ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.16 by root, Mon Nov 12 08:47:14 2007 UTC vs.
Revision 1.210 by root, Thu Oct 30 08:09:30 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 with events.	73	these event sources and provide your program with events.
14		74
15	To do this, it must take more or less complete control over your process	75	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	76	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	77	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	79	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	80	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	81	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	82	watcher.
23		83
24	=head1 FEATURES	84	=head2 FEATURES
25		85
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	86	Libev supports C<select>, C<poll>, the Linux-specific C<epoll>, the
27	kqueue mechanisms for file descriptor events, relative timers, absolute	87	BSD-specific C<kqueue> and the Solaris-specific event port mechanisms
28	timers with customised rescheduling, signal events, process status change	88	for file descriptor events (C<ev_io>), the Linux C<inotify> interface
29	events (related to SIGCHLD), and event watchers dealing with the event	89	(for C<ev_stat>), relative timers (C<ev_timer>), absolute timers
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	90	with customised rescheduling (C<ev_periodic>), synchronous signals
		91	(C<ev_signal>), process status change events (C<ev_child>), and event
		92	watchers dealing with the event loop mechanism itself (C<ev_idle>,
		93	C<ev_embed>, C<ev_prepare> and C<ev_check> watchers) as well as
		94	file watchers (C<ev_stat>) and even limited support for fork events
		95	(C<ev_fork>).
		96
		97	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	98	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	99	for example).
33		100
34	=head1 CONVENTIONS	101	=head2 CONVENTIONS
35		102
36	Libev is very configurable. In this manual the default configuration	103	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	104	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	105	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	106	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	107	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	108	name C<loop> (which is always of type C<ev_loop *>) will not have
42	will not have this argument.	109	this argument.
43		110
44	=head1 TIME AND OTHER GLOBAL FUNCTIONS	111	=head2 TIME REPRESENTATION
45		112
46	Libev represents time as a single floating point number, representing the	113	Libev represents time as a single floating point number, representing the
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	114	(fractional) number of seconds since the (POSIX) epoch (somewhere near
48	the beginning of 1970, details are complicated, don't ask). This type is	115	the beginning of 1970, details are complicated, don't ask). This type is
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	116	called C<ev_tstamp>, which is what you should use too. It usually aliases
50	to the double type in C.	117	to the C<double> type in C, and when you need to do any calculations on
		118	it, you should treat it as some 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.
51		147
52	=over 4	148	=over 4
53		149
54	=item ev_tstamp ev_time ()	150	=item ev_tstamp ev_time ()
55		151
56	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 ()>.
57		161
58	=item int ev_version_major ()	162	=item int ev_version_major ()
59		163
60	=item int ev_version_minor ()	164	=item int ev_version_minor ()
61		165
62	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
63	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
64	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
65	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
66	version of the library your program was compiled against.	170	version of the library your program was compiled against.
67		171
		172	These version numbers refer to the ABI version of the library, not the
		173	release version.
		174
68	Usually, it's a good idea to terminate if the major versions mismatch,	175	Usually, it's a good idea to terminate if the major versions mismatch,
69	as this indicates an incompatible change. Minor versions are usually	176	as this indicates an incompatible change. Minor versions are usually
70	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
71	not a problem.	178	not a problem.
72		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
73	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	219	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
74		220
75	Sets the allocation function to use (the prototype is similar to the	221	Sets the allocation function to use (the prototype is similar - the
76	realloc C function, the semantics are identical). It is used to allocate	222	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
77	and free memory (no surprises here). If it returns zero when memory	223	used to allocate and free memory (no surprises here). If it returns zero
78	needs to be allocated, the library might abort or take some potentially	224	when memory needs to be allocated (C<size != 0>), the library might abort
79	destructive action. The default is your system realloc function.	225	or take some potentially destructive action.
		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.
80		230
81	You could override this function in high-availability programs to, say,	231	You could override this function in high-availability programs to, say,
82	free some memory if it cannot allocate memory, to use a special allocator,	232	free some memory if it cannot allocate memory, to use a special allocator,
83	or even to sleep a while and retry until some memory is available.	233	or even to sleep a while and retry until some memory is available.
84		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
85	=item ev_set_syserr_cb (void (*cb)(const char *msg));	255	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
86		256
87	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
88	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
89	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
90	callback is set, then libev will expect it to remedy the sitution, no	260	callback is set, then libev will expect it to remedy the situation, no
91	matter what, when it returns. That is, libev will generally retry the	261	matter what, when it returns. That is, libev will generally retry the
92	requested operation, or, if the condition doesn't go away, do bad stuff	262	requested operation, or, if the condition doesn't go away, do bad stuff
93	(such as abort).	263	(such as abort).
94		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
95	=back	277	=back
96		278
97	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
98		280
99	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>
100	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>
101	events, and dynamically created loops which do not.	283	I<function>).
102		284
103	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
104	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
105	create, you also create another event loop. Libev itself does no locking	287	not.
106	whatsoever, so if you mix calls to the same event loop in different
107	threads, make sure you lock (this is usually a bad idea, though, even if
108	done correctly, because it's hideous and inefficient).
109		288
110	=over 4	289	=over 4
111		290
112	=item struct ev_loop *ev_default_loop (unsigned int flags)	291	=item struct ev_loop *ev_default_loop (unsigned int flags)
113		292
114	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
115	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
116	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
117	flags).	296	flags. If that is troubling you, check C<ev_backend ()> afterwards).
118		297
119	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
120	function.	299	function.
121		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 be shared 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
122	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
123	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>).
124		314
125	It supports the following flags:	315	The following flags are supported:
126		316
127	=over 4	317	=over 4
128		318
129	=item C<EVFLAG_AUTO>	319	=item C<EVFLAG_AUTO>
130		320
131	The default flags value. Use this if you have no clue (it's the right	321	The default flags value. Use this if you have no clue (it's the right
132	thing, believe me).	322	thing, believe me).
133		323
134	=item C<EVFLAG_NOENV>	324	=item C<EVFLAG_NOENV>
135		325
136	If this flag bit is ored into the flag value (or the program runs setuid	326	If this flag bit is or'ed into the flag value (or the program runs setuid
137	or setgid) then libev will I<not> look at the environment variable	327	or setgid) then libev will I<not> look at the environment variable
138	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	328	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
139	override the flags completely if it is found in the environment. This is	329	override the flags completely if it is found in the environment. This is
140	useful to try out specific backends to test their performance, or to work	330	useful to try out specific backends to test their performance, or to work
141	around bugs.	331	around bugs.
142		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
143	=item C<EVMETHOD_SELECT> (portable select backend)	353	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
144		354
		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.
		360
		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.
		367
		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).
		371
145	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	372	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
146		373
147	=item C<EVMETHOD_EPOLL> (linux 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	=item C<EVMETHOD_KQUEUE> (some bsds only)	381	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		382	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
150		383
151	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	384	=item C<EVBACKEND_EPOLL> (value 4, Linux)
152		385
153	=item C<EVMETHOD_PORT> (solaris 10 only)	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).
154		390
155	If one or more of these are ored into the flags value, then only these	391	The epoll mechanism deserves honorable mention as the most misdesigned
156	backends will be tried (in the reverse order as given here). If one are	392	of the more advanced event mechanisms: mere annoyances include silently
157	specified, any backend will do.	393	dropping file descriptors, requiring a system call per change per file
		394	descriptor (and unnecessary guessing of parameters), problems with dup and
		395	so on. The biggest issue is fork races, however - if a program forks then
		396	I<both> parent and child process have to recreate the epoll set, which can
		397	take considerable time (one syscall per file descriptor) and is of course
		398	hard to detect.
		399
		400	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		401	of course I<doesn't>, and epoll just loves to report events for totally
		402	I<different> file descriptors (even already closed ones, so one cannot
		403	even remove them from the set) than registered in the set (especially
		404	on SMP systems). Libev tries to counter these spurious notifications by
		405	employing an additional generation counter and comparing that against the
		406	events to filter out spurious ones, recreating the set when required.
		407
		408	While stopping, setting and starting an I/O watcher in the same iteration
		409	will result in some caching, there is still a system call per such
		410	incident (because the same I<file descriptor> could point to a different
		411	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		412	file descriptors might not work very well if you register events for both
		413	file descriptors.
		414
		415	Best performance from this backend is achieved by not unregistering all
		416	watchers for a file descriptor until it has been closed, if possible,
		417	i.e. keep at least one watcher active per fd at all times. Stopping and
		418	starting a watcher (without re-setting it) also usually doesn't cause
		419	extra overhead. A fork can both result in spurious notifications as well
		420	as in libev having to destroy and recreate the epoll object, which can
		421	take considerable time and thus should be avoided.
		422
		423	While nominally embeddable in other event loops, this feature is broken in
		424	all kernel versions tested so far.
		425
		426	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		427	C<EVBACKEND_POLL>.
		428
		429	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		430
		431	Kqueue deserves special mention, as at the time of this writing, it
		432	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		433	with anything but sockets and pipes, except on Darwin, where of course
		434	it's completely useless). Unlike epoll, however, whose brokenness
		435	is by design, these kqueue bugs can (and eventually will) be fixed
		436	without API changes to existing programs. For this reason it's not being
		437	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		438	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		439	system like NetBSD.
		440
		441	You still can embed kqueue into a normal poll or select backend and use it
		442	only for sockets (after having made sure that sockets work with kqueue on
		443	the target platform). See C<ev_embed> watchers for more info.
		444
		445	It scales in the same way as the epoll backend, but the interface to the
		446	kernel is more efficient (which says nothing about its actual speed, of
		447	course). While stopping, setting and starting an I/O watcher does never
		448	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		449	two event changes per incident. Support for C<fork ()> is very bad (but
		450	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		451	cases
		452
		453	This backend usually performs well under most conditions.
		454
		455	While nominally embeddable in other event loops, this doesn't work
		456	everywhere, so you might need to test for this. And since it is broken
		457	almost everywhere, you should only use it when you have a lot of sockets
		458	(for which it usually works), by embedding it into another event loop
		459	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
		460	using it only for sockets.
		461
		462	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		463	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		464	C<NOTE_EOF>.
		465
		466	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		467
		468	This is not implemented yet (and might never be, unless you send me an
		469	implementation). According to reports, C</dev/poll> only supports sockets
		470	and is not embeddable, which would limit the usefulness of this backend
		471	immensely.
		472
		473	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		474
		475	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		476	it's really slow, but it still scales very well (O(active_fds)).
		477
		478	Please note that Solaris event ports can deliver a lot of spurious
		479	notifications, so you need to use non-blocking I/O or other means to avoid
		480	blocking when no data (or space) is available.
		481
		482	While this backend scales well, it requires one system call per active
		483	file descriptor per loop iteration. For small and medium numbers of file
		484	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		485	might perform better.
		486
		487	On the positive side, with the exception of the spurious readiness
		488	notifications, this backend actually performed fully to specification
		489	in all tests and is fully embeddable, which is a rare feat among the
		490	OS-specific backends (I vastly prefer correctness over speed hacks).
		491
		492	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		493	C<EVBACKEND_POLL>.
		494
		495	=item C<EVBACKEND_ALL>
		496
		497	Try all backends (even potentially broken ones that wouldn't be tried
		498	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		499	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		500
		501	It is definitely not recommended to use this flag.
158		502
159	=back	503	=back
		504
		505	If one or more of these are or'ed into the flags value, then only these
		506	backends will be tried (in the reverse order as listed here). If none are
		507	specified, all backends in C<ev_recommended_backends ()> will be tried.
		508
		509	Example: This is the most typical usage.
		510
		511	if (!ev_default_loop (0))
		512	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		513
		514	Example: Restrict libev to the select and poll backends, and do not allow
		515	environment settings to be taken into account:
		516
		517	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		518
		519	Example: Use whatever libev has to offer, but make sure that kqueue is
		520	used if available (warning, breaks stuff, best use only with your own
		521	private event loop and only if you know the OS supports your types of
		522	fds):
		523
		524	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
160		525
161	=item struct ev_loop *ev_loop_new (unsigned int flags)	526	=item struct ev_loop *ev_loop_new (unsigned int flags)
162		527
163	Similar to C<ev_default_loop>, but always creates a new event loop that is	528	Similar to C<ev_default_loop>, but always creates a new event loop that is
164	always distinct from the default loop. Unlike the default loop, it cannot	529	always distinct from the default loop. Unlike the default loop, it cannot
165	handle signal and child watchers, and attempts to do so will be greeted by	530	handle signal and child watchers, and attempts to do so will be greeted by
166	undefined behaviour (or a failed assertion if assertions are enabled).	531	undefined behaviour (or a failed assertion if assertions are enabled).
167		532
		533	Note that this function I<is> thread-safe, and the recommended way to use
		534	libev with threads is indeed to create one loop per thread, and using the
		535	default loop in the "main" or "initial" thread.
		536
		537	Example: Try to create a event loop that uses epoll and nothing else.
		538
		539	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		540	if (!epoller)
		541	fatal ("no epoll found here, maybe it hides under your chair");
		542
168	=item ev_default_destroy ()	543	=item ev_default_destroy ()
169		544
170	Destroys the default loop again (frees all memory and kernel state	545	Destroys the default loop again (frees all memory and kernel state
171	etc.). This stops all registered event watchers (by not touching them in	546	etc.). None of the active event watchers will be stopped in the normal
172	any way whatsoever, although you cannot rely on this :).	547	sense, so e.g. C<ev_is_active> might still return true. It is your
		548	responsibility to either stop all watchers cleanly yourself I<before>
		549	calling this function, or cope with the fact afterwards (which is usually
		550	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		551	for example).
		552
		553	Note that certain global state, such as signal state (and installed signal
		554	handlers), will not be freed by this function, and related watchers (such
		555	as signal and child watchers) would need to be stopped manually.
		556
		557	In general it is not advisable to call this function except in the
		558	rare occasion where you really need to free e.g. the signal handling
		559	pipe fds. If you need dynamically allocated loops it is better to use
		560	C<ev_loop_new> and C<ev_loop_destroy>).
173		561
174	=item ev_loop_destroy (loop)	562	=item ev_loop_destroy (loop)
175		563
176	Like C<ev_default_destroy>, but destroys an event loop created by an	564	Like C<ev_default_destroy>, but destroys an event loop created by an
177	earlier call to C<ev_loop_new>.	565	earlier call to C<ev_loop_new>.
178		566
179	=item ev_default_fork ()	567	=item ev_default_fork ()
180		568
		569	This function sets a flag that causes subsequent C<ev_loop> iterations
181	This function reinitialises the kernel state for backends that have	570	to reinitialise the kernel state for backends that have one. Despite the
182	one. Despite the name, you can call it anytime, but it makes most sense	571	name, you can call it anytime, but it makes most sense after forking, in
183	after forking, in either the parent or child process (or both, but that	572	the child process (or both child and parent, but that again makes little
184	again makes little sense).	573	sense). You I<must> call it in the child before using any of the libev
		574	functions, and it will only take effect at the next C<ev_loop> iteration.
185		575
186	You I<must> call this function after forking if and only if you want to	576	On the other hand, you only need to call this function in the child
187	use the event library in both processes. If you just fork+exec, you don't	577	process if and only if you want to use the event library in the child. If
188	have to call it.	578	you just fork+exec, you don't have to call it at all.
189		579
190	The function itself is quite fast and it's usually not a problem to call	580	The function itself is quite fast and it's usually not a problem to call
191	it just in case after a fork. To make this easy, the function will fit in	581	it just in case after a fork. To make this easy, the function will fit in
192	quite nicely into a call to C<pthread_atfork>:	582	quite nicely into a call to C<pthread_atfork>:
193		583
…		…
195		585
196	=item ev_loop_fork (loop)	586	=item ev_loop_fork (loop)
197		587
198	Like C<ev_default_fork>, but acts on an event loop created by	588	Like C<ev_default_fork>, but acts on an event loop created by
199	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	589	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
200	after fork, and how you do this is entirely your own problem.	590	after fork that you want to re-use in the child, and how you do this is
		591	entirely your own problem.
201		592
		593	=item int ev_is_default_loop (loop)
		594
		595	Returns true when the given loop is, in fact, the default loop, and false
		596	otherwise.
		597
		598	=item unsigned int ev_loop_count (loop)
		599
		600	Returns the count of loop iterations for the loop, which is identical to
		601	the number of times libev did poll for new events. It starts at C<0> and
		602	happily wraps around with enough iterations.
		603
		604	This value can sometimes be useful as a generation counter of sorts (it
		605	"ticks" the number of loop iterations), as it roughly corresponds with
		606	C<ev_prepare> and C<ev_check> calls.
		607
202	=item unsigned int ev_method (loop)	608	=item unsigned int ev_backend (loop)
203		609
204	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	610	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
205	use.	611	use.
206		612
207	=item ev_tstamp ev_now (loop)	613	=item ev_tstamp ev_now (loop)
208		614
209	Returns the current "event loop time", which is the time the event loop	615	Returns the current "event loop time", which is the time the event loop
210	got events and started processing them. This timestamp does not change	616	received events and started processing them. This timestamp does not
211	as long as callbacks are being processed, and this is also the base time	617	change as long as callbacks are being processed, and this is also the base
212	used for relative timers. You can treat it as the timestamp of the event	618	time used for relative timers. You can treat it as the timestamp of the
213	occuring (or more correctly, the mainloop finding out about it).	619	event occurring (or more correctly, libev finding out about it).
		620
		621	=item ev_now_update (loop)
		622
		623	Establishes the current time by querying the kernel, updating the time
		624	returned by C<ev_now ()> in the progress. This is a costly operation and
		625	is usually done automatically within C<ev_loop ()>.
