2655	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2656	the loop iterates next and checks for the watcher to have become active,
		2657	it will reset the flag again. C<ev_async_pending> can be used to very
		2658	quickly check whether invoking the loop might be a good idea.
		2659
		2660	Not that this does I<not> check whether the watcher itself is pending, only
		2661	whether it has been requested to make this watcher pending.
		2662
		2663	=back
		2664
		2665
670	=head1 OTHER FUNCTIONS	2666	=head1 OTHER FUNCTIONS
671		2667
672	There are some other fucntions of possible interest. Described. Here. Now.	2668	There are some other functions of possible interest. Described. Here. Now.
673		2669
674	=over 4	2670	=over 4
675		2671
676	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2672	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
677		2673
678	This function combines a simple timer and an I/O watcher, calls your	2674	This function combines a simple timer and an I/O watcher, calls your
679	callback on whichever event happens first and automatically stop both	2675	callback on whichever event happens first and automatically stops both
680	watchers. This is useful if you want to wait for a single event on an fd	2676	watchers. This is useful if you want to wait for a single event on an fd
681	or timeout without havign to allocate/configure/start/stop/free one or	2677	or timeout without having to allocate/configure/start/stop/free one or
682	more watchers yourself.	2678	more watchers yourself.
683		2679
684	If C<fd> is less than 0, then no I/O watcher will be started and events is	2680	If C<fd> is less than 0, then no I/O watcher will be started and the
685	ignored. Otherwise, an ev_io watcher for the given C<fd> and C<events> set	2681	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
686	will be craeted and started.	2682	the given C<fd> and C<events> set will be created and started.
687		2683
688	If C<timeout> is less than 0, then no timeout watcher will be	2684	If C<timeout> is less than 0, then no timeout watcher will be
689	started. Otherwise an ev_timer watcher with after = C<timeout> (and repeat	2685	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
690	= 0) will be started.	2686	repeat = 0) will be started. C<0> is a valid timeout.
691		2687
692	The callback has the type C<void (*cb)(int revents, void *arg)> and	2688	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
693	gets passed an events set (normally a combination of EV_ERROR, EV_READ,	2689	passed an C<revents> set like normal event callbacks (a combination of
694	EV_WRITE or EV_TIMEOUT) and the C<arg> value passed to C<ev_once>:	2690	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
		2691	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2692	a timeout and an io event at the same time - you probably should give io
		2693	events precedence.
695		2694
		2695	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		2696
696	static void stdin_ready (int revents, void *arg)	2697	static void stdin_ready (int revents, void *arg)
697	{	2698	{
698	if (revents & EV_TIMEOUT)
699	/* doh, nothing entered */
700	else if (revents & EV_READ)	2699	if (revents & EV_READ)
701	/* stdin might have data for us, joy! */	2700	/* stdin might have data for us, joy! */;
		2701	else if (revents & EV_TIMEOUT)
		2702	/* doh, nothing entered */;
702	}	2703	}
703		2704
704	ev_once (STDIN_FILENO, EV_READm 10., stdin_ready, 0);	2705	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
705		2706
706	=item ev_feed_event (loop, watcher, int events)	2707	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
707		2708
708	Feeds the given event set into the event loop, as if the specified event	2709	Feeds the given event set into the event loop, as if the specified event
709	has happened for the specified watcher (which must be a pointer to an	2710	had happened for the specified watcher (which must be a pointer to an
710	initialised but not necessarily active event watcher).	2711	initialised but not necessarily started event watcher).
711		2712
712	=item ev_feed_fd_event (loop, int fd, int revents)	2713	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
713		2714
714	Feed an event on the given fd, as if a file descriptor backend detected it.	2715	Feed an event on the given fd, as if a file descriptor backend detected
		2716	the given events it.
715		2717
716	=item ev_feed_signal_event (loop, int signum)	2718	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
717		2719
718	Feed an event as if the given signal occured (loop must be the default loop!).	2720	Feed an event as if the given signal occurred (C<loop> must be the default
		2721	loop!).
719		2722
720	=back	2723	=back
721		2724
		2725
		2726	=head1 LIBEVENT EMULATION
		2727
		2728	Libev offers a compatibility emulation layer for libevent. It cannot
		2729	emulate the internals of libevent, so here are some usage hints:
		2730
		2731	=over 4
		2732
		2733	=item * Use it by including <event.h>, as usual.
		2734
		2735	=item * The following members are fully supported: ev_base, ev_callback,
		2736	ev_arg, ev_fd, ev_res, ev_events.
		2737
		2738	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		2739	maintained by libev, it does not work exactly the same way as in libevent (consider
		2740	it a private API).
		2741
		2742	=item * Priorities are not currently supported. Initialising priorities
		2743	will fail and all watchers will have the same priority, even though there
		2744	is an ev_pri field.
		2745
		2746	=item * In libevent, the last base created gets the signals, in libev, the
		2747	first base created (== the default loop) gets the signals.
		2748
		2749	=item * Other members are not supported.
		2750
		2751	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		2752	to use the libev header file and library.
		2753
		2754	=back
		2755
		2756	=head1 C++ SUPPORT
		2757
		2758	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		2759	you to use some convenience methods to start/stop watchers and also change
		2760	the callback model to a model using method callbacks on objects.
		2761
		2762	To use it,
		2763
		2764	#include <ev++.h>
		2765
		2766	This automatically includes F<ev.h> and puts all of its definitions (many
		2767	of them macros) into the global namespace. All C++ specific things are
		2768	put into the C<ev> namespace. It should support all the same embedding
		2769	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		2770
		2771	Care has been taken to keep the overhead low. The only data member the C++
		2772	classes add (compared to plain C-style watchers) is the event loop pointer
		2773	that the watcher is associated with (or no additional members at all if
		2774	you disable C<EV_MULTIPLICITY> when embedding libev).
		2775
		2776	Currently, functions, and static and non-static member functions can be
		2777	used as callbacks. Other types should be easy to add as long as they only
		2778	need one additional pointer for context. If you need support for other
		2779	types of functors please contact the author (preferably after implementing
		2780	it).
		2781
		2782	Here is a list of things available in the C<ev> namespace:
		2783
		2784	=over 4
		2785
		2786	=item C<ev::READ>, C<ev::WRITE> etc.
		2787
		2788	These are just enum values with the same values as the C<EV_READ> etc.
		2789	macros from F<ev.h>.
		2790
		2791	=item C<ev::tstamp>, C<ev::now>
		2792
		2793	Aliases to the same types/functions as with the C<ev_> prefix.
		2794
		2795	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		2796
		2797	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		2798	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		2799	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		2800	defines by many implementations.
		2801
		2802	All of those classes have these methods:
		2803
		2804	=over 4
		2805
		2806	=item ev::TYPE::TYPE ()
		2807
		2808	=item ev::TYPE::TYPE (struct ev_loop *)
		2809
		2810	=item ev::TYPE::~TYPE
		2811
		2812	The constructor (optionally) takes an event loop to associate the watcher
		2813	with. If it is omitted, it will use C<EV_DEFAULT>.
		2814
		2815	The constructor calls C<ev_init> for you, which means you have to call the
		2816	C<set> method before starting it.
		2817
		2818	It will not set a callback, however: You have to call the templated C<set>
		2819	method to set a callback before you can start the watcher.
		2820
		2821	(The reason why you have to use a method is a limitation in C++ which does
		2822	not allow explicit template arguments for constructors).
		2823
		2824	The destructor automatically stops the watcher if it is active.
		2825
		2826	=item w->set<class, &class::method> (object *)
		2827
		2828	This method sets the callback method to call. The method has to have a
		2829	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		2830	first argument and the C<revents> as second. The object must be given as
		2831	parameter and is stored in the C<data> member of the watcher.
		2832
		2833	This method synthesizes efficient thunking code to call your method from
		2834	the C callback that libev requires. If your compiler can inline your
		2835	callback (i.e. it is visible to it at the place of the C<set> call and
		2836	your compiler is good :), then the method will be fully inlined into the
		2837	thunking function, making it as fast as a direct C callback.
		2838
		2839	Example: simple class declaration and watcher initialisation
		2840
		2841	struct myclass
		2842	{
		2843	void io_cb (ev::io &w, int revents) { }
		2844	}
		2845
		2846	myclass obj;
		2847	ev::io iow;
		2848	iow.set <myclass, &myclass::io_cb> (&obj);
		2849
		2850	=item w->set<function> (void *data = 0)
		2851
		2852	Also sets a callback, but uses a static method or plain function as
		2853	callback. The optional C<data> argument will be stored in the watcher's
		2854	C<data> member and is free for you to use.
		2855
		2856	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		2857
		2858	See the method-C<set> above for more details.
		2859
		2860	Example: Use a plain function as callback.
		2861
		2862	static void io_cb (ev::io &w, int revents) { }
		2863	iow.set <io_cb> ();
		2864
		2865	=item w->set (struct ev_loop *)
		2866
		2867	Associates a different C<struct ev_loop> with this watcher. You can only
		2868	do this when the watcher is inactive (and not pending either).
		2869
		2870	=item w->set ([arguments])
		2871
		2872	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		2873	called at least once. Unlike the C counterpart, an active watcher gets
		2874	automatically stopped and restarted when reconfiguring it with this
		2875	method.
		2876
		2877	=item w->start ()
		2878
		2879	Starts the watcher. Note that there is no C<loop> argument, as the
		2880	constructor already stores the event loop.
		2881
		2882	=item w->stop ()
		2883
		2884	Stops the watcher if it is active. Again, no C<loop> argument.
		2885
		2886	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		2887
		2888	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		2889	C<ev_TYPE_again> function.
		2890
		2891	=item w->sweep () (C<ev::embed> only)
		2892
		2893	Invokes C<ev_embed_sweep>.
		2894
		2895	=item w->update () (C<ev::stat> only)
		2896
		2897	Invokes C<ev_stat_stat>.
		2898
		2899	=back
		2900
		2901	=back
		2902
		2903	Example: Define a class with an IO and idle watcher, start one of them in
		2904	the constructor.
		2905
		2906	class myclass
		2907	{
		2908	ev::io io ; void io_cb (ev::io &w, int revents);
		2909	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		2910
		2911	myclass (int fd)
		2912	{
		2913	io .set <myclass, &myclass::io_cb > (this);
		2914	idle.set <myclass, &myclass::idle_cb> (this);
		2915
		2916	io.start (fd, ev::READ);
		2917	}
		2918	};
		2919
		2920
		2921	=head1 OTHER LANGUAGE BINDINGS
		2922
		2923	Libev does not offer other language bindings itself, but bindings for a
		2924	number of languages exist in the form of third-party packages. If you know
		2925	any interesting language binding in addition to the ones listed here, drop
		2926	me a note.
		2927
		2928	=over 4
		2929
		2930	=item Perl
		2931
		2932	The EV module implements the full libev API and is actually used to test
		2933	libev. EV is developed together with libev. Apart from the EV core module,
		2934	there are additional modules that implement libev-compatible interfaces
		2935	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		2936	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		2937	and C<EV::Glib>).