		626
		627	This function is rarely useful, but when some event callback runs for a
		628	very long time without entering the event loop, updating libev's idea of
		629	the current time is a good idea.
		630
		631	See also "The special problem of time updates" in the C<ev_timer> section.
214		632
215	=item ev_loop (loop, int flags)	633	=item ev_loop (loop, int flags)
216		634
217	Finally, this is it, the event handler. This function usually is called	635	Finally, this is it, the event handler. This function usually is called
218	after you initialised all your watchers and you want to start handling	636	after you initialised all your watchers and you want to start handling
219	events.	637	events.
220		638
221	If the flags argument is specified as 0, it will not return until either	639	If the flags argument is specified as C<0>, it will not return until
222	no event watchers are active anymore or C<ev_unloop> was called.	640	either no event watchers are active anymore or C<ev_unloop> was called.
		641
		642	Please note that an explicit C<ev_unloop> is usually better than
		643	relying on all watchers to be stopped when deciding when a program has
		644	finished (especially in interactive programs), but having a program
		645	that automatically loops as long as it has to and no longer by virtue
		646	of relying on its watchers stopping correctly, that is truly a thing of
		647	beauty.
223		648
224	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	649	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
225	those events and any outstanding ones, but will not block your process in	650	those events and any already outstanding ones, but will not block your
226	case there are no events and will return after one iteration of the loop.	651	process in case there are no events and will return after one iteration of
		652	the loop.
227		653
228	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	654	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
229	neccessary) and will handle those and any outstanding ones. It will block	655	necessary) and will handle those and any already outstanding ones. It
230	your process until at least one new event arrives, and will return after	656	will block your process until at least one new event arrives (which could
		657	be an event internal to libev itself, so there is no guarantee that a
		658	user-registered callback will be called), and will return after one
231	one iteration of the loop.	659	iteration of the loop.
232		660
233	This flags value could be used to implement alternative looping	661	This is useful if you are waiting for some external event in conjunction
234	constructs, but the C<prepare> and C<check> watchers provide a better and	662	with something not expressible using other libev watchers (i.e. "roll your
235	more generic mechanism.	663	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		664	usually a better approach for this kind of thing.
		665
		666	Here are the gory details of what C<ev_loop> does:
		667
		668	- Before the first iteration, call any pending watchers.
		669	* If EVFLAG_FORKCHECK was used, check for a fork.
		670	- If a fork was detected (by any means), queue and call all fork watchers.
		671	- Queue and call all prepare watchers.
		672	- If we have been forked, detach and recreate the kernel state
		673	as to not disturb the other process.
		674	- Update the kernel state with all outstanding changes.
		675	- Update the "event loop time" (ev_now ()).
		676	- Calculate for how long to sleep or block, if at all
		677	(active idle watchers, EVLOOP_NONBLOCK or not having
		678	any active watchers at all will result in not sleeping).
		679	- Sleep if the I/O and timer collect interval say so.
		680	- Block the process, waiting for any events.
		681	- Queue all outstanding I/O (fd) events.
		682	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
		683	- Queue all expired timers.
		684	- Queue all expired periodics.
		685	- Unless any events are pending now, queue all idle watchers.
		686	- Queue all check watchers.
		687	- Call all queued watchers in reverse order (i.e. check watchers first).
		688	Signals and child watchers are implemented as I/O watchers, and will
		689	be handled here by queueing them when their watcher gets executed.
		690	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
		691	were used, or there are no active watchers, return, otherwise
		692	continue with step *.
		693
		694	Example: Queue some jobs and then loop until no events are outstanding
		695	anymore.
		696
		697	... queue jobs here, make sure they register event watchers as long
		698	... as they still have work to do (even an idle watcher will do..)
		699	ev_loop (my_loop, 0);
		700	... jobs done or somebody called unloop. yeah!
236		701
237	=item ev_unloop (loop, how)	702	=item ev_unloop (loop, how)
238		703
239	Can be used to make a call to C<ev_loop> return early (but only after it	704	Can be used to make a call to C<ev_loop> return early (but only after it
240	has processed all outstanding events). The C<how> argument must be either	705	has processed all outstanding events). The C<how> argument must be either
241	C<EVUNLOOP_ONCE>, which will make the innermost C<ev_loop> call return, or	706	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
242	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	707	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
		708
		709	This "unloop state" will be cleared when entering C<ev_loop> again.
		710
		711	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
243		712
244	=item ev_ref (loop)	713	=item ev_ref (loop)
245		714
246	=item ev_unref (loop)	715	=item ev_unref (loop)
247		716
248	Ref/unref can be used to add or remove a reference count on the event	717	Ref/unref can be used to add or remove a reference count on the event
249	loop: Every watcher keeps one reference, and as long as the reference	718	loop: Every watcher keeps one reference, and as long as the reference
250	count is nonzero, C<ev_loop> will not return on its own. If you have	719	count is nonzero, C<ev_loop> will not return on its own.
		720
251	a watcher you never unregister that should not keep C<ev_loop> from	721	If you have a watcher you never unregister that should not keep C<ev_loop>
252	returning, ev_unref() after starting, and ev_ref() before stopping it. For	722	from returning, call ev_unref() after starting, and ev_ref() before
		723	stopping it.
		724
253	example, libev itself uses this for its internal signal pipe: It is not	725	As an example, libev itself uses this for its internal signal pipe: It is
254	visible to the libev user and should not keep C<ev_loop> from exiting if	726	not visible to the libev user and should not keep C<ev_loop> from exiting
255	no event watchers registered by it are active. It is also an excellent	727	if no event watchers registered by it are active. It is also an excellent
256	way to do this for generic recurring timers or from within third-party	728	way to do this for generic recurring timers or from within third-party
257	libraries. Just remember to I<unref after start> and I<ref before stop>.	729	libraries. Just remember to I<unref after start> and I<ref before stop>
		730	(but only if the watcher wasn't active before, or was active before,
		731	respectively).
		732
		733	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		734	running when nothing else is active.
		735
		736	ev_signal exitsig;
		737	ev_signal_init (&exitsig, sig_cb, SIGINT);
		738	ev_signal_start (loop, &exitsig);
		739	evf_unref (loop);
		740
		741	Example: For some weird reason, unregister the above signal handler again.
		742
		743	ev_ref (loop);
		744	ev_signal_stop (loop, &exitsig);
		745
		746	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		747
		748	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		749
		750	These advanced functions influence the time that libev will spend waiting
		751	for events. Both time intervals are by default C<0>, meaning that libev
		752	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		753	latency.
		754
		755	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		756	allows libev to delay invocation of I/O and timer/periodic callbacks
		757	to increase efficiency of loop iterations (or to increase power-saving
		758	opportunities).
		759
		760	The idea is that sometimes your program runs just fast enough to handle
		761	one (or very few) event(s) per loop iteration. While this makes the
		762	program responsive, it also wastes a lot of CPU time to poll for new
		763	events, especially with backends like C<select ()> which have a high
		764	overhead for the actual polling but can deliver many events at once.
		765
		766	By setting a higher I<io collect interval> you allow libev to spend more
		767	time collecting I/O events, so you can handle more events per iteration,
		768	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		769	C<ev_timer>) will be not affected. Setting this to a non-null value will
		770	introduce an additional C<ev_sleep ()> call into most loop iterations.
		771
		772	Likewise, by setting a higher I<timeout collect interval> you allow libev
		773	to spend more time collecting timeouts, at the expense of increased
		774	latency/jitter/inexactness (the watcher callback will be called
		775	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		776	value will not introduce any overhead in libev.
		777
		778	Many (busy) programs can usually benefit by setting the I/O collect
		779	interval to a value near C<0.1> or so, which is often enough for
		780	interactive servers (of course not for games), likewise for timeouts. It
		781	usually doesn't make much sense to set it to a lower value than C<0.01>,
		782	as this approaches the timing granularity of most systems.
		783
		784	Setting the I<timeout collect interval> can improve the opportunity for
		785	saving power, as the program will "bundle" timer callback invocations that
		786	are "near" in time together, by delaying some, thus reducing the number of
		787	times the process sleeps and wakes up again. Another useful technique to
		788	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		789	they fire on, say, one-second boundaries only.
		790
		791	=item ev_loop_verify (loop)
		792
		793	This function only does something when C<EV_VERIFY> support has been
		794	compiled in, which is the default for non-minimal builds. It tries to go
		795	through all internal structures and checks them for validity. If anything
		796	is found to be inconsistent, it will print an error message to standard
		797	error and call C<abort ()>.
		798
		799	This can be used to catch bugs inside libev itself: under normal
		800	circumstances, this function will never abort as of course libev keeps its
		801	data structures consistent.
258		802
259	=back	803	=back
260		804
		805
261	=head1 ANATOMY OF A WATCHER	806	=head1 ANATOMY OF A WATCHER
		807
		808	In the following description, uppercase C<TYPE> in names stands for the
		809	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		810	watchers and C<ev_io_start> for I/O watchers.
262		811
263	A watcher is a structure that you create and register to record your	812	A watcher is a structure that you create and register to record your
264	interest in some event. For instance, if you want to wait for STDIN to	813	interest in some event. For instance, if you want to wait for STDIN to
265	become readable, you would create an C<ev_io> watcher for that:	814	become readable, you would create an C<ev_io> watcher for that:
266		815
267	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	816	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
268	{	817	{
269	ev_io_stop (w);	818	ev_io_stop (w);
270	ev_unloop (loop, EVUNLOOP_ALL);	819	ev_unloop (loop, EVUNLOOP_ALL);
271	}	820	}
272		821
273	struct ev_loop *loop = ev_default_loop (0);	822	struct ev_loop *loop = ev_default_loop (0);
		823
274	struct ev_io stdin_watcher;	824	ev_io stdin_watcher;
		825
275	ev_init (&stdin_watcher, my_cb);	826	ev_init (&stdin_watcher, my_cb);
276	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	827	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
277	ev_io_start (loop, &stdin_watcher);	828	ev_io_start (loop, &stdin_watcher);
		829
278	ev_loop (loop, 0);	830	ev_loop (loop, 0);
279		831
280	As you can see, you are responsible for allocating the memory for your	832	As you can see, you are responsible for allocating the memory for your
281	watcher structures (and it is usually a bad idea to do this on the stack,	833	watcher structures (and it is I<usually> a bad idea to do this on the
282	although this can sometimes be quite valid).	834	stack).
		835
		836	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		837	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
283		838
284	Each watcher structure must be initialised by a call to C<ev_init	839	Each watcher structure must be initialised by a call to C<ev_init
285	(watcher *, callback)>, which expects a callback to be provided. This	840	(watcher *, callback)>, which expects a callback to be provided. This
286	callback gets invoked each time the event occurs (or, in the case of io	841	callback gets invoked each time the event occurs (or, in the case of I/O
287	watchers, each time the event loop detects that the file descriptor given	842	watchers, each time the event loop detects that the file descriptor given
288	is readable and/or writable).	843	is readable and/or writable).
289		844
290	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	845	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
291	with arguments specific to this watcher type. There is also a macro	846	macro to configure it, with arguments specific to the watcher type. There
292	to combine initialisation and setting in one call: C<< ev_<type>_init	847	is also a macro to combine initialisation and setting in one call: C<<
293	(watcher *, callback, ...) >>.	848	ev_TYPE_init (watcher *, callback, ...) >>.
294		849
295	To make the watcher actually watch out for events, you have to start it	850	To make the watcher actually watch out for events, you have to start it
296	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	851	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
297	*) >>), and you can stop watching for events at any time by calling the	852	*) >>), and you can stop watching for events at any time by calling the
298	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	853	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
299		854
300	As long as your watcher is active (has been started but not stopped) you	855	As long as your watcher is active (has been started but not stopped) you
301	must not touch the values stored in it. Most specifically you must never	856	must not touch the values stored in it. Most specifically you must never
302	reinitialise it or call its set method.	857	reinitialise it or call its C<ev_TYPE_set> macro.
303
304	You can check whether an event is active by calling the C<ev_is_active
305	(watcher *)> macro. To see whether an event is outstanding (but the
306	callback for it has not been called yet) you can use the C<ev_is_pending
307	(watcher *)> macro.
308		858
309	Each and every callback receives the event loop pointer as first, the	859	Each and every callback receives the event loop pointer as first, the
310	registered watcher structure as second, and a bitset of received events as	860	registered watcher structure as second, and a bitset of received events as
311	third argument.	861	third argument.
312		862
…		…
336	The signal specified in the C<ev_signal> watcher has been received by a thread.	886	The signal specified in the C<ev_signal> watcher has been received by a thread.
337		887
338	=item C<EV_CHILD>	888	=item C<EV_CHILD>
339		889
340	The pid specified in the C<ev_child> watcher has received a status change.	890	The pid specified in the C<ev_child> watcher has received a status change.
		891
		892	=item C<EV_STAT>
		893
		894	The path specified in the C<ev_stat> watcher changed its attributes somehow.
341		895
342	=item C<EV_IDLE>	896	=item C<EV_IDLE>
343		897
344	The C<ev_idle> watcher has determined that you have nothing better to do.	898	The C<ev_idle> watcher has determined that you have nothing better to do.
345		899
…		…
353	received events. Callbacks of both watcher types can start and stop as	907	received events. Callbacks of both watcher types can start and stop as
354	many watchers as they want, and all of them will be taken into account	908	many watchers as they want, and all of them will be taken into account
355	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	909	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
356	C<ev_loop> from blocking).	910	C<ev_loop> from blocking).
357		911
		912	=item C<EV_EMBED>
		913
		914	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		915
		916	=item C<EV_FORK>
		917
		918	The event loop has been resumed in the child process after fork (see
		919	C<ev_fork>).
		920
		921	=item C<EV_ASYNC>
		922
		923	The given async watcher has been asynchronously notified (see C<ev_async>).
		924
358	=item C<EV_ERROR>	925	=item C<EV_ERROR>
359		926
360	An unspecified error has occured, the watcher has been stopped. This might	927	An unspecified error has occurred, the watcher has been stopped. This might
361	happen because the watcher could not be properly started because libev	928	happen because the watcher could not be properly started because libev
362	ran out of memory, a file descriptor was found to be closed or any other	929	ran out of memory, a file descriptor was found to be closed or any other
		930	problem. Libev considers these application bugs.
		931
363	problem. You best act on it by reporting the problem and somehow coping	932	You best act on it by reporting the problem and somehow coping with the
364	with the watcher being stopped.	933	watcher being stopped. Note that well-written programs should not receive
		934	an error ever, so when your watcher receives it, this usually indicates a
		935	bug in your program.
365		936
366	Libev will usually signal a few "dummy" events together with an error,	937	Libev will usually signal a few "dummy" events together with an error, for
367	for example it might indicate that a fd is readable or writable, and if	938	example it might indicate that a fd is readable or writable, and if your
368	your callbacks is well-written it can just attempt the operation and cope	939	callbacks is well-written it can just attempt the operation and cope with
369	with the error from read() or write(). This will not work in multithreaded	940	the error from read() or write(). This will not work in multi-threaded
370	programs, though, so beware.	941	programs, though, as the fd could already be closed and reused for another
		942	thing, so beware.
371		943
372	=back	944	=back
373		945
		946	=head2 GENERIC WATCHER FUNCTIONS
		947
		948	=over 4
		949
		950	=item C<ev_init> (ev_TYPE *watcher, callback)
		951
		952	This macro initialises the generic portion of a watcher. The contents
		953	of the watcher object can be arbitrary (so C<malloc> will do). Only
		954	the generic parts of the watcher are initialised, you I<need> to call
		955	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		956	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		957	which rolls both calls into one.
		958
		959	You can reinitialise a watcher at any time as long as it has been stopped
		960	(or never started) and there are no pending events outstanding.
		961
		962	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		963	int revents)>.
		964
		965	Example: Initialise an C<ev_io> watcher in two steps.
		966
		967	ev_io w;
		968	ev_init (&w, my_cb);
		969	ev_io_set (&w, STDIN_FILENO, EV_READ);
		970
		971	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		972
		973	This macro initialises the type-specific parts of a watcher. You need to
		974	call C<ev_init> at least once before you call this macro, but you can
		975	call C<ev_TYPE_set> any number of times. You must not, however, call this
		976	macro on a watcher that is active (it can be pending, however, which is a
		977	difference to the C<ev_init> macro).
		978
		979	Although some watcher types do not have type-specific arguments
		980	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		981
		982	See C<ev_init>, above, for an example.
		983
		984	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		985
		986	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		987	calls into a single call. This is the most convenient method to initialise
		988	a watcher. The same limitations apply, of course.
		989
		990	Example: Initialise and set an C<ev_io> watcher in one step.
		991
		992	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		993
		994	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		995
		996	Starts (activates) the given watcher. Only active watchers will receive
		997	events. If the watcher is already active nothing will happen.
		998
		999	Example: Start the C<ev_io> watcher that is being abused as example in this
		1000	whole section.
		1001
		1002	ev_io_start (EV_DEFAULT_UC, &w);
		1003
		1004	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		1005
		1006	Stops the given watcher if active, and clears the pending status (whether
		1007	the watcher was active or not).
		1008
		1009	It is possible that stopped watchers are pending - for example,
		1010	non-repeating timers are being stopped when they become pending - but
		1011	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1012	pending. If you want to free or reuse the memory used by the watcher it is
		1013	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1014
		1015	=item bool ev_is_active (ev_TYPE *watcher)
		1016
		1017	Returns a true value iff the watcher is active (i.e. it has been started
		1018	and not yet been stopped). As long as a watcher is active you must not modify
		1019	it.