		2938
		2939	It can be found and installed via CPAN, its homepage is at
		2940	L<http://software.schmorp.de/pkg/EV>.
		2941
		2942	=item Python
		2943
		2944	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		2945	seems to be quite complete and well-documented. Note, however, that the
		2946	patch they require for libev is outright dangerous as it breaks the ABI
		2947	for everybody else, and therefore, should never be applied in an installed
		2948	libev (if python requires an incompatible ABI then it needs to embed
		2949	libev).
		2950
		2951	=item Ruby
		2952
		2953	Tony Arcieri has written a ruby extension that offers access to a subset
		2954	of the libev API and adds file handle abstractions, asynchronous DNS and
		2955	more on top of it. It can be found via gem servers. Its homepage is at
		2956	L<http://rev.rubyforge.org/>.
		2957
		2958	=item D
		2959
		2960	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		2961	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		2962
		2963	=item Ocaml
		2964
		2965	Erkki Seppala has written Ocaml bindings for libev, to be found at
		2966	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		2967
		2968	=back
		2969
		2970
		2971	=head1 MACRO MAGIC
		2972
		2973	Libev can be compiled with a variety of options, the most fundamental
		2974	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		2975	functions and callbacks have an initial C<struct ev_loop *> argument.
		2976
		2977	To make it easier to write programs that cope with either variant, the
		2978	following macros are defined:
		2979
		2980	=over 4
		2981
		2982	=item C<EV_A>, C<EV_A_>
		2983
		2984	This provides the loop I<argument> for functions, if one is required ("ev
		2985	loop argument"). The C<EV_A> form is used when this is the sole argument,
		2986	C<EV_A_> is used when other arguments are following. Example:
		2987
		2988	ev_unref (EV_A);
		2989	ev_timer_add (EV_A_ watcher);
		2990	ev_loop (EV_A_ 0);
		2991
		2992	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		2993	which is often provided by the following macro.
		2994
		2995	=item C<EV_P>, C<EV_P_>
		2996
		2997	This provides the loop I<parameter> for functions, if one is required ("ev
		2998	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		2999	C<EV_P_> is used when other parameters are following. Example:
		3000
		3001	// this is how ev_unref is being declared
		3002	static void ev_unref (EV_P);
		3003
		3004	// this is how you can declare your typical callback
		3005	static void cb (EV_P_ ev_timer *w, int revents)
		3006
		3007	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3008	suitable for use with C<EV_A>.
		3009
		3010	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3011
		3012	Similar to the other two macros, this gives you the value of the default
		3013	loop, if multiple loops are supported ("ev loop default").
		3014
		3015	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3016
		3017	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3018	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3019	is undefined when the default loop has not been initialised by a previous
		3020	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3021
		3022	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3023	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3024
		3025	=back
		3026
		3027	Example: Declare and initialise a check watcher, utilising the above
		3028	macros so it will work regardless of whether multiple loops are supported
		3029	or not.
		3030
		3031	static void
		3032	check_cb (EV_P_ ev_timer *w, int revents)
		3033	{
		3034	ev_check_stop (EV_A_ w);
		3035	}
		3036
		3037	ev_check check;
		3038	ev_check_init (&check, check_cb);
		3039	ev_check_start (EV_DEFAULT_ &check);
		3040	ev_loop (EV_DEFAULT_ 0);
		3041
		3042	=head1 EMBEDDING
		3043
		3044	Libev can (and often is) directly embedded into host
		3045	applications. Examples of applications that embed it include the Deliantra
		3046	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3047	and rxvt-unicode.
		3048
		3049	The goal is to enable you to just copy the necessary files into your
		3050	source directory without having to change even a single line in them, so
		3051	you can easily upgrade by simply copying (or having a checked-out copy of
		3052	libev somewhere in your source tree).
		3053
		3054	=head2 FILESETS
		3055
		3056	Depending on what features you need you need to include one or more sets of files
		3057	in your application.
		3058
		3059	=head3 CORE EVENT LOOP
		3060
		3061	To include only the libev core (all the C<ev_*> functions), with manual
		3062	configuration (no autoconf):
		3063
		3064	#define EV_STANDALONE 1
		3065	#include "ev.c"
		3066
		3067	This will automatically include F<ev.h>, too, and should be done in a
		3068	single C source file only to provide the function implementations. To use
		3069	it, do the same for F<ev.h> in all files wishing to use this API (best
		3070	done by writing a wrapper around F<ev.h> that you can include instead and
		3071	where you can put other configuration options):
		3072
		3073	#define EV_STANDALONE 1
		3074	#include "ev.h"
		3075
		3076	Both header files and implementation files can be compiled with a C++
		3077	compiler (at least, thats a stated goal, and breakage will be treated
		3078	as a bug).
		3079
		3080	You need the following files in your source tree, or in a directory
		3081	in your include path (e.g. in libev/ when using -Ilibev):
		3082
		3083	ev.h
		3084	ev.c
		3085	ev_vars.h
		3086	ev_wrap.h
		3087
		3088	ev_win32.c required on win32 platforms only
		3089
		3090	ev_select.c only when select backend is enabled (which is enabled by default)
		3091	ev_poll.c only when poll backend is enabled (disabled by default)
		3092	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3093	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3094	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3095
		3096	F<ev.c> includes the backend files directly when enabled, so you only need
		3097	to compile this single file.
		3098
		3099	=head3 LIBEVENT COMPATIBILITY API
		3100
		3101	To include the libevent compatibility API, also include:
		3102
		3103	#include "event.c"
		3104
		3105	in the file including F<ev.c>, and:
		3106
		3107	#include "event.h"
		3108
		3109	in the files that want to use the libevent API. This also includes F<ev.h>.
		3110
		3111	You need the following additional files for this:
		3112
		3113	event.h
		3114	event.c
		3115
		3116	=head3 AUTOCONF SUPPORT
		3117
		3118	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3119	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3120	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3121	include F<config.h> and configure itself accordingly.
		3122
		3123	For this of course you need the m4 file:
		3124
		3125	libev.m4
		3126
		3127	=head2 PREPROCESSOR SYMBOLS/MACROS
		3128
		3129	Libev can be configured via a variety of preprocessor symbols you have to
		3130	define before including any of its files. The default in the absence of
		3131	autoconf is documented for every option.
		3132
		3133	=over 4
		3134
		3135	=item EV_STANDALONE
		3136
		3137	Must always be C<1> if you do not use autoconf configuration, which
		3138	keeps libev from including F<config.h>, and it also defines dummy
		3139	implementations for some libevent functions (such as logging, which is not
		3140	supported). It will also not define any of the structs usually found in
		3141	F<event.h> that are not directly supported by the libev core alone.
		3142
		3143	=item EV_USE_MONOTONIC
		3144
		3145	If defined to be C<1>, libev will try to detect the availability of the
		3146	monotonic clock option at both compile time and runtime. Otherwise no use
		3147	of the monotonic clock option will be attempted. If you enable this, you
		3148	usually have to link against librt or something similar. Enabling it when
		3149	the functionality isn't available is safe, though, although you have
		3150	to make sure you link against any libraries where the C<clock_gettime>
		3151	function is hiding in (often F<-lrt>).
		3152
		3153	=item EV_USE_REALTIME
		3154
		3155	If defined to be C<1>, libev will try to detect the availability of the
		3156	real-time clock option at compile time (and assume its availability at
		3157	runtime if successful). Otherwise no use of the real-time clock option will
		3158	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
		3159	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
		3160	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
		3161
		3162	=item EV_USE_NANOSLEEP
		3163
		3164	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3165	and will use it for delays. Otherwise it will use C<select ()>.
		3166
		3167	=item EV_USE_EVENTFD
		3168
		3169	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3170	available and will probe for kernel support at runtime. This will improve
		3171	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3172	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3173	2.7 or newer, otherwise disabled.
		3174
		3175	=item EV_USE_SELECT
		3176
		3177	If undefined or defined to be C<1>, libev will compile in support for the
		3178	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3179	other method takes over, select will be it. Otherwise the select backend
		3180	will not be compiled in.
		3181
		3182	=item EV_SELECT_USE_FD_SET
		3183
		3184	If defined to C<1>, then the select backend will use the system C<fd_set>
		3185	structure. This is useful if libev doesn't compile due to a missing
		3186	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on
		3187	exotic systems. This usually limits the range of file descriptors to some
		3188	low limit such as 1024 or might have other limitations (winsocket only
		3189	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
		3190	influence the size of the C<fd_set> used.
		3191
		3192	=item EV_SELECT_IS_WINSOCKET
		3193
		3194	When defined to C<1>, the select backend will assume that
		3195	select/socket/connect etc. don't understand file descriptors but
		3196	wants osf handles on win32 (this is the case when the select to
		3197	be used is the winsock select). This means that it will call
		3198	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3199	it is assumed that all these functions actually work on fds, even
		3200	on win32. Should not be defined on non-win32 platforms.
		3201
		3202	=item EV_FD_TO_WIN32_HANDLE
		3203
		3204	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3205	file descriptors to socket handles. When not defining this symbol (the
		3206	default), then libev will call C<_get_osfhandle>, which is usually
		3207	correct. In some cases, programs use their own file descriptor management,
		3208	in which case they can provide this function to map fds to socket handles.
		3209
		3210	=item EV_USE_POLL
		3211
		3212	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3213	backend. Otherwise it will be enabled on non-win32 platforms. It
		3214	takes precedence over select.
		3215
		3216	=item EV_USE_EPOLL
		3217
		3218	If defined to be C<1>, libev will compile in support for the Linux
		3219	C<epoll>(7) backend. Its availability will be detected at runtime,
		3220	otherwise another method will be used as fallback. This is the preferred
		3221	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3222	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3223
		3224	=item EV_USE_KQUEUE
		3225
		3226	If defined to be C<1>, libev will compile in support for the BSD style
		3227	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3228	otherwise another method will be used as fallback. This is the preferred
		3229	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3230	supports some types of fds correctly (the only platform we found that
		3231	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3232	not be used unless explicitly requested. The best way to use it is to find
		3233	out whether kqueue supports your type of fd properly and use an embedded
		3234	kqueue loop.
		3235
		3236	=item EV_USE_PORT
		3237
		3238	If defined to be C<1>, libev will compile in support for the Solaris
		3239	10 port style backend. Its availability will be detected at runtime,
		3240	otherwise another method will be used as fallback. This is the preferred
		3241	backend for Solaris 10 systems.
		3242
		3243	=item EV_USE_DEVPOLL
		3244
		3245	Reserved for future expansion, works like the USE symbols above.
		3246
		3247	=item EV_USE_INOTIFY
		3248
		3249	If defined to be C<1>, libev will compile in support for the Linux inotify
		3250	interface to speed up C<ev_stat> watchers. Its actual availability will
		3251	be detected at runtime. If undefined, it will be enabled if the headers
		3252	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3253
		3254	=item EV_ATOMIC_T
		3255
		3256	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3257	access is atomic with respect to other threads or signal contexts. No such
		3258	type is easily found in the C language, so you can provide your own type
		3259	that you know is safe for your purposes. It is used both for signal handler "locking"
		3260	as well as for signal and thread safety in C<ev_async> watchers.