		1020
		1021	=item bool ev_is_pending (ev_TYPE *watcher)
		1022
		1023	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1024	events but its callback has not yet been invoked). As long as a watcher
		1025	is pending (but not active) you must not call an init function on it (but
		1026	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1027	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1028	it).
		1029
		1030	=item callback ev_cb (ev_TYPE *watcher)
		1031
		1032	Returns the callback currently set on the watcher.
		1033
		1034	=item ev_cb_set (ev_TYPE *watcher, callback)
		1035
		1036	Change the callback. You can change the callback at virtually any time
		1037	(modulo threads).
		1038
		1039	=item ev_set_priority (ev_TYPE *watcher, priority)
		1040
		1041	=item int ev_priority (ev_TYPE *watcher)
		1042
		1043	Set and query the priority of the watcher. The priority is a small
		1044	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1045	(default: C<-2>). Pending watchers with higher priority will be invoked
		1046	before watchers with lower priority, but priority will not keep watchers
		1047	from being executed (except for C<ev_idle> watchers).
		1048
		1049	This means that priorities are I<only> used for ordering callback
		1050	invocation after new events have been received. This is useful, for
		1051	example, to reduce latency after idling, or more often, to bind two
		1052	watchers on the same event and make sure one is called first.
		1053
		1054	If you need to suppress invocation when higher priority events are pending
		1055	you need to look at C<ev_idle> watchers, which provide this functionality.
		1056
		1057	You I<must not> change the priority of a watcher as long as it is active or
		1058	pending.
		1059
		1060	The default priority used by watchers when no priority has been set is
		1061	always C<0>, which is supposed to not be too high and not be too low :).
		1062
		1063	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1064	fine, as long as you do not mind that the priority value you query might
		1065	or might not have been clamped to the valid range.
		1066
		1067	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1068
		1069	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1070	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1071	can deal with that fact, as both are simply passed through to the
		1072	callback.
		1073
		1074	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1075
		1076	If the watcher is pending, this function clears its pending status and
		1077	returns its C<revents> bitset (as if its callback was invoked). If the
		1078	watcher isn't pending it does nothing and returns C<0>.
		1079
		1080	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1081	callback to be invoked, which can be accomplished with this function.
		1082
		1083	=back
		1084
		1085
374	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1086	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
375		1087
376	Each watcher has, by default, a member C<void *data> that you can change	1088	Each watcher has, by default, a member C<void *data> that you can change
377	and read at any time, libev will completely ignore it. This can be used	1089	and read at any time: libev will completely ignore it. This can be used
378	to associate arbitrary data with your watcher. If you need more data and	1090	to associate arbitrary data with your watcher. If you need more data and
379	don't want to allocate memory and store a pointer to it in that data	1091	don't want to allocate memory and store a pointer to it in that data
380	member, you can also "subclass" the watcher type and provide your own	1092	member, you can also "subclass" the watcher type and provide your own
381	data:	1093	data:
382		1094
383	struct my_io	1095	struct my_io
384	{	1096	{
385	struct ev_io io;	1097	ev_io io;
386	int otherfd;	1098	int otherfd;
387	void *somedata;	1099	void *somedata;
388	struct whatever *mostinteresting;	1100	struct whatever *mostinteresting;
389	}	1101	};
		1102
		1103	...
		1104	struct my_io w;
		1105	ev_io_init (&w.io, my_cb, fd, EV_READ);
390		1106
391	And since your callback will be called with a pointer to the watcher, you	1107	And since your callback will be called with a pointer to the watcher, you
392	can cast it back to your own type:	1108	can cast it back to your own type:
393		1109
394	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1110	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
395	{	1111	{
396	struct my_io *w = (struct my_io *)w_;	1112	struct my_io *w = (struct my_io *)w_;
397	...	1113	...
398	}	1114	}
399		1115
400	More interesting and less C-conformant ways of catsing your callback type	1116	More interesting and less C-conformant ways of casting your callback type
401	have been omitted....	1117	instead have been omitted.
		1118
		1119	Another common scenario is to use some data structure with multiple
		1120	embedded watchers:
		1121
		1122	struct my_biggy
		1123	{
		1124	int some_data;
		1125	ev_timer t1;
		1126	ev_timer t2;
		1127	}
		1128
		1129	In this case getting the pointer to C<my_biggy> is a bit more
		1130	complicated: Either you store the address of your C<my_biggy> struct
		1131	in the C<data> member of the watcher (for woozies), or you need to use
		1132	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1133	programmers):
		1134
		1135	#include <stddef.h>
		1136
		1137	static void
		1138	t1_cb (EV_P_ ev_timer *w, int revents)
		1139	{
		1140	struct my_biggy big = (struct my_biggy *
		1141	(((char *)w) - offsetof (struct my_biggy, t1));
		1142	}
		1143
		1144	static void
		1145	t2_cb (EV_P_ ev_timer *w, int revents)
		1146	{
		1147	struct my_biggy big = (struct my_biggy *
		1148	(((char *)w) - offsetof (struct my_biggy, t2));
		1149	}
402		1150
403		1151
404	=head1 WATCHER TYPES	1152	=head1 WATCHER TYPES
405		1153
406	This section describes each watcher in detail, but will not repeat	1154	This section describes each watcher in detail, but will not repeat
407	information given in the last section.	1155	information given in the last section. Any initialisation/set macros,
		1156	functions and members specific to the watcher type are explained.
408		1157
		1158	Members are additionally marked with either I<[read-only]>, meaning that,
		1159	while the watcher is active, you can look at the member and expect some
		1160	sensible content, but you must not modify it (you can modify it while the
		1161	watcher is stopped to your hearts content), or I<[read-write]>, which
		1162	means you can expect it to have some sensible content while the watcher
		1163	is active, but you can also modify it. Modifying it may not do something
		1164	sensible or take immediate effect (or do anything at all), but libev will
		1165	not crash or malfunction in any way.
		1166
		1167
409	=head2 C<ev_io> - is this file descriptor readable or writable	1168	=head2 C<ev_io> - is this file descriptor readable or writable?
410		1169
411	I/O watchers check whether a file descriptor is readable or writable	1170	I/O watchers check whether a file descriptor is readable or writable
412	in each iteration of the event loop (This behaviour is called	1171	in each iteration of the event loop, or, more precisely, when reading
413	level-triggering because you keep receiving events as long as the	1172	would not block the process and writing would at least be able to write
414	condition persists. Remember you can stop the watcher if you don't want to	1173	some data. This behaviour is called level-triggering because you keep
415	act on the event and neither want to receive future events).	1174	receiving events as long as the condition persists. Remember you can stop
		1175	the watcher if you don't want to act on the event and neither want to
		1176	receive future events.
416		1177
417	In general you can register as many read and/or write event watchers oer	1178	In general you can register as many read and/or write event watchers per
418	fd as you want (as long as you don't confuse yourself). Setting all file	1179	fd as you want (as long as you don't confuse yourself). Setting all file
419	descriptors to non-blocking mode is also usually a good idea (but not	1180	descriptors to non-blocking mode is also usually a good idea (but not
420	required if you know what you are doing).	1181	required if you know what you are doing).
421		1182
422	You have to be careful with dup'ed file descriptors, though. Some backends	1183	If you cannot use non-blocking mode, then force the use of a
423	(the linux epoll backend is a notable example) cannot handle dup'ed file	1184	known-to-be-good backend (at the time of this writing, this includes only
424	descriptors correctly if you register interest in two or more fds pointing	1185	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
425	to the same file/socket etc. description.
426		1186
427	If you must do this, then force the use of a known-to-be-good backend	1187	Another thing you have to watch out for is that it is quite easy to
428	(at the time of this writing, this includes only EVMETHOD_SELECT and	1188	receive "spurious" readiness notifications, that is your callback might
429	EVMETHOD_POLL).	1189	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1190	because there is no data. Not only are some backends known to create a
		1191	lot of those (for example Solaris ports), it is very easy to get into
		1192	this situation even with a relatively standard program structure. Thus
		1193	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1194	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1195
		1196	If you cannot run the fd in non-blocking mode (for example you should
		1197	not play around with an Xlib connection), then you have to separately
		1198	re-test whether a file descriptor is really ready with a known-to-be good
		1199	interface such as poll (fortunately in our Xlib example, Xlib already
		1200	does this on its own, so its quite safe to use). Some people additionally
		1201	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1202	indefinitely.
		1203
		1204	But really, best use non-blocking mode.
		1205
		1206	=head3 The special problem of disappearing file descriptors
		1207
		1208	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1209	descriptor (either due to calling C<close> explicitly or any other means,
		1210	such as C<dup2>). The reason is that you register interest in some file
		1211	descriptor, but when it goes away, the operating system will silently drop
		1212	this interest. If another file descriptor with the same number then is
		1213	registered with libev, there is no efficient way to see that this is, in
		1214	fact, a different file descriptor.
		1215
		1216	To avoid having to explicitly tell libev about such cases, libev follows
		1217	the following policy: Each time C<ev_io_set> is being called, libev
		1218	will assume that this is potentially a new file descriptor, otherwise
		1219	it is assumed that the file descriptor stays the same. That means that
		1220	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1221	descriptor even if the file descriptor number itself did not change.
		1222
		1223	This is how one would do it normally anyway, the important point is that
		1224	the libev application should not optimise around libev but should leave
		1225	optimisations to libev.
		1226
		1227	=head3 The special problem of dup'ed file descriptors
		1228
		1229	Some backends (e.g. epoll), cannot register events for file descriptors,
		1230	but only events for the underlying file descriptions. That means when you
		1231	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1232	events for them, only one file descriptor might actually receive events.
		1233
		1234	There is no workaround possible except not registering events
		1235	for potentially C<dup ()>'ed file descriptors, or to resort to
		1236	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1237
		1238	=head3 The special problem of fork
		1239
		1240	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1241	useless behaviour. Libev fully supports fork, but needs to be told about
		1242	it in the child.
		1243
		1244	To support fork in your programs, you either have to call
		1245	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1246	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1247	C<EVBACKEND_POLL>.
		1248
		1249	=head3 The special problem of SIGPIPE
		1250
		1251	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1252	when writing to a pipe whose other end has been closed, your program gets
		1253	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1254	this is sensible behaviour, for daemons, this is usually undesirable.
		1255
		1256	So when you encounter spurious, unexplained daemon exits, make sure you
		1257	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1258	somewhere, as that would have given you a big clue).
		1259
		1260
		1261	=head3 Watcher-Specific Functions
430		1262
431	=over 4	1263	=over 4
432		1264
433	=item ev_io_init (ev_io *, callback, int fd, int events)	1265	=item ev_io_init (ev_io *, callback, int fd, int events)
434		1266
435	=item ev_io_set (ev_io *, int fd, int events)	1267	=item ev_io_set (ev_io *, int fd, int events)
436		1268
437	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1269	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
438	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1270	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
439	EV_WRITE> to receive the given events.	1271	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1272
		1273	=item int fd [read-only]
		1274
		1275	The file descriptor being watched.
		1276
		1277	=item int events [read-only]
		1278
		1279	The events being watched.
440		1280
441	=back	1281	=back
442		1282
		1283	=head3 Examples
		1284
		1285	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1286	readable, but only once. Since it is likely line-buffered, you could
		1287	attempt to read a whole line in the callback.
		1288
		1289	static void
		1290	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1291	{
		1292	ev_io_stop (loop, w);
		1293	.. read from stdin here (or from w->fd) and handle any I/O errors
		1294	}
		1295
		1296	...
		1297	struct ev_loop *loop = ev_default_init (0);
		1298	ev_io stdin_readable;
		1299	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1300	ev_io_start (loop, &stdin_readable);
		1301	ev_loop (loop, 0);
		1302
		1303
443	=head2 C<ev_timer> - relative and optionally recurring timeouts	1304	=head2 C<ev_timer> - relative and optionally repeating timeouts
444		1305
445	Timer watchers are simple relative timers that generate an event after a	1306	Timer watchers are simple relative timers that generate an event after a
446	given time, and optionally repeating in regular intervals after that.	1307	given time, and optionally repeating in regular intervals after that.
447		1308
448	The timers are based on real time, that is, if you register an event that	1309	The timers are based on real time, that is, if you register an event that
449	times out after an hour and youreset your system clock to last years	1310	times out after an hour and you reset your system clock to January last
450	time, it will still time out after (roughly) and hour. "Roughly" because	1311	year, it will still time out after (roughly) one hour. "Roughly" because
451	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1312	detecting time jumps is hard, and some inaccuracies are unavoidable (the
452	monotonic clock option helps a lot here).	1313	monotonic clock option helps a lot here).
		1314
		1315	The callback is guaranteed to be invoked only I<after> its timeout has
		1316	passed, but if multiple timers become ready during the same loop iteration
		1317	then order of execution is undefined.
		1318
		1319	=head3 Be smart about timeouts
		1320
		1321	Many real-world problems involve some kind of timeout, usually for error
		1322	recovery. A typical example is an HTTP request - if the other side hangs,
		1323	you want to raise some error after a while.
		1324
		1325	What follows are some ways to handle this problem, from obvious and
		1326	inefficient to smart and efficient.
		1327
		1328	In the following, a 60 second activity timeout is assumed - a timeout that
		1329	gets reset to 60 seconds each time there is activity (e.g. each time some
		1330	data or other life sign was received).
		1331
		1332	=over 4
		1333
		1334	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1335
		1336	This is the most obvious, but not the most simple way: In the beginning,
		1337	start the watcher:
		1338
		1339	ev_timer_init (timer, callback, 60., 0.);
		1340	ev_timer_start (loop, timer);
		1341
		1342	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1343	and start it again:
		1344
		1345	ev_timer_stop (loop, timer);
		1346	ev_timer_set (timer, 60., 0.);
		1347	ev_timer_start (loop, timer);
		1348
		1349	This is relatively simple to implement, but means that each time there is
		1350	some activity, libev will first have to remove the timer from its internal
		1351	data structure and then add it again. Libev tries to be fast, but it's
		1352	still not a constant-time operation.
		1353
		1354	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1355
		1356	This is the easiest way, and involves using C<ev_timer_again> instead of
		1357	C<ev_timer_start>.
		1358
		1359	To implement this, configure an C<ev_timer> with a C<repeat> value
		1360	of C<60> and then call C<ev_timer_again> at start and each time you
		1361	successfully read or write some data. If you go into an idle state where
		1362	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1363	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1364
		1365	That means you can ignore both the C<ev_timer_start> function and the
		1366	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1367	member and C<ev_timer_again>.
		1368
		1369	At start:
		1370
		1371	ev_timer_init (timer, callback);
		1372	timer->repeat = 60.;
		1373	ev_timer_again (loop, timer);
		1374
		1375	Each time there is some activity:
		1376
		1377	ev_timer_again (loop, timer);
		1378
		1379	It is even possible to change the time-out on the fly, regardless of
		1380	whether the watcher is active or not:
		1381
		1382	timer->repeat = 30.;
		1383	ev_timer_again (loop, timer);
		1384
		1385	This is slightly more efficient then stopping/starting the timer each time
		1386	you want to modify its timeout value, as libev does not have to completely
		1387	remove and re-insert the timer from/into its internal data structure.
		1388
		1389	It is, however, even simpler than the "obvious" way to do it.
		1390
		1391	=item 3. Let the timer time out, but then re-arm it as required.
		1392
		1393	This method is more tricky, but usually most efficient: Most timeouts are
		1394	relatively long compared to the intervals between other activity - in
		1395	our example, within 60 seconds, there are usually many I/O events with
		1396	associated activity resets.
		1397
		1398	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1399	but remember the time of last activity, and check for a real timeout only
		1400	within the callback:
		1401
		1402	ev_tstamp last_activity; // time of last activity
		1403
		1404	static void
		1405	callback (EV_P_ ev_timer *w, int revents)
		1406	{
		1407	ev_tstamp now = ev_now (EV_A);
		1408	ev_tstamp timeout = last_activity + 60.;
		1409
		1410	// if last_activity + 60. is older than now, we did time out
		1411	if (timeout < now)
		1412	{
		1413	// timeout occured, take action
		1414	}
		1415	else
		1416	{
		1417	// callback was invoked, but there was some activity, re-arm
		1418	// the watcher to fire in last_activity + 60, which is
		1419	// guaranteed to be in the future, so "again" is positive:
		1420	w->again = timeout - now;
		1421	ev_timer_again (EV_A_ w);
		1422	}
		1423	}
		1424
		1425	To summarise the callback: first calculate the real timeout (defined
		1426	as "60 seconds after the last activity"), then check if that time has
		1427	been reached, which means something I<did>, in fact, time out. Otherwise
		1428	the callback was invoked too early (C<timeout> is in the future), so
		1429	re-schedule the timer to fire at that future time, to see if maybe we have
		1430	a timeout then.
		1431
		1432	Note how C<ev_timer_again> is used, taking advantage of the
		1433	C<ev_timer_again> optimisation when the timer is already running.
		1434
		1435	This scheme causes more callback invocations (about one every 60 seconds
		1436	minus half the average time between activity), but virtually no calls to
		1437	libev to change the timeout.