		3261
		3262	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3263	(from F<signal.h>), which is usually good enough on most platforms.
		3264
		3265	=item EV_H
		3266
		3267	The name of the F<ev.h> header file used to include it. The default if
		3268	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3269	used to virtually rename the F<ev.h> header file in case of conflicts.
		3270
		3271	=item EV_CONFIG_H
		3272
		3273	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3274	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3275	C<EV_H>, above.
		3276
		3277	=item EV_EVENT_H
		3278
		3279	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3280	of how the F<event.h> header can be found, the default is C<"event.h">.
		3281
		3282	=item EV_PROTOTYPES
		3283
		3284	If defined to be C<0>, then F<ev.h> will not define any function
		3285	prototypes, but still define all the structs and other symbols. This is
		3286	occasionally useful if you want to provide your own wrapper functions
		3287	around libev functions.
		3288
		3289	=item EV_MULTIPLICITY
		3290
		3291	If undefined or defined to C<1>, then all event-loop-specific functions
		3292	will have the C<struct ev_loop *> as first argument, and you can create
		3293	additional independent event loops. Otherwise there will be no support
		3294	for multiple event loops and there is no first event loop pointer
		3295	argument. Instead, all functions act on the single default loop.
		3296
		3297	=item EV_MINPRI
		3298
		3299	=item EV_MAXPRI
		3300
		3301	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3302	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3303	provide for more priorities by overriding those symbols (usually defined
		3304	to be C<-2> and C<2>, respectively).
		3305
		3306	When doing priority-based operations, libev usually has to linearly search
		3307	all the priorities, so having many of them (hundreds) uses a lot of space
		3308	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3309	fine.
		3310
		3311	If your embedding application does not need any priorities, defining these
		3312	both to C<0> will save some memory and CPU.
		3313
		3314	=item EV_PERIODIC_ENABLE
		3315
		3316	If undefined or defined to be C<1>, then periodic timers are supported. If
		3317	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3318	code.
		3319
		3320	=item EV_IDLE_ENABLE
		3321
		3322	If undefined or defined to be C<1>, then idle watchers are supported. If
		3323	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3324	code.
		3325
		3326	=item EV_EMBED_ENABLE
		3327
		3328	If undefined or defined to be C<1>, then embed watchers are supported. If
		3329	defined to be C<0>, then they are not. Embed watchers rely on most other
		3330	watcher types, which therefore must not be disabled.
		3331
		3332	=item EV_STAT_ENABLE
		3333
		3334	If undefined or defined to be C<1>, then stat watchers are supported. If
		3335	defined to be C<0>, then they are not.
		3336
		3337	=item EV_FORK_ENABLE
		3338
		3339	If undefined or defined to be C<1>, then fork watchers are supported. If
		3340	defined to be C<0>, then they are not.
		3341
		3342	=item EV_ASYNC_ENABLE
		3343
		3344	If undefined or defined to be C<1>, then async watchers are supported. If
		3345	defined to be C<0>, then they are not.
		3346
		3347	=item EV_MINIMAL
		3348
		3349	If you need to shave off some kilobytes of code at the expense of some
		3350	speed, define this symbol to C<1>. Currently this is used to override some
		3351	inlining decisions, saves roughly 30% code size on amd64. It also selects a
		3352	much smaller 2-heap for timer management over the default 4-heap.
		3353
		3354	=item EV_PID_HASHSIZE
		3355
		3356	C<ev_child> watchers use a small hash table to distribute workload by
		3357	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3358	than enough. If you need to manage thousands of children you might want to
		3359	increase this value (I<must> be a power of two).
		3360
		3361	=item EV_INOTIFY_HASHSIZE
		3362
		3363	C<ev_stat> watchers use a small hash table to distribute workload by
		3364	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3365	usually more than enough. If you need to manage thousands of C<ev_stat>
		3366	watchers you might want to increase this value (I<must> be a power of
		3367	two).
		3368
		3369	=item EV_USE_4HEAP
		3370
		3371	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3372	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3373	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3374	faster performance with many (thousands) of watchers.
		3375
		3376	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3377	(disabled).
		3378
		3379	=item EV_HEAP_CACHE_AT
		3380
		3381	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3382	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3383	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3384	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3385	but avoids random read accesses on heap changes. This improves performance
		3386	noticeably with many (hundreds) of watchers.
		3387
		3388	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3389	(disabled).
		3390
		3391	=item EV_VERIFY
		3392
		3393	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3394	be done: If set to C<0>, no internal verification code will be compiled
		3395	in. If set to C<1>, then verification code will be compiled in, but not
		3396	called. If set to C<2>, then the internal verification code will be
		3397	called once per loop, which can slow down libev. If set to C<3>, then the
		3398	verification code will be called very frequently, which will slow down
		3399	libev considerably.
		3400
		3401	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3402	C<0>.
		3403
		3404	=item EV_COMMON
		3405
		3406	By default, all watchers have a C<void *data> member. By redefining
		3407	this macro to a something else you can include more and other types of
		3408	members. You have to define it each time you include one of the files,
		3409	though, and it must be identical each time.
		3410
		3411	For example, the perl EV module uses something like this:
		3412
		3413	#define EV_COMMON \
		3414	SV *self; /* contains this struct */ \
		3415	SV *cb_sv, *fh /* note no trailing ";" */
		3416
		3417	=item EV_CB_DECLARE (type)
		3418
		3419	=item EV_CB_INVOKE (watcher, revents)
		3420
		3421	=item ev_set_cb (ev, cb)
		3422
		3423	Can be used to change the callback member declaration in each watcher,
		3424	and the way callbacks are invoked and set. Must expand to a struct member
		3425	definition and a statement, respectively. See the F<ev.h> header file for
		3426	their default definitions. One possible use for overriding these is to
		3427	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3428	method calls instead of plain function calls in C++.