		1438
		1439	To start the timer, simply initialise the watcher and set C<last_activity>
		1440	to the current time (meaning we just have some activity :), then call the
		1441	callback, which will "do the right thing" and start the timer:
		1442
		1443	ev_timer_init (timer, callback);
		1444	last_activity = ev_now (loop);
		1445	callback (loop, timer, EV_TIMEOUT);
		1446
		1447	And when there is some activity, simply store the current time in
		1448	C<last_activity>, no libev calls at all:
		1449
		1450	last_actiivty = ev_now (loop);
		1451
		1452	This technique is slightly more complex, but in most cases where the
		1453	time-out is unlikely to be triggered, much more efficient.
		1454
		1455	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1456	callback :) - just change the timeout and invoke the callback, which will
		1457	fix things for you.
		1458
		1459	=item 4. Wee, just use a double-linked list for your timeouts.
		1460
		1461	If there is not one request, but many thousands (millions...), all
		1462	employing some kind of timeout with the same timeout value, then one can
		1463	do even better:
		1464
		1465	When starting the timeout, calculate the timeout value and put the timeout
		1466	at the I<end> of the list.
		1467
		1468	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1469	the list is expected to fire (for example, using the technique #3).
		1470
		1471	When there is some activity, remove the timer from the list, recalculate
		1472	the timeout, append it to the end of the list again, and make sure to
		1473	update the C<ev_timer> if it was taken from the beginning of the list.
		1474
		1475	This way, one can manage an unlimited number of timeouts in O(1) time for
		1476	starting, stopping and updating the timers, at the expense of a major
		1477	complication, and having to use a constant timeout. The constant timeout
		1478	ensures that the list stays sorted.
		1479
		1480	=back
		1481
		1482	So which method the best?
		1483
		1484	Method #2 is a simple no-brain-required solution that is adequate in most
		1485	situations. Method #3 requires a bit more thinking, but handles many cases
		1486	better, and isn't very complicated either. In most case, choosing either
		1487	one is fine, with #3 being better in typical situations.
		1488
		1489	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1490	rather complicated, but extremely efficient, something that really pays
		1491	off after the first million or so of active timers, i.e. it's usually
		1492	overkill :)
		1493
		1494	=head3 The special problem of time updates
		1495
		1496	Establishing the current time is a costly operation (it usually takes at
		1497	least two system calls): EV therefore updates its idea of the current
		1498	time only before and after C<ev_loop> collects new events, which causes a
		1499	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1500	lots of events in one iteration.
453		1501
454	The relative timeouts are calculated relative to the C<ev_now ()>	1502	The relative timeouts are calculated relative to the C<ev_now ()>
455	time. This is usually the right thing as this timestamp refers to the time	1503	time. This is usually the right thing as this timestamp refers to the time
456	of the event triggering whatever timeout you are modifying/starting. If	1504	of the event triggering whatever timeout you are modifying/starting. If
457	you suspect event processing to be delayed and you *need* to base the timeout	1505	you suspect event processing to be delayed and you I<need> to base the
458	ion the current time, use something like this to adjust for this:	1506	timeout on the current time, use something like this to adjust for this:
459		1507
460	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1508	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
461		1509
		1510	If the event loop is suspended for a long time, you can also force an
		1511	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1512	()>.
		1513
		1514	=head3 Watcher-Specific Functions and Data Members
		1515
462	=over 4	1516	=over 4
463		1517
464	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1518	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
465		1519
466	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1520	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
467		1521
468	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1522	Configure the timer to trigger after C<after> seconds. If C<repeat>
469	C<0.>, then it will automatically be stopped. If it is positive, then the	1523	is C<0.>, then it will automatically be stopped once the timeout is
470	timer will automatically be configured to trigger again C<repeat> seconds	1524	reached. If it is positive, then the timer will automatically be
471	later, again, and again, until stopped manually.	1525	configured to trigger again C<repeat> seconds later, again, and again,
		1526	until stopped manually.
472		1527
473	The timer itself will do a best-effort at avoiding drift, that is, if you	1528	The timer itself will do a best-effort at avoiding drift, that is, if
474	configure a timer to trigger every 10 seconds, then it will trigger at	1529	you configure a timer to trigger every 10 seconds, then it will normally
475	exactly 10 second intervals. If, however, your program cannot keep up with	1530	trigger at exactly 10 second intervals. If, however, your program cannot
476	the timer (ecause it takes longer than those 10 seconds to do stuff) the	1531	keep up with the timer (because it takes longer than those 10 seconds to
477	timer will not fire more than once per event loop iteration.	1532	do stuff) the timer will not fire more than once per event loop iteration.
478		1533
479	=item ev_timer_again (loop)	1534	=item ev_timer_again (loop, ev_timer *)
480		1535
481	This will act as if the timer timed out and restart it again if it is	1536	This will act as if the timer timed out and restart it again if it is
482	repeating. The exact semantics are:	1537	repeating. The exact semantics are:
483		1538
		1539	If the timer is pending, its pending status is cleared.
		1540
484	If the timer is started but nonrepeating, stop it.	1541	If the timer is started but non-repeating, stop it (as if it timed out).
485		1542
486	If the timer is repeating, either start it if necessary (with the repeat	1543	If the timer is repeating, either start it if necessary (with the
487	value), or reset the running timer to the repeat value.	1544	C<repeat> value), or reset the running timer to the C<repeat> value.
488		1545
489	This sounds a bit complicated, but here is a useful and typical	1546	This sounds a bit complicated, see "Be smart about timeouts", above, for a
490	example: Imagine you have a tcp connection and you want a so-called idle	1547	usage example.
491	timeout, that is, you want to be called when there have been, say, 60	1548
492	seconds of inactivity on the socket. The easiest way to do this is to	1549	=item ev_tstamp repeat [read-write]
493	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1550
494	time you successfully read or write some data. If you go into an idle	1551	The current C<repeat> value. Will be used each time the watcher times out
495	state where you do not expect data to travel on the socket, you can stop	1552	or C<ev_timer_again> is called, and determines the next timeout (if any),
496	the timer, and again will automatically restart it if need be.	1553	which is also when any modifications are taken into account.
497		1554
498	=back	1555	=back
499		1556
		1557	=head3 Examples
		1558
		1559	Example: Create a timer that fires after 60 seconds.
		1560
		1561	static void
		1562	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1563	{
		1564	.. one minute over, w is actually stopped right here
		1565	}
		1566
		1567	ev_timer mytimer;
		1568	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1569	ev_timer_start (loop, &mytimer);
		1570
		1571	Example: Create a timeout timer that times out after 10 seconds of
		1572	inactivity.
		1573
		1574	static void
		1575	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1576	{
		1577	.. ten seconds without any activity
		1578	}
		1579
		1580	ev_timer mytimer;
		1581	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1582	ev_timer_again (&mytimer); /* start timer */
		1583	ev_loop (loop, 0);
		1584
		1585	// and in some piece of code that gets executed on any "activity":
		1586	// reset the timeout to start ticking again at 10 seconds
		1587	ev_timer_again (&mytimer);
		1588
		1589
500	=head2 C<ev_periodic> - to cron or not to cron	1590	=head2 C<ev_periodic> - to cron or not to cron?
501		1591
502	Periodic watchers are also timers of a kind, but they are very versatile	1592	Periodic watchers are also timers of a kind, but they are very versatile
503	(and unfortunately a bit complex).	1593	(and unfortunately a bit complex).
504		1594
505	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1595	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
506	but on wallclock time (absolute time). You can tell a periodic watcher	1596	but on wall clock time (absolute time). You can tell a periodic watcher
507	to trigger "at" some specific point in time. For example, if you tell a	1597	to trigger after some specific point in time. For example, if you tell a
508	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1598	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()
509	+ 10.>) and then reset your system clock to the last year, then it will	1599	+ 10.>, that is, an absolute time not a delay) and then reset your system
		1600	clock to January of the previous year, then it will take more than year
510	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1601	to trigger the event (unlike an C<ev_timer>, which would still trigger
511	roughly 10 seconds later and of course not if you reset your system time	1602	roughly 10 seconds later as it uses a relative timeout).
512	again).
513		1603
514	They can also be used to implement vastly more complex timers, such as	1604	C<ev_periodic>s can also be used to implement vastly more complex timers,
515	triggering an event on eahc midnight, local time.	1605	such as triggering an event on each "midnight, local time", or other
		1606	complicated rules.
		1607
		1608	As with timers, the callback is guaranteed to be invoked only when the
		1609	time (C<at>) has passed, but if multiple periodic timers become ready
		1610	during the same loop iteration, then order of execution is undefined.
		1611
		1612	=head3 Watcher-Specific Functions and Data Members
516		1613
517	=over 4	1614	=over 4
518		1615
519	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1616	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
520		1617
521	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1618	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
522		1619
523	Lots of arguments, lets sort it out... There are basically three modes of	1620	Lots of arguments, lets sort it out... There are basically three modes of
524	operation, and we will explain them from simplest to complex:	1621	operation, and we will explain them from simplest to most complex:
525
526		1622
527	=over 4	1623	=over 4
528		1624
529	=item * absolute timer (interval = reschedule_cb = 0)	1625	=item * absolute timer (at = time, interval = reschedule_cb = 0)
530		1626
531	In this configuration the watcher triggers an event at the wallclock time	1627	In this configuration the watcher triggers an event after the wall clock
532	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1628	time C<at> has passed. It will not repeat and will not adjust when a time
533	that is, if it is to be run at January 1st 2011 then it will run when the	1629	jump occurs, that is, if it is to be run at January 1st 2011 then it will
534	system time reaches or surpasses this time.	1630	only run when the system clock reaches or surpasses this time.
535		1631
536	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1632	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
537		1633
538	In this mode the watcher will always be scheduled to time out at the next	1634	In this mode the watcher will always be scheduled to time out at the next
539	C<at + N * interval> time (for some integer N) and then repeat, regardless	1635	C<at + N * interval> time (for some integer N, which can also be negative)
540	of any time jumps.	1636	and then repeat, regardless of any time jumps.
541		1637
542	This can be used to create timers that do not drift with respect to system	1638	This can be used to create timers that do not drift with respect to the
543	time:	1639	system clock, for example, here is a C<ev_periodic> that triggers each
		1640	hour, on the hour:
544		1641
545	ev_periodic_set (&periodic, 0., 3600., 0);	1642	ev_periodic_set (&periodic, 0., 3600., 0);
546		1643
547	This doesn't mean there will always be 3600 seconds in between triggers,	1644	This doesn't mean there will always be 3600 seconds in between triggers,
548	but only that the the callback will be called when the system time shows a	1645	but only that the callback will be called when the system time shows a
549	full hour (UTC), or more correctly, when the system time is evenly divisible	1646	full hour (UTC), or more correctly, when the system time is evenly divisible
550	by 3600.	1647	by 3600.
551		1648
552	Another way to think about it (for the mathematically inclined) is that	1649	Another way to think about it (for the mathematically inclined) is that
553	C<ev_periodic> will try to run the callback in this mode at the next possible	1650	C<ev_periodic> will try to run the callback in this mode at the next possible
554	time where C<time = at (mod interval)>, regardless of any time jumps.	1651	time where C<time = at (mod interval)>, regardless of any time jumps.
555		1652
		1653	For numerical stability it is preferable that the C<at> value is near
		1654	C<ev_now ()> (the current time), but there is no range requirement for
		1655	this value, and in fact is often specified as zero.
		1656
		1657	Note also that there is an upper limit to how often a timer can fire (CPU
		1658	speed for example), so if C<interval> is very small then timing stability
		1659	will of course deteriorate. Libev itself tries to be exact to be about one
		1660	millisecond (if the OS supports it and the machine is fast enough).
		1661
556	=item * manual reschedule mode (reschedule_cb = callback)	1662	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
557		1663
558	In this mode the values for C<interval> and C<at> are both being	1664	In this mode the values for C<interval> and C<at> are both being
559	ignored. Instead, each time the periodic watcher gets scheduled, the	1665	ignored. Instead, each time the periodic watcher gets scheduled, the
560	reschedule callback will be called with the watcher as first, and the	1666	reschedule callback will be called with the watcher as first, and the
561	current time as second argument.	1667	current time as second argument.
562		1668
563	NOTE: I<This callback MUST NOT stop or destroy the periodic or any other	1669	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
564	periodic watcher, ever, or make any event loop modifications>. If you need	1670	ever, or make ANY event loop modifications whatsoever>.
		1671
565	to stop it, return C<now + 1e30> (or so, fudge fudge) and stop it afterwards.	1672	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1673	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1674	only event loop modification you are allowed to do).
566		1675
567	Also, I<< this callback must always return a time that is later than the
568	passed C<now> value >>. Not even C<now> itself will be ok.
569
570	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1676	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
571	ev_tstamp now)>, e.g.:	1677	*w, ev_tstamp now)>, e.g.:
572		1678
		1679	static ev_tstamp
573	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1680	my_rescheduler (ev_periodic *w, ev_tstamp now)
574	{	1681	{
575	return now + 60.;	1682	return now + 60.;
576	}	1683	}
577		1684
578	It must return the next time to trigger, based on the passed time value	1685	It must return the next time to trigger, based on the passed time value
579	(that is, the lowest time value larger than to the second argument). It	1686	(that is, the lowest time value larger than to the second argument). It
580	will usually be called just before the callback will be triggered, but	1687	will usually be called just before the callback will be triggered, but
581	might be called at other times, too.	1688	might be called at other times, too.
582		1689
		1690	NOTE: I<< This callback must always return a time that is higher than or
		1691	equal to the passed C<now> value >>.
		1692
583	This can be used to create very complex timers, such as a timer that	1693	This can be used to create very complex timers, such as a timer that
584	triggers on each midnight, local time. To do this, you would calculate the	1694	triggers on "next midnight, local time". To do this, you would calculate the
585	next midnight after C<now> and return the timestamp value for this. How you do this	1695	next midnight after C<now> and return the timestamp value for this. How
586	is, again, up to you (but it is not trivial).	1696	you do this is, again, up to you (but it is not trivial, which is the main
		1697	reason I omitted it as an example).
587		1698
588	=back	1699	=back
589		1700
590	=item ev_periodic_again (loop, ev_periodic *)	1701	=item ev_periodic_again (loop, ev_periodic *)
591		1702
592	Simply stops and restarts the periodic watcher again. This is only useful	1703	Simply stops and restarts the periodic watcher again. This is only useful
593	when you changed some parameters or the reschedule callback would return	1704	when you changed some parameters or the reschedule callback would return
594	a different time than the last time it was called (e.g. in a crond like	1705	a different time than the last time it was called (e.g. in a crond like
595	program when the crontabs have changed).	1706	program when the crontabs have changed).
596		1707
		1708	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1709
		1710	When active, returns the absolute time that the watcher is supposed to
		1711	trigger next.
		1712
		1713	=item ev_tstamp offset [read-write]
		1714
		1715	When repeating, this contains the offset value, otherwise this is the
		1716	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
		1717
		1718	Can be modified any time, but changes only take effect when the periodic
		1719	timer fires or C<ev_periodic_again> is being called.
		1720
		1721	=item ev_tstamp interval [read-write]
		1722
		1723	The current interval value. Can be modified any time, but changes only
		1724	take effect when the periodic timer fires or C<ev_periodic_again> is being
		1725	called.
		1726
		1727	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		1728
		1729	The current reschedule callback, or C<0>, if this functionality is
		1730	switched off. Can be changed any time, but changes only take effect when
		1731	the periodic timer fires or C<ev_periodic_again> is being called.
		1732
597	=back	1733	=back
598		1734
		1735	=head3 Examples
		1736
		1737	Example: Call a callback every hour, or, more precisely, whenever the
		1738	system time is divisible by 3600. The callback invocation times have
		1739	potentially a lot of jitter, but good long-term stability.
		1740
		1741	static void
		1742	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		1743	{
		1744	... its now a full hour (UTC, or TAI or whatever your clock follows)
		1745	}
		1746
		1747	ev_periodic hourly_tick;
		1748	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		1749	ev_periodic_start (loop, &hourly_tick);
		1750
		1751	Example: The same as above, but use a reschedule callback to do it:
		1752
		1753	#include <math.h>
		1754
		1755	static ev_tstamp
		1756	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		1757	{
		1758	return now + (3600. - fmod (now, 3600.));
		1759	}
		1760
		1761	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		1762
		1763	Example: Call a callback every hour, starting now:
		1764
		1765	ev_periodic hourly_tick;
		1766	ev_periodic_init (&hourly_tick, clock_cb,
		1767	fmod (ev_now (loop), 3600.), 3600., 0);
		1768	ev_periodic_start (loop, &hourly_tick);
		1769
		1770
599	=head2 C<ev_signal> - signal me when a signal gets signalled	1771	=head2 C<ev_signal> - signal me when a signal gets signalled!
600		1772
601	Signal watchers will trigger an event when the process receives a specific	1773	Signal watchers will trigger an event when the process receives a specific
602	signal one or more times. Even though signals are very asynchronous, libev	1774	signal one or more times. Even though signals are very asynchronous, libev
603	will try it's best to deliver signals synchronously, i.e. as part of the	1775	will try it's best to deliver signals synchronously, i.e. as part of the
604	normal event processing, like any other event.	1776	normal event processing, like any other event.
605		1777
		1778	If you want signals asynchronously, just use C<sigaction> as you would
		1779	do without libev and forget about sharing the signal. You can even use
		1780	C<ev_async> from a signal handler to synchronously wake up an event loop.
		1781
606	You can configure as many watchers as you like per signal. Only when the	1782	You can configure as many watchers as you like per signal. Only when the
607	first watcher gets started will libev actually register a signal watcher	1783	first watcher gets started will libev actually register a signal handler
608	with the kernel (thus it coexists with your own signal handlers as long	1784	with the kernel (thus it coexists with your own signal handlers as long as
609	as you don't register any with libev). Similarly, when the last signal	1785	you don't register any with libev for the same signal). Similarly, when
610	watcher for a signal is stopped libev will reset the signal handler to	1786	the last signal watcher for a signal is stopped, libev will reset the
611	SIG_DFL (regardless of what it was set to before).	1787	signal handler to SIG_DFL (regardless of what it was set to before).
		1788
		1789	If possible and supported, libev will install its handlers with
		1790	C<SA_RESTART> behaviour enabled, so system calls should not be unduly
		1791	interrupted. If you have a problem with system calls getting interrupted by
		1792	signals you can block all signals in an C<ev_check> watcher and unblock
		1793	them in an C<ev_prepare> watcher.
		1794
		1795	=head3 Watcher-Specific Functions and Data Members
612		1796
613	=over 4	1797	=over 4
614		1798
615	=item ev_signal_init (ev_signal *, callback, int signum)	1799	=item ev_signal_init (ev_signal *, callback, int signum)
616		1800
617	=item ev_signal_set (ev_signal *, int signum)	1801	=item ev_signal_set (ev_signal *, int signum)
618		1802
619	Configures the watcher to trigger on the given signal number (usually one	1803	Configures the watcher to trigger on the given signal number (usually one
620	of the C<SIGxxx> constants).	1804	of the C<SIGxxx> constants).
621		1805
		1806	=item int signum [read-only]
		1807
		1808	The signal the watcher watches out for.
		1809
622	=back	1810	=back
623		1811
		1812	=head3 Examples
		1813
		1814	Example: Try to exit cleanly on SIGINT.
		1815
		1816	static void
		1817	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		1818	{
		1819	ev_unloop (loop, EVUNLOOP_ALL);
		1820	}
		1821
		1822	ev_signal signal_watcher;
		1823	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1824	ev_signal_start (loop, &signal_watcher);
		1825
		1826
624	=head2 C<ev_child> - wait for pid status changes	1827	=head2 C<ev_child> - watch out for process status changes
625		1828
626	Child watchers trigger when your process receives a SIGCHLD in response to	1829	Child watchers trigger when your process receives a SIGCHLD in response to
627	some child status changes (most typically when a child of yours dies).	1830	some child status changes (most typically when a child of yours dies or
		1831	exits). It is permissible to install a child watcher I<after> the child
		1832	has been forked (which implies it might have already exited), as long
		1833	as the event loop isn't entered (or is continued from a watcher), i.e.,
		1834	forking and then immediately registering a watcher for the child is fine,
		1835	but forking and registering a watcher a few event loop iterations later is
		1836	not.
		1837
		1838	Only the default event loop is capable of handling signals, and therefore
		1839	you can only register child watchers in the default event loop.
		1840
		1841	=head3 Process Interaction
		1842
		1843	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1844	initialised. This is necessary to guarantee proper behaviour even if
		1845	the first child watcher is started after the child exits. The occurrence
		1846	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1847	synchronously as part of the event loop processing. Libev always reaps all
		1848	children, even ones not watched.
		1849
		1850	=head3 Overriding the Built-In Processing
		1851
		1852	Libev offers no special support for overriding the built-in child
		1853	processing, but if your application collides with libev's default child
		1854	handler, you can override it easily by installing your own handler for
		1855	C<SIGCHLD> after initialising the default loop, and making sure the
		1856	default loop never gets destroyed. You are encouraged, however, to use an
		1857	event-based approach to child reaping and thus use libev's support for
		1858	that, so other libev users can use C<ev_child> watchers freely.
		1859
		1860	=head3 Stopping the Child Watcher
		1861
		1862	Currently, the child watcher never gets stopped, even when the
		1863	child terminates, so normally one needs to stop the watcher in the
		1864	callback. Future versions of libev might stop the watcher automatically
		1865	when a child exit is detected.
		1866
		1867	=head3 Watcher-Specific Functions and Data Members
628		1868
629	=over 4	1869	=over 4
630		1870
631	=item ev_child_init (ev_child *, callback, int pid)	1871	=item ev_child_init (ev_child *, callback, int pid, int trace)
632		1872
633	=item ev_child_set (ev_child *, int pid)	1873	=item ev_child_set (ev_child *, int pid, int trace)
634		1874
635	Configures the watcher to wait for status changes of process C<pid> (or	1875	Configures the watcher to wait for status changes of process C<pid> (or
636	I<any> process if C<pid> is specified as C<0>). The callback can look	1876	I<any> process if C<pid> is specified as C<0>). The callback can look
637	at the C<rstatus> member of the C<ev_child> watcher structure to see	1877	at the C<rstatus> member of the C<ev_child> watcher structure to see
638	the status word (use the macros from C<sys/wait.h> and see your systems	1878	the status word (use the macros from C<sys/wait.h> and see your systems
639	C<waitpid> documentation). The C<rpid> member contains the pid of the	1879	C<waitpid> documentation). The C<rpid> member contains the pid of the
640	process causing the status change.	1880	process causing the status change. C<trace> must be either C<0> (only
		1881	activate the watcher when the process terminates) or C<1> (additionally
		1882	activate the watcher when the process is stopped or continued).
		1883
		1884	=item int pid [read-only]
		1885
		1886	The process id this watcher watches out for, or C<0>, meaning any process id.
		1887
		1888	=item int rpid [read-write]
		1889
		1890	The process id that detected a status change.
		1891
		1892	=item int rstatus [read-write]
		1893
		1894	The process exit/trace status caused by C<rpid> (see your systems
		1895	C<waitpid> and C<sys/wait.h> documentation for details).
641		1896
642	=back	1897	=back
643		1898
		1899	=head3 Examples
		1900
		1901	Example: C<fork()> a new process and install a child handler to wait for
		1902	its completion.
		1903
		1904	ev_child cw;
		1905
		1906	static void
		1907	child_cb (EV_P_ ev_child *w, int revents)
		1908	{
		1909	ev_child_stop (EV_A_ w);
		1910	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		1911	}
		1912
		1913	pid_t pid = fork ();
		1914
		1915	if (pid < 0)
		1916	// error
		1917	else if (pid == 0)
		1918	{
		1919	// the forked child executes here
		1920	exit (1);
		1921	}
		1922	else
		1923	{
		1924	ev_child_init (&cw, child_cb, pid, 0);
		1925	ev_child_start (EV_DEFAULT_ &cw);
		1926	}
		1927
		1928
		1929	=head2 C<ev_stat> - did the file attributes just change?
		1930
		1931	This watches a file system path for attribute changes. That is, it calls
		1932	C<stat> on that path in regular intervals (or when the OS says it changed)
		1933	and sees if it changed compared to the last time, invoking the callback if
		1934	it did.
		1935
		1936	The path does not need to exist: changing from "path exists" to "path does
		1937	not exist" is a status change like any other. The condition "path does
		1938	not exist" is signified by the C<st_nlink> field being zero (which is
		1939	otherwise always forced to be at least one) and all the other fields of
		1940	the stat buffer having unspecified contents.
		1941
		1942	The path I<must not> end in a slash or contain special components such as
		1943	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		1944	your working directory changes, then the behaviour is undefined.
		1945
		1946	Since there is no portable change notification interface available, the
		1947	portable implementation simply calls C<stat(2)> regularly on the path
		1948	to see if it changed somehow. You can specify a recommended polling
		1949	interval for this case. If you specify a polling interval of C<0> (highly
		1950	recommended!) then a I<suitable, unspecified default> value will be used
		1951	(which you can expect to be around five seconds, although this might
		1952	change dynamically). Libev will also impose a minimum interval which is
		1953	currently around C<0.1>, but that's usually overkill.
		1954
		1955	This watcher type is not meant for massive numbers of stat watchers,
		1956	as even with OS-supported change notifications, this can be
		1957	resource-intensive.
		1958
		1959	At the time of this writing, the only OS-specific interface implemented
		1960	is the Linux inotify interface (implementing kqueue support is left as
		1961	an exercise for the reader. Note, however, that the author sees no way
		1962	of implementing C<ev_stat> semantics with kqueue).
		1963
		1964	=head3 ABI Issues (Largefile Support)
		1965
		1966	Libev by default (unless the user overrides this) uses the default
		1967	compilation environment, which means that on systems with large file
		1968	support disabled by default, you get the 32 bit version of the stat
		1969	structure. When using the library from programs that change the ABI to
		1970	use 64 bit file offsets the programs will fail. In that case you have to
		1971	compile libev with the same flags to get binary compatibility. This is
		1972	obviously the case with any flags that change the ABI, but the problem is
		1973	most noticeably displayed with ev_stat and large file support.
		1974
		1975	The solution for this is to lobby your distribution maker to make large
		1976	file interfaces available by default (as e.g. FreeBSD does) and not
		1977	optional. Libev cannot simply switch on large file support because it has
		1978	to exchange stat structures with application programs compiled using the
		1979	default compilation environment.
		1980
		1981	=head3 Inotify and Kqueue
		1982
		1983	When C<inotify (7)> support has been compiled into libev (generally
		1984	only available with Linux 2.6.25 or above due to bugs in earlier
		1985	implementations) and present at runtime, it will be used to speed up
		1986	change detection where possible. The inotify descriptor will be created
		1987	lazily when the first C<ev_stat> watcher is being started.
		1988
		1989	Inotify presence does not change the semantics of C<ev_stat> watchers
		1990	except that changes might be detected earlier, and in some cases, to avoid
		1991	making regular C<stat> calls. Even in the presence of inotify support
		1992	there are many cases where libev has to resort to regular C<stat> polling,
		1993	but as long as the path exists, libev usually gets away without polling.
		1994
		1995	There is no support for kqueue, as apparently it cannot be used to
		1996	implement this functionality, due to the requirement of having a file
		1997	descriptor open on the object at all times, and detecting renames, unlinks
		1998	etc. is difficult.
		1999
		2000	=head3 The special problem of stat time resolution
		2001
		2002	The C<stat ()> system call only supports full-second resolution portably,
		2003	and even on systems where the resolution is higher, most file systems
		2004	still only support whole seconds.
		2005
		2006	That means that, if the time is the only thing that changes, you can
		2007	easily miss updates: on the first update, C<ev_stat> detects a change and
		2008	calls your callback, which does something. When there is another update
		2009	within the same second, C<ev_stat> will be unable to detect unless the
		2010	stat data does change in other ways (e.g. file size).
		2011
		2012	The solution to this is to delay acting on a change for slightly more
		2013	than a second (or till slightly after the next full second boundary), using
		2014	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2015	ev_timer_again (loop, w)>).
		2016
		2017	The C<.02> offset is added to work around small timing inconsistencies
		2018	of some operating systems (where the second counter of the current time
		2019	might be be delayed. One such system is the Linux kernel, where a call to
		2020	C<gettimeofday> might return a timestamp with a full second later than
		2021	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2022	update file times then there will be a small window where the kernel uses
		2023	the previous second to update file times but libev might already execute
		2024	the timer callback).
		2025
		2026	=head3 Watcher-Specific Functions and Data Members
		2027
		2028	=over 4
		2029
		2030	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2031
		2032	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2033
		2034	Configures the watcher to wait for status changes of the given
		2035	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2036	be detected and should normally be specified as C<0> to let libev choose
		2037	a suitable value. The memory pointed to by C<path> must point to the same
		2038	path for as long as the watcher is active.
		2039
		2040	The callback will receive an C<EV_STAT> event when a change was detected,
		2041	relative to the attributes at the time the watcher was started (or the
		2042	last change was detected).
		2043
		2044	=item ev_stat_stat (loop, ev_stat *)
		2045
		2046	Updates the stat buffer immediately with new values. If you change the
		2047	watched path in your callback, you could call this function to avoid
		2048	detecting this change (while introducing a race condition if you are not
		2049	the only one changing the path). Can also be useful simply to find out the
		2050	new values.
		2051
		2052	=item ev_statdata attr [read-only]
		2053
		2054	The most-recently detected attributes of the file. Although the type is
		2055	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2056	suitable for your system, but you can only rely on the POSIX-standardised
		2057	members to be present. If the C<st_nlink> member is C<0>, then there was
		2058	some error while C<stat>ing the file.
		2059
		2060	=item ev_statdata prev [read-only]
		2061
		2062	The previous attributes of the file. The callback gets invoked whenever
		2063	C<prev> != C<attr>, or, more precisely, one or more of these members
		2064	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2065	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2066
		2067	=item ev_tstamp interval [read-only]
		2068
		2069	The specified interval.
		2070
		2071	=item const char *path [read-only]
		2072
		2073	The file system path that is being watched.
		2074
		2075	=back
		2076
		2077	=head3 Examples
		2078
		2079	Example: Watch C</etc/passwd> for attribute changes.
		2080
		2081	static void
		2082	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2083	{
		2084	/* /etc/passwd changed in some way */
		2085	if (w->attr.st_nlink)
		2086	{
		2087	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2088	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2089	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2090	}
		2091	else
		2092	/* you shalt not abuse printf for puts */
		2093	puts ("wow, /etc/passwd is not there, expect problems. "
		2094	"if this is windows, they already arrived\n");
		2095	}
		2096
		2097	...
		2098	ev_stat passwd;
		2099
		2100	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2101	ev_stat_start (loop, &passwd);
		2102
		2103	Example: Like above, but additionally use a one-second delay so we do not
		2104	miss updates (however, frequent updates will delay processing, too, so
		2105	one might do the work both on C<ev_stat> callback invocation I<and> on
		2106	C<ev_timer> callback invocation).
		2107
		2108	static ev_stat passwd;
		2109	static ev_timer timer;
		2110
		2111	static void
		2112	timer_cb (EV_P_ ev_timer *w, int revents)
		2113	{
		2114	ev_timer_stop (EV_A_ w);
		2115
		2116	/* now it's one second after the most recent passwd change */
		2117	}
		2118
		2119	static void
		2120	stat_cb (EV_P_ ev_stat *w, int revents)
		2121	{
		2122	/* reset the one-second timer */
		2123	ev_timer_again (EV_A_ &timer);
		2124	}
		2125
		2126	...
		2127	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2128	ev_stat_start (loop, &passwd);
		2129	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2130
		2131
644	=head2 C<ev_idle> - when you've got nothing better to do	2132	=head2 C<ev_idle> - when you've got nothing better to do...
645		2133
646	Idle watchers trigger events when there are no other events are pending	2134	Idle watchers trigger events when no other events of the same or higher
647	(prepare, check and other idle watchers do not count). That is, as long	2135	priority are pending (prepare, check and other idle watchers do not count
648	as your process is busy handling sockets or timeouts (or even signals,	2136	as receiving "events").
649	imagine) it will not be triggered. But when your process is idle all idle	2137
650	watchers are being called again and again, once per event loop iteration -	2138	That is, as long as your process is busy handling sockets or timeouts
		2139	(or even signals, imagine) of the same or higher priority it will not be
		2140	triggered. But when your process is idle (or only lower-priority watchers
		2141	are pending), the idle watchers are being called once per event loop
651	until stopped, that is, or your process receives more events and becomes	2142	iteration - until stopped, that is, or your process receives more events
652	busy.	2143	and becomes busy again with higher priority stuff.
653		2144
654	The most noteworthy effect is that as long as any idle watchers are	2145	The most noteworthy effect is that as long as any idle watchers are
655	active, the process will not block when waiting for new events.	2146	active, the process will not block when waiting for new events.
656		2147
657	Apart from keeping your process non-blocking (which is a useful	2148	Apart from keeping your process non-blocking (which is a useful
658	effect on its own sometimes), idle watchers are a good place to do	2149	effect on its own sometimes), idle watchers are a good place to do
659	"pseudo-background processing", or delay processing stuff to after the	2150	"pseudo-background processing", or delay processing stuff to after the
660	event loop has handled all outstanding events.	2151	event loop has handled all outstanding events.
661		2152
		2153	=head3 Watcher-Specific Functions and Data Members
		2154
662	=over 4	2155	=over 4
663		2156
664	=item ev_idle_init (ev_signal *, callback)	2157	=item ev_idle_init (ev_signal *, callback)
665		2158
666	Initialises and configures the idle watcher - it has no parameters of any	2159	Initialises and configures the idle watcher - it has no parameters of any
667	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2160	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
668	believe me.	2161	believe me.
669		2162
670	=back	2163	=back
671		2164
		2165	=head3 Examples
		2166
		2167	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2168	callback, free it. Also, use no error checking, as usual.
		2169
		2170	static void
		2171	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2172	{
		2173	free (w);
		2174	// now do something you wanted to do when the program has
		2175	// no longer anything immediate to do.
		2176	}
		2177
		2178	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2179	ev_idle_init (idle_watcher, idle_cb);
		2180	ev_idle_start (loop, idle_cb);
		2181
		2182
672	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2183	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
673		2184
674	Prepare and check watchers are usually (but not always) used in tandem:	2185	Prepare and check watchers are usually (but not always) used in pairs:
675	Prepare watchers get invoked before the process blocks and check watchers	2186	prepare watchers get invoked before the process blocks and check watchers
676	afterwards.	2187	afterwards.
677		2188
		2189	You I<must not> call C<ev_loop> or similar functions that enter
		2190	the current event loop from either C<ev_prepare> or C<ev_check>
		2191	watchers. Other loops than the current one are fine, however. The
		2192	rationale behind this is that you do not need to check for recursion in
		2193	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2194	C<ev_check> so if you have one watcher of each kind they will always be
		2195	called in pairs bracketing the blocking call.