		3429
		3430	=back
		3431
		3432	=head2 EXPORTED API SYMBOLS
		3433
		3434	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3435	exported symbols, you can use the provided F<Symbol.*> files which list
		3436	all public symbols, one per line:
		3437
		3438	Symbols.ev for libev proper
		3439	Symbols.event for the libevent emulation
		3440
		3441	This can also be used to rename all public symbols to avoid clashes with
		3442	multiple versions of libev linked together (which is obviously bad in
		3443	itself, but sometimes it is inconvenient to avoid this).
		3444
		3445	A sed command like this will create wrapper C<#define>'s that you need to
		3446	include before including F<ev.h>:
		3447
		3448	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3449
		3450	This would create a file F<wrap.h> which essentially looks like this:
		3451
		3452	#define ev_backend myprefix_ev_backend
		3453	#define ev_check_start myprefix_ev_check_start
		3454	#define ev_check_stop myprefix_ev_check_stop
		3455	...
		3456
		3457	=head2 EXAMPLES
		3458
		3459	For a real-world example of a program the includes libev
		3460	verbatim, you can have a look at the EV perl module
		3461	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3462	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3463	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3464	will be compiled. It is pretty complex because it provides its own header
		3465	file.
		3466
		3467	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3468	that everybody includes and which overrides some configure choices:
		3469
		3470	#define EV_MINIMAL 1
		3471	#define EV_USE_POLL 0
		3472	#define EV_MULTIPLICITY 0
		3473	#define EV_PERIODIC_ENABLE 0
		3474	#define EV_STAT_ENABLE 0
		3475	#define EV_FORK_ENABLE 0
		3476	#define EV_CONFIG_H <config.h>
		3477	#define EV_MINPRI 0
		3478	#define EV_MAXPRI 0
		3479
		3480	#include "ev++.h"
		3481
		3482	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3483
		3484	#include "ev_cpp.h"
		3485	#include "ev.c"
		3486
		3487	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3488
		3489	=head2 THREADS AND COROUTINES
		3490
		3491	=head3 THREADS
		3492
		3493	All libev functions are reentrant and thread-safe unless explicitly
		3494	documented otherwise, but libev implements no locking itself. This means
		3495	that you can use as many loops as you want in parallel, as long as there
		3496	are no concurrent calls into any libev function with the same loop
		3497	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3498	of course): libev guarantees that different event loops share no data
		3499	structures that need any locking.
		3500
		3501	Or to put it differently: calls with different loop parameters can be done
		3502	concurrently from multiple threads, calls with the same loop parameter
		3503	must be done serially (but can be done from different threads, as long as
		3504	only one thread ever is inside a call at any point in time, e.g. by using
		3505	a mutex per loop).
		3506
		3507	Specifically to support threads (and signal handlers), libev implements
		3508	so-called C<ev_async> watchers, which allow some limited form of
		3509	concurrency on the same event loop, namely waking it up "from the
		3510	outside".
		3511
		3512	If you want to know which design (one loop, locking, or multiple loops
		3513	without or something else still) is best for your problem, then I cannot
		3514	help you, but here is some generic advice:
		3515
		3516	=over 4
		3517
		3518	=item * most applications have a main thread: use the default libev loop
		3519	in that thread, or create a separate thread running only the default loop.
		3520
		3521	This helps integrating other libraries or software modules that use libev
		3522	themselves and don't care/know about threading.
		3523
		3524	=item * one loop per thread is usually a good model.
		3525
		3526	Doing this is almost never wrong, sometimes a better-performance model
		3527	exists, but it is always a good start.
		3528
		3529	=item * other models exist, such as the leader/follower pattern, where one
		3530	loop is handed through multiple threads in a kind of round-robin fashion.
		3531
		3532	Choosing a model is hard - look around, learn, know that usually you can do
		3533	better than you currently do :-)
		3534
		3535	=item * often you need to talk to some other thread which blocks in the
		3536	event loop.
		3537
		3538	C<ev_async> watchers can be used to wake them up from other threads safely
		3539	(or from signal contexts...).
		3540
		3541	An example use would be to communicate signals or other events that only
		3542	work in the default loop by registering the signal watcher with the
		3543	default loop and triggering an C<ev_async> watcher from the default loop
		3544	watcher callback into the event loop interested in the signal.
		3545
		3546	=back
		3547
		3548	=head3 COROUTINES
		3549
		3550	Libev is very accommodating to coroutines ("cooperative threads"):
		3551	libev fully supports nesting calls to its functions from different
		3552	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3553	different coroutines, and switch freely between both coroutines running the
		3554	loop, as long as you don't confuse yourself). The only exception is that
		3555	you must not do this from C<ev_periodic> reschedule callbacks.
		3556
		3557	Care has been taken to ensure that libev does not keep local state inside
		3558	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3559	they do not clal any callbacks.
		3560
		3561	=head2 COMPILER WARNINGS
		3562
		3563	Depending on your compiler and compiler settings, you might get no or a
		3564	lot of warnings when compiling libev code. Some people are apparently
		3565	scared by this.
		3566
		3567	However, these are unavoidable for many reasons. For one, each compiler
		3568	has different warnings, and each user has different tastes regarding
		3569	warning options. "Warn-free" code therefore cannot be a goal except when
		3570	targeting a specific compiler and compiler-version.
		3571
		3572	Another reason is that some compiler warnings require elaborate
		3573	workarounds, or other changes to the code that make it less clear and less
		3574	maintainable.
		3575
		3576	And of course, some compiler warnings are just plain stupid, or simply
		3577	wrong (because they don't actually warn about the condition their message
		3578	seems to warn about). For example, certain older gcc versions had some
		3579	warnings that resulted an extreme number of false positives. These have
		3580	been fixed, but some people still insist on making code warn-free with
		3581	such buggy versions.
		3582
		3583	While libev is written to generate as few warnings as possible,
		3584	"warn-free" code is not a goal, and it is recommended not to build libev
		3585	with any compiler warnings enabled unless you are prepared to cope with
		3586	them (e.g. by ignoring them). Remember that warnings are just that:
		3587	warnings, not errors, or proof of bugs.
		3588
		3589
		3590	=head2 VALGRIND
		3591
		3592	Valgrind has a special section here because it is a popular tool that is
		3593	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		3594
		3595	If you think you found a bug (memory leak, uninitialised data access etc.)
		3596	in libev, then check twice: If valgrind reports something like:
		3597
		3598	==2274== definitely lost: 0 bytes in 0 blocks.
		3599	==2274== possibly lost: 0 bytes in 0 blocks.
		3600	==2274== still reachable: 256 bytes in 1 blocks.
		3601
		3602	Then there is no memory leak, just as memory accounted to global variables
		3603	is not a memleak - the memory is still being refernced, and didn't leak.
		3604
		3605	Similarly, under some circumstances, valgrind might report kernel bugs
		3606	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3607	although an acceptable workaround has been found here), or it might be
		3608	confused.
		3609
		3610	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		3611	make it into some kind of religion.
		3612
		3613	If you are unsure about something, feel free to contact the mailing list
		3614	with the full valgrind report and an explanation on why you think this
		3615	is a bug in libev (best check the archives, too :). However, don't be
		3616	annoyed when you get a brisk "this is no bug" answer and take the chance
		3617	of learning how to interpret valgrind properly.
		3618
		3619	If you need, for some reason, empty reports from valgrind for your project
		3620	I suggest using suppression lists.
		3621
		3622
		3623	=head1 PORTABILITY NOTES
		3624
		3625	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		3626
		3627	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3628	requires, and its I/O model is fundamentally incompatible with the POSIX
		3629	model. Libev still offers limited functionality on this platform in
		3630	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3631	descriptors. This only applies when using Win32 natively, not when using
		3632	e.g. cygwin.
		3633
		3634	Lifting these limitations would basically require the full
		3635	re-implementation of the I/O system. If you are into these kinds of
		3636	things, then note that glib does exactly that for you in a very portable
		3637	way (note also that glib is the slowest event library known to man).
		3638
		3639	There is no supported compilation method available on windows except
		3640	embedding it into other applications.
		3641
		3642	Not a libev limitation but worth mentioning: windows apparently doesn't
		3643	accept large writes: instead of resulting in a partial write, windows will
		3644	either accept everything or return C<ENOBUFS> if the buffer is too large,
		3645	so make sure you only write small amounts into your sockets (less than a
		3646	megabyte seems safe, but this apparently depends on the amount of memory
		3647	available).
		3648
		3649	Due to the many, low, and arbitrary limits on the win32 platform and
		3650	the abysmal performance of winsockets, using a large number of sockets
		3651	is not recommended (and not reasonable). If your program needs to use
		3652	more than a hundred or so sockets, then likely it needs to use a totally
		3653	different implementation for windows, as libev offers the POSIX readiness
		3654	notification model, which cannot be implemented efficiently on windows
		3655	(Microsoft monopoly games).
		3656
		3657	A typical way to use libev under windows is to embed it (see the embedding
		3658	section for details) and use the following F<evwrap.h> header file instead
		3659	of F<ev.h>:
		3660
		3661	#define EV_STANDALONE /* keeps ev from requiring config.h */
		3662	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		3663
		3664	#include "ev.h"
		3665
		3666	And compile the following F<evwrap.c> file into your project (make sure
		3667	you do I<not> compile the F<ev.c> or any other embedded source files!):
		3668
		3669	#include "evwrap.h"
		3670	#include "ev.c"
		3671
		3672	=over 4
		3673
		3674	=item The winsocket select function
		3675
		3676	The winsocket C<select> function doesn't follow POSIX in that it
		3677	requires socket I<handles> and not socket I<file descriptors> (it is
		3678	also extremely buggy). This makes select very inefficient, and also
		3679	requires a mapping from file descriptors to socket handles (the Microsoft
		3680	C runtime provides the function C<_open_osfhandle> for this). See the
		3681	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		3682	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		3683
		3684	The configuration for a "naked" win32 using the Microsoft runtime
		3685	libraries and raw winsocket select is:
		3686
		3687	#define EV_USE_SELECT 1
		3688	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3689
		3690	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3691	complexity in the O(n²) range when using win32.
		3692
		3693	=item Limited number of file descriptors
		3694
		3695	Windows has numerous arbitrary (and low) limits on things.
		3696
		3697	Early versions of winsocket's select only supported waiting for a maximum
		3698	of C<64> handles (probably owning to the fact that all windows kernels
		3699	can only wait for C<64> things at the same time internally; Microsoft
		3700	recommends spawning a chain of threads and wait for 63 handles and the
		3701	previous thread in each. Great).
		3702
		3703	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3704	to some high number (e.g. C<2048>) before compiling the winsocket select
		3705	call (which might be in libev or elsewhere, for example, perl does its own
		3706	select emulation on windows).