		2196
678	Their main purpose is to integrate other event mechanisms into libev. This	2197	Their main purpose is to integrate other event mechanisms into libev and
679	could be used, for example, to track variable changes, implement your own	2198	their use is somewhat advanced. They could be used, for example, to track
680	watchers, integrate net-snmp or a coroutine library and lots more.	2199	variable changes, implement your own watchers, integrate net-snmp or a
		2200	coroutine library and lots more. They are also occasionally useful if
		2201	you cache some data and want to flush it before blocking (for example,
		2202	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2203	watcher).
681		2204
682	This is done by examining in each prepare call which file descriptors need	2205	This is done by examining in each prepare call which file descriptors
683	to be watched by the other library, registering C<ev_io> watchers for	2206	need to be watched by the other library, registering C<ev_io> watchers
684	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2207	for them and starting an C<ev_timer> watcher for any timeouts (many
685	provide just this functionality). Then, in the check watcher you check for	2208	libraries provide exactly this functionality). Then, in the check watcher,
686	any events that occured (by checking the pending status of all watchers	2209	you check for any events that occurred (by checking the pending status
687	and stopping them) and call back into the library. The I/O and timer	2210	of all watchers and stopping them) and call back into the library. The
688	callbacks will never actually be called (but must be valid neverthelles,	2211	I/O and timer callbacks will never actually be called (but must be valid
689	because you never know, you know?).	2212	nevertheless, because you never know, you know?).
690		2213
691	As another example, the Perl Coro module uses these hooks to integrate	2214	As another example, the Perl Coro module uses these hooks to integrate
692	coroutines into libev programs, by yielding to other active coroutines	2215	coroutines into libev programs, by yielding to other active coroutines
693	during each prepare and only letting the process block if no coroutines	2216	during each prepare and only letting the process block if no coroutines
694	are ready to run (its actually more complicated, it only runs coroutines	2217	are ready to run (it's actually more complicated: it only runs coroutines
695	with priority higher than the event loop and one lower priority once,	2218	with priority higher than or equal to the event loop and one coroutine
696	using idle watchers to keep the event loop from blocking if lower-priority	2219	of lower priority, but only once, using idle watchers to keep the event
697	coroutines exist, thus mapping low-priority coroutines to idle/background	2220	loop from blocking if lower-priority coroutines are active, thus mapping
698	tasks).	2221	low-priority coroutines to idle/background tasks).
		2222
		2223	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2224	priority, to ensure that they are being run before any other watchers
		2225	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2226
		2227	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2228	activate ("feed") events into libev. While libev fully supports this, they
		2229	might get executed before other C<ev_check> watchers did their job. As
		2230	C<ev_check> watchers are often used to embed other (non-libev) event
		2231	loops those other event loops might be in an unusable state until their
		2232	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2233	others).
		2234
		2235	=head3 Watcher-Specific Functions and Data Members
699		2236
700	=over 4	2237	=over 4
701		2238
702	=item ev_prepare_init (ev_prepare *, callback)	2239	=item ev_prepare_init (ev_prepare *, callback)
703		2240
704	=item ev_check_init (ev_check *, callback)	2241	=item ev_check_init (ev_check *, callback)
705		2242
706	Initialises and configures the prepare or check watcher - they have no	2243	Initialises and configures the prepare or check watcher - they have no
707	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2244	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
708	macros, but using them is utterly, utterly and completely pointless.	2245	macros, but using them is utterly, utterly, utterly and completely
		2246	pointless.
709		2247
710	=back	2248	=back
711		2249
		2250	=head3 Examples
		2251
		2252	There are a number of principal ways to embed other event loops or modules
		2253	into libev. Here are some ideas on how to include libadns into libev
		2254	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2255	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2256	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2257	Glib event loop).
		2258
		2259	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2260	and in a check watcher, destroy them and call into libadns. What follows
		2261	is pseudo-code only of course. This requires you to either use a low
		2262	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2263	the callbacks for the IO/timeout watchers might not have been called yet.
		2264
		2265	static ev_io iow [nfd];
		2266	static ev_timer tw;
		2267
		2268	static void
		2269	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2270	{
		2271	}
		2272
		2273	// create io watchers for each fd and a timer before blocking
		2274	static void
		2275	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2276	{
		2277	int timeout = 3600000;
		2278	struct pollfd fds [nfd];
		2279	// actual code will need to loop here and realloc etc.
		2280	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2281
		2282	/* the callback is illegal, but won't be called as we stop during check */
		2283	ev_timer_init (&tw, 0, timeout * 1e-3);
		2284	ev_timer_start (loop, &tw);
		2285
		2286	// create one ev_io per pollfd
		2287	for (int i = 0; i < nfd; ++i)
		2288	{
		2289	ev_io_init (iow + i, io_cb, fds [i].fd,
		2290	((fds [i].events & POLLIN ? EV_READ : 0)
		2291	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2292
		2293	fds [i].revents = 0;
		2294	ev_io_start (loop, iow + i);
		2295	}
		2296	}
		2297
		2298	// stop all watchers after blocking
		2299	static void
		2300	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2301	{
		2302	ev_timer_stop (loop, &tw);
		2303
		2304	for (int i = 0; i < nfd; ++i)
		2305	{
		2306	// set the relevant poll flags
		2307	// could also call adns_processreadable etc. here
		2308	struct pollfd *fd = fds + i;
		2309	int revents = ev_clear_pending (iow + i);
		2310	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2311	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2312
		2313	// now stop the watcher
		2314	ev_io_stop (loop, iow + i);
		2315	}
		2316
		2317	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2318	}
		2319
		2320	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2321	in the prepare watcher and would dispose of the check watcher.
		2322
		2323	Method 3: If the module to be embedded supports explicit event
		2324	notification (libadns does), you can also make use of the actual watcher
		2325	callbacks, and only destroy/create the watchers in the prepare watcher.
		2326
		2327	static void
		2328	timer_cb (EV_P_ ev_timer *w, int revents)
		2329	{
		2330	adns_state ads = (adns_state)w->data;
		2331	update_now (EV_A);
		2332
		2333	adns_processtimeouts (ads, &tv_now);
		2334	}
		2335
		2336	static void
		2337	io_cb (EV_P_ ev_io *w, int revents)
		2338	{
		2339	adns_state ads = (adns_state)w->data;
		2340	update_now (EV_A);
		2341
		2342	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2343	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2344	}
		2345
		2346	// do not ever call adns_afterpoll
		2347
		2348	Method 4: Do not use a prepare or check watcher because the module you
		2349	want to embed is not flexible enough to support it. Instead, you can
		2350	override their poll function. The drawback with this solution is that the
		2351	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2352	this approach, effectively embedding EV as a client into the horrible
		2353	libglib event loop.
		2354
		2355	static gint
		2356	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2357	{
		2358	int got_events = 0;
		2359
		2360	for (n = 0; n < nfds; ++n)
		2361	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2362
		2363	if (timeout >= 0)
		2364	// create/start timer
		2365
		2366	// poll
		2367	ev_loop (EV_A_ 0);
		2368
		2369	// stop timer again
		2370	if (timeout >= 0)
		2371	ev_timer_stop (EV_A_ &to);
		2372
		2373	// stop io watchers again - their callbacks should have set
		2374	for (n = 0; n < nfds; ++n)
		2375	ev_io_stop (EV_A_ iow [n]);
		2376
		2377	return got_events;
		2378	}
		2379
		2380
		2381	=head2 C<ev_embed> - when one backend isn't enough...
		2382
		2383	This is a rather advanced watcher type that lets you embed one event loop
		2384	into another (currently only C<ev_io> events are supported in the embedded
		2385	loop, other types of watchers might be handled in a delayed or incorrect
		2386	fashion and must not be used).
		2387
		2388	There are primarily two reasons you would want that: work around bugs and
		2389	prioritise I/O.
		2390
		2391	As an example for a bug workaround, the kqueue backend might only support
		2392	sockets on some platform, so it is unusable as generic backend, but you
		2393	still want to make use of it because you have many sockets and it scales
		2394	so nicely. In this case, you would create a kqueue-based loop and embed
		2395	it into your default loop (which might use e.g. poll). Overall operation
		2396	will be a bit slower because first libev has to call C<poll> and then
		2397	C<kevent>, but at least you can use both mechanisms for what they are
		2398	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2399
		2400	As for prioritising I/O: under rare circumstances you have the case where
		2401	some fds have to be watched and handled very quickly (with low latency),
		2402	and even priorities and idle watchers might have too much overhead. In
		2403	this case you would put all the high priority stuff in one loop and all
		2404	the rest in a second one, and embed the second one in the first.
		2405
		2406	As long as the watcher is active, the callback will be invoked every time
		2407	there might be events pending in the embedded loop. The callback must then
		2408	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
		2409	their callbacks (you could also start an idle watcher to give the embedded
		2410	loop strictly lower priority for example). You can also set the callback
		2411	to C<0>, in which case the embed watcher will automatically execute the
		2412	embedded loop sweep.
		2413
		2414	As long as the watcher is started it will automatically handle events. The
		2415	callback will be invoked whenever some events have been handled. You can
		2416	set the callback to C<0> to avoid having to specify one if you are not
		2417	interested in that.
		2418
		2419	Also, there have not currently been made special provisions for forking:
		2420	when you fork, you not only have to call C<ev_loop_fork> on both loops,
		2421	but you will also have to stop and restart any C<ev_embed> watchers
		2422	yourself - but you can use a fork watcher to handle this automatically,
		2423	and future versions of libev might do just that.
		2424
		2425	Unfortunately, not all backends are embeddable: only the ones returned by
		2426	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2427	portable one.
		2428
		2429	So when you want to use this feature you will always have to be prepared
		2430	that you cannot get an embeddable loop. The recommended way to get around
		2431	this is to have a separate variables for your embeddable loop, try to
		2432	create it, and if that fails, use the normal loop for everything.
		2433
		2434	=head3 C<ev_embed> and fork
		2435
		2436	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2437	automatically be applied to the embedded loop as well, so no special
		2438	fork handling is required in that case. When the watcher is not running,
		2439	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2440	as applicable.
		2441
		2442	=head3 Watcher-Specific Functions and Data Members
		2443
		2444	=over 4
		2445
		2446	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2447
		2448	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2449
		2450	Configures the watcher to embed the given loop, which must be
		2451	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2452	invoked automatically, otherwise it is the responsibility of the callback
		2453	to invoke it (it will continue to be called until the sweep has been done,
		2454	if you do not want that, you need to temporarily stop the embed watcher).
		2455
		2456	=item ev_embed_sweep (loop, ev_embed *)
		2457
		2458	Make a single, non-blocking sweep over the embedded loop. This works
		2459	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2460	appropriate way for embedded loops.
		2461
		2462	=item struct ev_loop *other [read-only]
		2463
		2464	The embedded event loop.
		2465
		2466	=back
		2467
		2468	=head3 Examples
		2469
		2470	Example: Try to get an embeddable event loop and embed it into the default
		2471	event loop. If that is not possible, use the default loop. The default
		2472	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2473	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2474	used).
		2475
		2476	struct ev_loop *loop_hi = ev_default_init (0);
		2477	struct ev_loop *loop_lo = 0;
		2478	ev_embed embed;
		2479
		2480	// see if there is a chance of getting one that works
		2481	// (remember that a flags value of 0 means autodetection)
		2482	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2483	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2484	: 0;
		2485
		2486	// if we got one, then embed it, otherwise default to loop_hi
		2487	if (loop_lo)
		2488	{
		2489	ev_embed_init (&embed, 0, loop_lo);
		2490	ev_embed_start (loop_hi, &embed);
		2491	}
		2492	else
		2493	loop_lo = loop_hi;
		2494
		2495	Example: Check if kqueue is available but not recommended and create
		2496	a kqueue backend for use with sockets (which usually work with any
		2497	kqueue implementation). Store the kqueue/socket-only event loop in
		2498	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2499
		2500	struct ev_loop *loop = ev_default_init (0);
		2501	struct ev_loop *loop_socket = 0;
		2502	ev_embed embed;
		2503
		2504	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2505	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2506	{
		2507	ev_embed_init (&embed, 0, loop_socket);
		2508	ev_embed_start (loop, &embed);
		2509	}
		2510
		2511	if (!loop_socket)
		2512	loop_socket = loop;
		2513
		2514	// now use loop_socket for all sockets, and loop for everything else
		2515
		2516
		2517	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2518
		2519	Fork watchers are called when a C<fork ()> was detected (usually because
		2520	whoever is a good citizen cared to tell libev about it by calling
		2521	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2522	event loop blocks next and before C<ev_check> watchers are being called,
		2523	and only in the child after the fork. If whoever good citizen calling
		2524	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2525	handlers will be invoked, too, of course.
		2526
		2527	=head3 Watcher-Specific Functions and Data Members
		2528
		2529	=over 4
		2530
		2531	=item ev_fork_init (ev_signal *, callback)
		2532
		2533	Initialises and configures the fork watcher - it has no parameters of any
		2534	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2535	believe me.
		2536
		2537	=back
		2538
		2539
		2540	=head2 C<ev_async> - how to wake up another event loop
		2541
		2542	In general, you cannot use an C<ev_loop> from multiple threads or other
		2543	asynchronous sources such as signal handlers (as opposed to multiple event
		2544	loops - those are of course safe to use in different threads).
		2545
		2546	Sometimes, however, you need to wake up another event loop you do not
		2547	control, for example because it belongs to another thread. This is what
		2548	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2549	can signal it by calling C<ev_async_send>, which is thread- and signal
		2550	safe.
		2551
		2552	This functionality is very similar to C<ev_signal> watchers, as signals,
		2553	too, are asynchronous in nature, and signals, too, will be compressed
		2554	(i.e. the number of callback invocations may be less than the number of
		2555	C<ev_async_sent> calls).
		2556
		2557	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2558	just the default loop.
		2559
		2560	=head3 Queueing
		2561
		2562	C<ev_async> does not support queueing of data in any way. The reason
		2563	is that the author does not know of a simple (or any) algorithm for a
		2564	multiple-writer-single-reader queue that works in all cases and doesn't
		2565	need elaborate support such as pthreads.
		2566
		2567	That means that if you want to queue data, you have to provide your own
		2568	queue. But at least I can tell you how to implement locking around your
		2569	queue:
		2570
		2571	=over 4
		2572
		2573	=item queueing from a signal handler context
		2574
		2575	To implement race-free queueing, you simply add to the queue in the signal
		2576	handler but you block the signal handler in the watcher callback. Here is
		2577	an example that does that for some fictitious SIGUSR1 handler:
		2578
		2579	static ev_async mysig;
		2580
		2581	static void
		2582	sigusr1_handler (void)
		2583	{
		2584	sometype data;
		2585
		2586	// no locking etc.
		2587	queue_put (data);
		2588	ev_async_send (EV_DEFAULT_ &mysig);
		2589	}
		2590
		2591	static void
		2592	mysig_cb (EV_P_ ev_async *w, int revents)
		2593	{
		2594	sometype data;
		2595	sigset_t block, prev;
		2596
		2597	sigemptyset (&block);
		2598	sigaddset (&block, SIGUSR1);
		2599	sigprocmask (SIG_BLOCK, &block, &prev);
		2600
		2601	while (queue_get (&data))
		2602	process (data);
		2603
		2604	if (sigismember (&prev, SIGUSR1)
		2605	sigprocmask (SIG_UNBLOCK, &block, 0);
		2606	}
		2607
		2608	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2609	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2610	either...).
		2611
		2612	=item queueing from a thread context
		2613
		2614	The strategy for threads is different, as you cannot (easily) block
		2615	threads but you can easily preempt them, so to queue safely you need to
		2616	employ a traditional mutex lock, such as in this pthread example:
		2617
		2618	static ev_async mysig;
		2619	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2620
		2621	static void
		2622	otherthread (void)
		2623	{
		2624	// only need to lock the actual queueing operation
		2625	pthread_mutex_lock (&mymutex);
		2626	queue_put (data);
		2627	pthread_mutex_unlock (&mymutex);
		2628
		2629	ev_async_send (EV_DEFAULT_ &mysig);
		2630	}
		2631
		2632	static void
		2633	mysig_cb (EV_P_ ev_async *w, int revents)
		2634	{
		2635	pthread_mutex_lock (&mymutex);
		2636
		2637	while (queue_get (&data))
		2638	process (data);
		2639
		2640	pthread_mutex_unlock (&mymutex);
		2641	}
		2642
		2643	=back
		2644
		2645
		2646	=head3 Watcher-Specific Functions and Data Members
		2647
		2648	=over 4
		2649
		2650	=item ev_async_init (ev_async *, callback)
		2651
		2652	Initialises and configures the async watcher - it has no parameters of any
		2653	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		2654	trust me.
		2655
		2656	=item ev_async_send (loop, ev_async *)
		2657
		2658	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2659	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2660	C<ev_feed_event>, this call is safe to do from other threads, signal or
		2661	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		2662	section below on what exactly this means).
		2663
		2664	This call incurs the overhead of a system call only once per loop iteration,
		2665	so while the overhead might be noticeable, it doesn't apply to repeated
		2666	calls to C<ev_async_send>.
		2667
		2668	=item bool = ev_async_pending (ev_async *)
		2669
		2670	Returns a non-zero value when C<ev_async_send> has been called on the
		2671	watcher but the event has not yet been processed (or even noted) by the
		2672	event loop.
		2673
		2674	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2675	the loop iterates next and checks for the watcher to have become active,
		2676	it will reset the flag again. C<ev_async_pending> can be used to very
		2677	quickly check whether invoking the loop might be a good idea.
		2678
		2679	Not that this does I<not> check whether the watcher itself is pending, only
		2680	whether it has been requested to make this watcher pending.
		2681
		2682	=back
		2683
		2684
712	=head1 OTHER FUNCTIONS	2685	=head1 OTHER FUNCTIONS
713		2686
714	There are some other functions of possible interest. Described. Here. Now.	2687	There are some other functions of possible interest. Described. Here. Now.
715		2688
716	=over 4	2689	=over 4
717		2690
718	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2691	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
719		2692
720	This function combines a simple timer and an I/O watcher, calls your	2693	This function combines a simple timer and an I/O watcher, calls your
721	callback on whichever event happens first and automatically stop both	2694	callback on whichever event happens first and automatically stops both
722	watchers. This is useful if you want to wait for a single event on an fd	2695	watchers. This is useful if you want to wait for a single event on an fd
723	or timeout without havign to allocate/configure/start/stop/free one or	2696	or timeout without having to allocate/configure/start/stop/free one or
724	more watchers yourself.	2697	more watchers yourself.
725		2698
726	If C<fd> is less than 0, then no I/O watcher will be started and events	2699	If C<fd> is less than 0, then no I/O watcher will be started and the
727	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2700	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
728	C<events> set will be craeted and started.	2701	the given C<fd> and C<events> set will be created and started.
729		2702
730	If C<timeout> is less than 0, then no timeout watcher will be	2703	If C<timeout> is less than 0, then no timeout watcher will be
731	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2704	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
732	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2705	repeat = 0) will be started. C<0> is a valid timeout.
733	dubious value.
734		2706
735	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	2707	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
736	passed an events set like normal event callbacks (with a combination of	2708	passed an C<revents> set like normal event callbacks (a combination of
737	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2709	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
738	value passed to C<ev_once>:	2710	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2711	a timeout and an io event at the same time - you probably should give io
		2712	events precedence.
739		2713
		2714	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		2715
740	static void stdin_ready (int revents, void *arg)	2716	static void stdin_ready (int revents, void *arg)
741	{	2717	{
742	if (revents & EV_TIMEOUT)
743	/* doh, nothing entered */;
744	else if (revents & EV_READ)	2718	if (revents & EV_READ)
745	/* stdin might have data for us, joy! */;	2719	/* stdin might have data for us, joy! */;
		2720	else if (revents & EV_TIMEOUT)
		2721	/* doh, nothing entered */;
746	}	2722	}
747		2723
748	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2724	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
749		2725
750	=item ev_feed_event (loop, watcher, int events)	2726	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
751		2727
752	Feeds the given event set into the event loop, as if the specified event	2728	Feeds the given event set into the event loop, as if the specified event
753	had happened for the specified watcher (which must be a pointer to an	2729	had happened for the specified watcher (which must be a pointer to an
754	initialised but not necessarily started event watcher).	2730	initialised but not necessarily started event watcher).
755		2731
756	=item ev_feed_fd_event (loop, int fd, int revents)	2732	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
757		2733
758	Feed an event on the given fd, as if a file descriptor backend detected	2734	Feed an event on the given fd, as if a file descriptor backend detected
759	the given events it.	2735	the given events it.
760		2736
761	=item ev_feed_signal_event (loop, int signum)	2737	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
762		2738
763	Feed an event as if the given signal occured (loop must be the default loop!).	2739	Feed an event as if the given signal occurred (C<loop> must be the default
		2740	loop!).
764		2741
765	=back	2742	=back
766		2743
		2744
		2745	=head1 LIBEVENT EMULATION
		2746
		2747	Libev offers a compatibility emulation layer for libevent. It cannot
		2748	emulate the internals of libevent, so here are some usage hints:
		2749
		2750	=over 4
		2751
		2752	=item * Use it by including <event.h>, as usual.
		2753
		2754	=item * The following members are fully supported: ev_base, ev_callback,
		2755	ev_arg, ev_fd, ev_res, ev_events.
		2756
		2757	=item * Avoid using ev_flags and the EVLIST_*-macros, while it is
		2758	maintained by libev, it does not work exactly the same way as in libevent (consider
		2759	it a private API).
		2760
		2761	=item * Priorities are not currently supported. Initialising priorities
		2762	will fail and all watchers will have the same priority, even though there
		2763	is an ev_pri field.
		2764
		2765	=item * In libevent, the last base created gets the signals, in libev, the
		2766	first base created (== the default loop) gets the signals.
		2767
		2768	=item * Other members are not supported.
		2769
		2770	=item * The libev emulation is I<not> ABI compatible to libevent, you need
		2771	to use the libev header file and library.
		2772
		2773	=back
		2774
		2775	=head1 C++ SUPPORT
		2776
		2777	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		2778	you to use some convenience methods to start/stop watchers and also change
		2779	the callback model to a model using method callbacks on objects.
		2780
		2781	To use it,
		2782
		2783	#include <ev++.h>
		2784
		2785	This automatically includes F<ev.h> and puts all of its definitions (many
		2786	of them macros) into the global namespace. All C++ specific things are
		2787	put into the C<ev> namespace. It should support all the same embedding
		2788	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		2789
		2790	Care has been taken to keep the overhead low. The only data member the C++
		2791	classes add (compared to plain C-style watchers) is the event loop pointer
		2792	that the watcher is associated with (or no additional members at all if
		2793	you disable C<EV_MULTIPLICITY> when embedding libev).
		2794
		2795	Currently, functions, and static and non-static member functions can be
		2796	used as callbacks. Other types should be easy to add as long as they only
		2797	need one additional pointer for context. If you need support for other
		2798	types of functors please contact the author (preferably after implementing
		2799	it).
		2800
		2801	Here is a list of things available in the C<ev> namespace:
		2802
		2803	=over 4
		2804
		2805	=item C<ev::READ>, C<ev::WRITE> etc.
		2806
		2807	These are just enum values with the same values as the C<EV_READ> etc.
		2808	macros from F<ev.h>.
		2809
		2810	=item C<ev::tstamp>, C<ev::now>
		2811
		2812	Aliases to the same types/functions as with the C<ev_> prefix.
		2813
		2814	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		2815
		2816	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		2817	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		2818	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		2819	defines by many implementations.
		2820
		2821	All of those classes have these methods:
		2822
		2823	=over 4
		2824
		2825	=item ev::TYPE::TYPE ()
		2826
		2827	=item ev::TYPE::TYPE (struct ev_loop *)
		2828
		2829	=item ev::TYPE::~TYPE
		2830
		2831	The constructor (optionally) takes an event loop to associate the watcher
		2832	with. If it is omitted, it will use C<EV_DEFAULT>.
		2833
		2834	The constructor calls C<ev_init> for you, which means you have to call the
		2835	C<set> method before starting it.
		2836
		2837	It will not set a callback, however: You have to call the templated C<set>
		2838	method to set a callback before you can start the watcher.
		2839
		2840	(The reason why you have to use a method is a limitation in C++ which does
		2841	not allow explicit template arguments for constructors).
		2842
		2843	The destructor automatically stops the watcher if it is active.
		2844
		2845	=item w->set<class, &class::method> (object *)
		2846
		2847	This method sets the callback method to call. The method has to have a
		2848	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		2849	first argument and the C<revents> as second. The object must be given as
		2850	parameter and is stored in the C<data> member of the watcher.
		2851
		2852	This method synthesizes efficient thunking code to call your method from
		2853	the C callback that libev requires. If your compiler can inline your
		2854	callback (i.e. it is visible to it at the place of the C<set> call and
		2855	your compiler is good :), then the method will be fully inlined into the
		2856	thunking function, making it as fast as a direct C callback.
		2857
		2858	Example: simple class declaration and watcher initialisation
		2859
		2860	struct myclass
		2861	{
		2862	void io_cb (ev::io &w, int revents) { }
		2863	}
		2864
		2865	myclass obj;
		2866	ev::io iow;
		2867	iow.set <myclass, &myclass::io_cb> (&obj);
		2868
		2869	=item w->set<function> (void *data = 0)
		2870
		2871	Also sets a callback, but uses a static method or plain function as
		2872	callback. The optional C<data> argument will be stored in the watcher's
		2873	C<data> member and is free for you to use.
		2874
		2875	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		2876
		2877	See the method-C<set> above for more details.
		2878
		2879	Example: Use a plain function as callback.
		2880
		2881	static void io_cb (ev::io &w, int revents) { }
		2882	iow.set <io_cb> ();
		2883
		2884	=item w->set (struct ev_loop *)
		2885
		2886	Associates a different C<struct ev_loop> with this watcher. You can only
		2887	do this when the watcher is inactive (and not pending either).
		2888
		2889	=item w->set ([arguments])
		2890
		2891	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		2892	called at least once. Unlike the C counterpart, an active watcher gets
		2893	automatically stopped and restarted when reconfiguring it with this
		2894	method.
		2895
		2896	=item w->start ()
		2897
		2898	Starts the watcher. Note that there is no C<loop> argument, as the
		2899	constructor already stores the event loop.
		2900
		2901	=item w->stop ()
		2902
		2903	Stops the watcher if it is active. Again, no C<loop> argument.
		2904
		2905	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		2906
		2907	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		2908	C<ev_TYPE_again> function.
		2909
		2910	=item w->sweep () (C<ev::embed> only)
		2911
		2912	Invokes C<ev_embed_sweep>.
		2913
		2914	=item w->update () (C<ev::stat> only)
		2915
		2916	Invokes C<ev_stat_stat>.
		2917
		2918	=back
		2919
		2920	=back
		2921
		2922	Example: Define a class with an IO and idle watcher, start one of them in
		2923	the constructor.
		2924
		2925	class myclass
		2926	{
		2927	ev::io io ; void io_cb (ev::io &w, int revents);
		2928	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		2929
		2930	myclass (int fd)
		2931	{
		2932	io .set <myclass, &myclass::io_cb > (this);
		2933	idle.set <myclass, &myclass::idle_cb> (this);
		2934
		2935	io.start (fd, ev::READ);
		2936	}
		2937	};
		2938
		2939
		2940	=head1 OTHER LANGUAGE BINDINGS
		2941
		2942	Libev does not offer other language bindings itself, but bindings for a
		2943	number of languages exist in the form of third-party packages. If you know
		2944	any interesting language binding in addition to the ones listed here, drop
		2945	me a note.
		2946
		2947	=over 4
		2948
		2949	=item Perl
		2950
		2951	The EV module implements the full libev API and is actually used to test
		2952	libev. EV is developed together with libev. Apart from the EV core module,
		2953	there are additional modules that implement libev-compatible interfaces
		2954	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		2955	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		2956	and C<EV::Glib>).
		2957
		2958	It can be found and installed via CPAN, its homepage is at
		2959	L<http://software.schmorp.de/pkg/EV>.
		2960
		2961	=item Python
		2962
		2963	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		2964	seems to be quite complete and well-documented. Note, however, that the
		2965	patch they require for libev is outright dangerous as it breaks the ABI
		2966	for everybody else, and therefore, should never be applied in an installed
		2967	libev (if python requires an incompatible ABI then it needs to embed
		2968	libev).
		2969
		2970	=item Ruby
		2971
		2972	Tony Arcieri has written a ruby extension that offers access to a subset
		2973	of the libev API and adds file handle abstractions, asynchronous DNS and
		2974	more on top of it. It can be found via gem servers. Its homepage is at
		2975	L<http://rev.rubyforge.org/>.
		2976
		2977	=item D
		2978
		2979	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		2980	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		2981
		2982	=item Ocaml
		2983
		2984	Erkki Seppala has written Ocaml bindings for libev, to be found at
		2985	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		2986
		2987	=back
		2988
		2989
		2990	=head1 MACRO MAGIC
		2991
		2992	Libev can be compiled with a variety of options, the most fundamental
		2993	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		2994	functions and callbacks have an initial C<struct ev_loop *> argument.
		2995
		2996	To make it easier to write programs that cope with either variant, the
		2997	following macros are defined:
		2998
		2999	=over 4
		3000
		3001	=item C<EV_A>, C<EV_A_>
		3002
		3003	This provides the loop I<argument> for functions, if one is required ("ev
		3004	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3005	C<EV_A_> is used when other arguments are following. Example:
		3006
		3007	ev_unref (EV_A);
		3008	ev_timer_add (EV_A_ watcher);
		3009	ev_loop (EV_A_ 0);
		3010
		3011	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3012	which is often provided by the following macro.
		3013
		3014	=item C<EV_P>, C<EV_P_>
		3015
		3016	This provides the loop I<parameter> for functions, if one is required ("ev
		3017	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3018	C<EV_P_> is used when other parameters are following. Example:
		3019
		3020	// this is how ev_unref is being declared
		3021	static void ev_unref (EV_P);
		3022
		3023	// this is how you can declare your typical callback
		3024	static void cb (EV_P_ ev_timer *w, int revents)
		3025
		3026	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3027	suitable for use with C<EV_A>.
		3028
		3029	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3030
		3031	Similar to the other two macros, this gives you the value of the default
		3032	loop, if multiple loops are supported ("ev loop default").
		3033
		3034	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3035
		3036	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3037	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3038	is undefined when the default loop has not been initialised by a previous
		3039	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3040
		3041	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3042	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3043
		3044	=back
		3045
		3046	Example: Declare and initialise a check watcher, utilising the above
		3047	macros so it will work regardless of whether multiple loops are supported
		3048	or not.
		3049
		3050	static void
		3051	check_cb (EV_P_ ev_timer *w, int revents)
		3052	{
		3053	ev_check_stop (EV_A_ w);
		3054	}
		3055
		3056	ev_check check;
		3057	ev_check_init (&check, check_cb);
		3058	ev_check_start (EV_DEFAULT_ &check);
		3059	ev_loop (EV_DEFAULT_ 0);
		3060
		3061	=head1 EMBEDDING
		3062
		3063	Libev can (and often is) directly embedded into host
		3064	applications. Examples of applications that embed it include the Deliantra
		3065	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3066	and rxvt-unicode.
		3067
		3068	The goal is to enable you to just copy the necessary files into your
		3069	source directory without having to change even a single line in them, so
		3070	you can easily upgrade by simply copying (or having a checked-out copy of
		3071	libev somewhere in your source tree).
		3072
		3073	=head2 FILESETS
		3074
		3075	Depending on what features you need you need to include one or more sets of files
		3076	in your application.
		3077
		3078	=head3 CORE EVENT LOOP
		3079
		3080	To include only the libev core (all the C<ev_*> functions), with manual
		3081	configuration (no autoconf):
		3082
		3083	#define EV_STANDALONE 1
		3084	#include "ev.c"
		3085
		3086	This will automatically include F<ev.h>, too, and should be done in a
		3087	single C source file only to provide the function implementations. To use
		3088	it, do the same for F<ev.h> in all files wishing to use this API (best
		3089	done by writing a wrapper around F<ev.h> that you can include instead and
		3090	where you can put other configuration options):
		3091
		3092	#define EV_STANDALONE 1
		3093	#include "ev.h"
		3094
		3095	Both header files and implementation files can be compiled with a C++
		3096	compiler (at least, that's a stated goal, and breakage will be treated
		3097	as a bug).
		3098
		3099	You need the following files in your source tree, or in a directory
		3100	in your include path (e.g. in libev/ when using -Ilibev):
		3101
		3102	ev.h
		3103	ev.c
		3104	ev_vars.h
		3105	ev_wrap.h
		3106
		3107	ev_win32.c required on win32 platforms only
		3108
		3109	ev_select.c only when select backend is enabled (which is enabled by default)
		3110	ev_poll.c only when poll backend is enabled (disabled by default)
		3111	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3112	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3113	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3114
		3115	F<ev.c> includes the backend files directly when enabled, so you only need
		3116	to compile this single file.
		3117
		3118	=head3 LIBEVENT COMPATIBILITY API
		3119
		3120	To include the libevent compatibility API, also include:
		3121
		3122	#include "event.c"
		3123
		3124	in the file including F<ev.c>, and:
		3125
		3126	#include "event.h"
		3127
		3128	in the files that want to use the libevent API. This also includes F<ev.h>.
		3129
		3130	You need the following additional files for this:
		3131
		3132	event.h
		3133	event.c
		3134
		3135	=head3 AUTOCONF SUPPORT
		3136
		3137	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3138	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3139	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3140	include F<config.h> and configure itself accordingly.
		3141
		3142	For this of course you need the m4 file:
		3143
		3144	libev.m4
		3145
		3146	=head2 PREPROCESSOR SYMBOLS/MACROS
		3147
		3148	Libev can be configured via a variety of preprocessor symbols you have to
		3149	define before including any of its files. The default in the absence of
		3150	autoconf is documented for every option.
		3151
		3152	=over 4
		3153
		3154	=item EV_STANDALONE
		3155
		3156	Must always be C<1> if you do not use autoconf configuration, which
		3157	keeps libev from including F<config.h>, and it also defines dummy
		3158	implementations for some libevent functions (such as logging, which is not
		3159	supported). It will also not define any of the structs usually found in
		3160	F<event.h> that are not directly supported by the libev core alone.
		3161
		3162	=item EV_USE_MONOTONIC
		3163
		3164	If defined to be C<1>, libev will try to detect the availability of the
		3165	monotonic clock option at both compile time and runtime. Otherwise no use
		3166	of the monotonic clock option will be attempted. If you enable this, you
		3167	usually have to link against librt or something similar. Enabling it when
		3168	the functionality isn't available is safe, though, although you have
		3169	to make sure you link against any libraries where the C<clock_gettime>
		3170	function is hiding in (often F<-lrt>).