		3707
		3708	Another limit is the number of file descriptors in the Microsoft runtime
		3709	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3710	or something like this inside Microsoft). You can increase this by calling
		3711	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3712	arbitrary limit), but is broken in many versions of the Microsoft runtime
		3713	libraries.
		3714
		3715	This might get you to about C<512> or C<2048> sockets (depending on
		3716	windows version and/or the phase of the moon). To get more, you need to
		3717	wrap all I/O functions and provide your own fd management, but the cost of
		3718	calling select (O(n²)) will likely make this unworkable.
		3719
		3720	=back
		3721
		3722	=head2 PORTABILITY REQUIREMENTS
		3723
		3724	In addition to a working ISO-C implementation and of course the
		3725	backend-specific APIs, libev relies on a few additional extensions:
		3726
		3727	=over 4
		3728
		3729	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		3730	calling conventions regardless of C<ev_watcher_type *>.
		3731
		3732	Libev assumes not only that all watcher pointers have the same internal
		3733	structure (guaranteed by POSIX but not by ISO C for example), but it also
		3734	assumes that the same (machine) code can be used to call any watcher
		3735	callback: The watcher callbacks have different type signatures, but libev
		3736	calls them using an C<ev_watcher *> internally.
		3737
		3738	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3739
		3740	The type C<sig_atomic_t volatile> (or whatever is defined as
		3741	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		3742	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3743	believed to be sufficiently portable.
		3744
		3745	=item C<sigprocmask> must work in a threaded environment
		3746
		3747	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3748	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3749	pthread implementations will either allow C<sigprocmask> in the "main
		3750	thread" or will block signals process-wide, both behaviours would
		3751	be compatible with libev. Interaction between C<sigprocmask> and
		3752	C<pthread_sigmask> could complicate things, however.
		3753
		3754	The most portable way to handle signals is to block signals in all threads
		3755	except the initial one, and run the default loop in the initial thread as
		3756	well.
		3757
		3758	=item C<long> must be large enough for common memory allocation sizes
		3759
		3760	To improve portability and simplify its API, libev uses C<long> internally
		3761	instead of C<size_t> when allocating its data structures. On non-POSIX
		3762	systems (Microsoft...) this might be unexpectedly low, but is still at
		3763	least 31 bits everywhere, which is enough for hundreds of millions of
		3764	watchers.
		3765
		3766	=item C<double> must hold a time value in seconds with enough accuracy
		3767
		3768	The type C<double> is used to represent timestamps. It is required to
		3769	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3770	enough for at least into the year 4000. This requirement is fulfilled by
		3771	implementations implementing IEEE 754 (basically all existing ones).
		3772
		3773	=back
		3774
		3775	If you know of other additional requirements drop me a note.
		3776
		3777
		3778	=head1 ALGORITHMIC COMPLEXITIES
		3779
		3780	In this section the complexities of (many of) the algorithms used inside
		3781	libev will be documented. For complexity discussions about backends see
		3782	the documentation for C<ev_default_init>.
		3783
		3784	All of the following are about amortised time: If an array needs to be
		3785	extended, libev needs to realloc and move the whole array, but this
		3786	happens asymptotically rarer with higher number of elements, so O(1) might
		3787	mean that libev does a lengthy realloc operation in rare cases, but on
		3788	average it is much faster and asymptotically approaches constant time.
		3789
		3790	=over 4
		3791
		3792	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		3793
		3794	This means that, when you have a watcher that triggers in one hour and
		3795	there are 100 watchers that would trigger before that, then inserting will
		3796	have to skip roughly seven (C<ld 100>) of these watchers.
		3797
		3798	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		3799
		3800	That means that changing a timer costs less than removing/adding them,
		3801	as only the relative motion in the event queue has to be paid for.
		3802
		3803	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		3804
		3805	These just add the watcher into an array or at the head of a list.
		3806
		3807	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		3808
		3809	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		3810
		3811	These watchers are stored in lists, so they need to be walked to find the
		3812	correct watcher to remove. The lists are usually short (you don't usually
		3813	have many watchers waiting for the same fd or signal: one is typical, two
		3814	is rare).
		3815
		3816	=item Finding the next timer in each loop iteration: O(1)
		3817
		3818	By virtue of using a binary or 4-heap, the next timer is always found at a
		3819	fixed position in the storage array.
		3820
		3821	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3822
		3823	A change means an I/O watcher gets started or stopped, which requires
		3824	libev to recalculate its status (and possibly tell the kernel, depending
		3825	on backend and whether C<ev_io_set> was used).
		3826
		3827	=item Activating one watcher (putting it into the pending state): O(1)
		3828
		3829	=item Priority handling: O(number_of_priorities)
		3830
		3831	Priorities are implemented by allocating some space for each
		3832	priority. When doing priority-based operations, libev usually has to
		3833	linearly search all the priorities, but starting/stopping and activating
		3834	watchers becomes O(1) with respect to priority handling.
		3835
		3836	=item Sending an ev_async: O(1)
		3837
		3838	=item Processing ev_async_send: O(number_of_async_watchers)
		3839
		3840	=item Processing signals: O(max_signal_number)
		3841
		3842	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3843	calls in the current loop iteration. Checking for async and signal events
		3844	involves iterating over all running async watchers or all signal numbers.
		3845
		3846	=back
		3847
		3848
722	=head1 AUTHOR	3849	=head1 AUTHOR
723		3850
724	Marc Lehmann <libev@schmorp.de>.	3851	Marc Lehmann <libev@schmorp.de>.
725		3852

Diff Legend

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