		3171
		3172	=item EV_USE_REALTIME
		3173
		3174	If defined to be C<1>, libev will try to detect the availability of the
		3175	real-time clock option at compile time (and assume its availability at
		3176	runtime if successful). Otherwise no use of the real-time clock option will
		3177	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
		3178	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
		3179	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
		3180
		3181	=item EV_USE_NANOSLEEP
		3182
		3183	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3184	and will use it for delays. Otherwise it will use C<select ()>.
		3185
		3186	=item EV_USE_EVENTFD
		3187
		3188	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3189	available and will probe for kernel support at runtime. This will improve
		3190	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3191	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3192	2.7 or newer, otherwise disabled.
		3193
		3194	=item EV_USE_SELECT
		3195
		3196	If undefined or defined to be C<1>, libev will compile in support for the
		3197	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3198	other method takes over, select will be it. Otherwise the select backend
		3199	will not be compiled in.
		3200
		3201	=item EV_SELECT_USE_FD_SET
		3202
		3203	If defined to C<1>, then the select backend will use the system C<fd_set>
		3204	structure. This is useful if libev doesn't compile due to a missing
		3205	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on
		3206	exotic systems. This usually limits the range of file descriptors to some
		3207	low limit such as 1024 or might have other limitations (winsocket only
		3208	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
		3209	influence the size of the C<fd_set> used.
		3210
		3211	=item EV_SELECT_IS_WINSOCKET
		3212
		3213	When defined to C<1>, the select backend will assume that
		3214	select/socket/connect etc. don't understand file descriptors but
		3215	wants osf handles on win32 (this is the case when the select to
		3216	be used is the winsock select). This means that it will call
		3217	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3218	it is assumed that all these functions actually work on fds, even
		3219	on win32. Should not be defined on non-win32 platforms.
		3220
		3221	=item EV_FD_TO_WIN32_HANDLE
		3222
		3223	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3224	file descriptors to socket handles. When not defining this symbol (the
		3225	default), then libev will call C<_get_osfhandle>, which is usually
		3226	correct. In some cases, programs use their own file descriptor management,
		3227	in which case they can provide this function to map fds to socket handles.
		3228
		3229	=item EV_USE_POLL
		3230
		3231	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3232	backend. Otherwise it will be enabled on non-win32 platforms. It
		3233	takes precedence over select.
		3234
		3235	=item EV_USE_EPOLL
		3236
		3237	If defined to be C<1>, libev will compile in support for the Linux
		3238	C<epoll>(7) backend. Its availability will be detected at runtime,
		3239	otherwise another method will be used as fallback. This is the preferred
		3240	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3241	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3242
		3243	=item EV_USE_KQUEUE
		3244
		3245	If defined to be C<1>, libev will compile in support for the BSD style
		3246	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3247	otherwise another method will be used as fallback. This is the preferred
		3248	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3249	supports some types of fds correctly (the only platform we found that
		3250	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3251	not be used unless explicitly requested. The best way to use it is to find
		3252	out whether kqueue supports your type of fd properly and use an embedded
		3253	kqueue loop.
		3254
		3255	=item EV_USE_PORT
		3256
		3257	If defined to be C<1>, libev will compile in support for the Solaris
		3258	10 port style backend. Its availability will be detected at runtime,
		3259	otherwise another method will be used as fallback. This is the preferred
		3260	backend for Solaris 10 systems.
		3261
		3262	=item EV_USE_DEVPOLL
		3263
		3264	Reserved for future expansion, works like the USE symbols above.
		3265
		3266	=item EV_USE_INOTIFY
		3267
		3268	If defined to be C<1>, libev will compile in support for the Linux inotify
		3269	interface to speed up C<ev_stat> watchers. Its actual availability will
		3270	be detected at runtime. If undefined, it will be enabled if the headers
		3271	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3272
		3273	=item EV_ATOMIC_T
		3274
		3275	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3276	access is atomic with respect to other threads or signal contexts. No such
		3277	type is easily found in the C language, so you can provide your own type
		3278	that you know is safe for your purposes. It is used both for signal handler "locking"
		3279	as well as for signal and thread safety in C<ev_async> watchers.
		3280
		3281	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3282	(from F<signal.h>), which is usually good enough on most platforms.
		3283
		3284	=item EV_H
		3285
		3286	The name of the F<ev.h> header file used to include it. The default if
		3287	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3288	used to virtually rename the F<ev.h> header file in case of conflicts.
		3289
		3290	=item EV_CONFIG_H
		3291
		3292	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3293	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3294	C<EV_H>, above.
		3295
		3296	=item EV_EVENT_H
		3297
		3298	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3299	of how the F<event.h> header can be found, the default is C<"event.h">.
		3300
		3301	=item EV_PROTOTYPES
		3302
		3303	If defined to be C<0>, then F<ev.h> will not define any function
		3304	prototypes, but still define all the structs and other symbols. This is
		3305	occasionally useful if you want to provide your own wrapper functions
		3306	around libev functions.
		3307
		3308	=item EV_MULTIPLICITY
		3309
		3310	If undefined or defined to C<1>, then all event-loop-specific functions
		3311	will have the C<struct ev_loop *> as first argument, and you can create
		3312	additional independent event loops. Otherwise there will be no support
		3313	for multiple event loops and there is no first event loop pointer
		3314	argument. Instead, all functions act on the single default loop.
		3315
		3316	=item EV_MINPRI
		3317
		3318	=item EV_MAXPRI
		3319
		3320	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3321	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3322	provide for more priorities by overriding those symbols (usually defined
		3323	to be C<-2> and C<2>, respectively).
		3324
		3325	When doing priority-based operations, libev usually has to linearly search
		3326	all the priorities, so having many of them (hundreds) uses a lot of space
		3327	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3328	fine.
		3329
		3330	If your embedding application does not need any priorities, defining these
		3331	both to C<0> will save some memory and CPU.
		3332
		3333	=item EV_PERIODIC_ENABLE
		3334
		3335	If undefined or defined to be C<1>, then periodic timers are supported. If
		3336	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3337	code.
		3338
		3339	=item EV_IDLE_ENABLE
		3340
		3341	If undefined or defined to be C<1>, then idle watchers are supported. If
		3342	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3343	code.
		3344
		3345	=item EV_EMBED_ENABLE
		3346
		3347	If undefined or defined to be C<1>, then embed watchers are supported. If
		3348	defined to be C<0>, then they are not. Embed watchers rely on most other
		3349	watcher types, which therefore must not be disabled.
		3350
		3351	=item EV_STAT_ENABLE
		3352
		3353	If undefined or defined to be C<1>, then stat watchers are supported. If
		3354	defined to be C<0>, then they are not.
		3355
		3356	=item EV_FORK_ENABLE
		3357
		3358	If undefined or defined to be C<1>, then fork watchers are supported. If
		3359	defined to be C<0>, then they are not.
		3360
		3361	=item EV_ASYNC_ENABLE
		3362
		3363	If undefined or defined to be C<1>, then async watchers are supported. If
		3364	defined to be C<0>, then they are not.
		3365
		3366	=item EV_MINIMAL
		3367
		3368	If you need to shave off some kilobytes of code at the expense of some
		3369	speed, define this symbol to C<1>. Currently this is used to override some
		3370	inlining decisions, saves roughly 30% code size on amd64. It also selects a
		3371	much smaller 2-heap for timer management over the default 4-heap.
		3372
		3373	=item EV_PID_HASHSIZE
		3374
		3375	C<ev_child> watchers use a small hash table to distribute workload by
		3376	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3377	than enough. If you need to manage thousands of children you might want to
		3378	increase this value (I<must> be a power of two).
		3379
		3380	=item EV_INOTIFY_HASHSIZE
		3381
		3382	C<ev_stat> watchers use a small hash table to distribute workload by
		3383	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3384	usually more than enough. If you need to manage thousands of C<ev_stat>
		3385	watchers you might want to increase this value (I<must> be a power of
		3386	two).
		3387
		3388	=item EV_USE_4HEAP
		3389
		3390	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3391	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3392	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3393	faster performance with many (thousands) of watchers.
		3394
		3395	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3396	(disabled).
		3397
		3398	=item EV_HEAP_CACHE_AT
		3399
		3400	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3401	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3402	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3403	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3404	but avoids random read accesses on heap changes. This improves performance
		3405	noticeably with many (hundreds) of watchers.
		3406
		3407	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3408	(disabled).
		3409
		3410	=item EV_VERIFY
		3411
		3412	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3413	be done: If set to C<0>, no internal verification code will be compiled
		3414	in. If set to C<1>, then verification code will be compiled in, but not
		3415	called. If set to C<2>, then the internal verification code will be
		3416	called once per loop, which can slow down libev. If set to C<3>, then the
		3417	verification code will be called very frequently, which will slow down
		3418	libev considerably.
		3419
		3420	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3421	C<0>.
		3422
		3423	=item EV_COMMON
		3424
		3425	By default, all watchers have a C<void *data> member. By redefining
		3426	this macro to a something else you can include more and other types of
		3427	members. You have to define it each time you include one of the files,
		3428	though, and it must be identical each time.
		3429
		3430	For example, the perl EV module uses something like this:
		3431
		3432	#define EV_COMMON \
		3433	SV *self; /* contains this struct */ \
		3434	SV *cb_sv, *fh /* note no trailing ";" */
		3435
		3436	=item EV_CB_DECLARE (type)
		3437
		3438	=item EV_CB_INVOKE (watcher, revents)
		3439
		3440	=item ev_set_cb (ev, cb)
		3441
		3442	Can be used to change the callback member declaration in each watcher,
		3443	and the way callbacks are invoked and set. Must expand to a struct member
		3444	definition and a statement, respectively. See the F<ev.h> header file for
		3445	their default definitions. One possible use for overriding these is to
		3446	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3447	method calls instead of plain function calls in C++.
		3448
		3449	=back
		3450
		3451	=head2 EXPORTED API SYMBOLS
		3452
		3453	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3454	exported symbols, you can use the provided F<Symbol.*> files which list
		3455	all public symbols, one per line:
		3456
		3457	Symbols.ev for libev proper
		3458	Symbols.event for the libevent emulation
		3459
		3460	This can also be used to rename all public symbols to avoid clashes with
		3461	multiple versions of libev linked together (which is obviously bad in
		3462	itself, but sometimes it is inconvenient to avoid this).
		3463
		3464	A sed command like this will create wrapper C<#define>'s that you need to
		3465	include before including F<ev.h>:
		3466
		3467	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3468
		3469	This would create a file F<wrap.h> which essentially looks like this:
		3470
		3471	#define ev_backend myprefix_ev_backend
		3472	#define ev_check_start myprefix_ev_check_start
		3473	#define ev_check_stop myprefix_ev_check_stop
		3474	...
		3475
		3476	=head2 EXAMPLES
		3477
		3478	For a real-world example of a program the includes libev
		3479	verbatim, you can have a look at the EV perl module
		3480	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3481	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3482	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3483	will be compiled. It is pretty complex because it provides its own header
		3484	file.
		3485
		3486	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3487	that everybody includes and which overrides some configure choices:
		3488
		3489	#define EV_MINIMAL 1
		3490	#define EV_USE_POLL 0
		3491	#define EV_MULTIPLICITY 0
		3492	#define EV_PERIODIC_ENABLE 0
		3493	#define EV_STAT_ENABLE 0
		3494	#define EV_FORK_ENABLE 0
		3495	#define EV_CONFIG_H <config.h>
		3496	#define EV_MINPRI 0
		3497	#define EV_MAXPRI 0
		3498
		3499	#include "ev++.h"
		3500
		3501	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3502
		3503	#include "ev_cpp.h"
		3504	#include "ev.c"
		3505
		3506	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3507
		3508	=head2 THREADS AND COROUTINES
		3509
		3510	=head3 THREADS
		3511
		3512	All libev functions are reentrant and thread-safe unless explicitly
		3513	documented otherwise, but libev implements no locking itself. This means
		3514	that you can use as many loops as you want in parallel, as long as there
		3515	are no concurrent calls into any libev function with the same loop
		3516	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3517	of course): libev guarantees that different event loops share no data
		3518	structures that need any locking.
		3519
		3520	Or to put it differently: calls with different loop parameters can be done
		3521	concurrently from multiple threads, calls with the same loop parameter
		3522	must be done serially (but can be done from different threads, as long as
		3523	only one thread ever is inside a call at any point in time, e.g. by using
		3524	a mutex per loop).
		3525
		3526	Specifically to support threads (and signal handlers), libev implements
		3527	so-called C<ev_async> watchers, which allow some limited form of
		3528	concurrency on the same event loop, namely waking it up "from the
		3529	outside".
		3530
		3531	If you want to know which design (one loop, locking, or multiple loops
		3532	without or something else still) is best for your problem, then I cannot
		3533	help you, but here is some generic advice:
		3534
		3535	=over 4
		3536
		3537	=item * most applications have a main thread: use the default libev loop
		3538	in that thread, or create a separate thread running only the default loop.
		3539
		3540	This helps integrating other libraries or software modules that use libev
		3541	themselves and don't care/know about threading.
		3542
		3543	=item * one loop per thread is usually a good model.
		3544
		3545	Doing this is almost never wrong, sometimes a better-performance model
		3546	exists, but it is always a good start.
		3547
		3548	=item * other models exist, such as the leader/follower pattern, where one
		3549	loop is handed through multiple threads in a kind of round-robin fashion.
		3550
		3551	Choosing a model is hard - look around, learn, know that usually you can do
		3552	better than you currently do :-)
		3553
		3554	=item * often you need to talk to some other thread which blocks in the
		3555	event loop.
		3556
		3557	C<ev_async> watchers can be used to wake them up from other threads safely
		3558	(or from signal contexts...).
		3559
		3560	An example use would be to communicate signals or other events that only
		3561	work in the default loop by registering the signal watcher with the
		3562	default loop and triggering an C<ev_async> watcher from the default loop
		3563	watcher callback into the event loop interested in the signal.
		3564
		3565	=back
		3566
		3567	=head3 COROUTINES
		3568
		3569	Libev is very accommodating to coroutines ("cooperative threads"):
		3570	libev fully supports nesting calls to its functions from different
		3571	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3572	different coroutines, and switch freely between both coroutines running the
		3573	loop, as long as you don't confuse yourself). The only exception is that
		3574	you must not do this from C<ev_periodic> reschedule callbacks.
		3575
		3576	Care has been taken to ensure that libev does not keep local state inside
		3577	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3578	they do not call any callbacks.
		3579
		3580	=head2 COMPILER WARNINGS
		3581
		3582	Depending on your compiler and compiler settings, you might get no or a
		3583	lot of warnings when compiling libev code. Some people are apparently
		3584	scared by this.
		3585
		3586	However, these are unavoidable for many reasons. For one, each compiler
		3587	has different warnings, and each user has different tastes regarding
		3588	warning options. "Warn-free" code therefore cannot be a goal except when
		3589	targeting a specific compiler and compiler-version.
		3590
		3591	Another reason is that some compiler warnings require elaborate
		3592	workarounds, or other changes to the code that make it less clear and less
		3593	maintainable.
		3594
		3595	And of course, some compiler warnings are just plain stupid, or simply
		3596	wrong (because they don't actually warn about the condition their message
		3597	seems to warn about). For example, certain older gcc versions had some
		3598	warnings that resulted an extreme number of false positives. These have
		3599	been fixed, but some people still insist on making code warn-free with
		3600	such buggy versions.
		3601
		3602	While libev is written to generate as few warnings as possible,
		3603	"warn-free" code is not a goal, and it is recommended not to build libev
		3604	with any compiler warnings enabled unless you are prepared to cope with
		3605	them (e.g. by ignoring them). Remember that warnings are just that:
		3606	warnings, not errors, or proof of bugs.
		3607
		3608
		3609	=head2 VALGRIND
		3610
		3611	Valgrind has a special section here because it is a popular tool that is
		3612	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		3613
		3614	If you think you found a bug (memory leak, uninitialised data access etc.)
		3615	in libev, then check twice: If valgrind reports something like:
		3616
		3617	==2274== definitely lost: 0 bytes in 0 blocks.
		3618	==2274== possibly lost: 0 bytes in 0 blocks.
		3619	==2274== still reachable: 256 bytes in 1 blocks.
		3620
		3621	Then there is no memory leak, just as memory accounted to global variables
		3622	is not a memleak - the memory is still being referenced, and didn't leak.
		3623
		3624	Similarly, under some circumstances, valgrind might report kernel bugs
		3625	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3626	although an acceptable workaround has been found here), or it might be
		3627	confused.
		3628
		3629	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		3630	make it into some kind of religion.
		3631
		3632	If you are unsure about something, feel free to contact the mailing list
		3633	with the full valgrind report and an explanation on why you think this
		3634	is a bug in libev (best check the archives, too :). However, don't be
		3635	annoyed when you get a brisk "this is no bug" answer and take the chance
		3636	of learning how to interpret valgrind properly.
		3637
		3638	If you need, for some reason, empty reports from valgrind for your project
		3639	I suggest using suppression lists.
		3640
		3641
		3642	=head1 PORTABILITY NOTES
		3643
		3644	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		3645
		3646	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3647	requires, and its I/O model is fundamentally incompatible with the POSIX
		3648	model. Libev still offers limited functionality on this platform in
		3649	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3650	descriptors. This only applies when using Win32 natively, not when using
		3651	e.g. cygwin.
		3652
		3653	Lifting these limitations would basically require the full
		3654	re-implementation of the I/O system. If you are into these kinds of
		3655	things, then note that glib does exactly that for you in a very portable
		3656	way (note also that glib is the slowest event library known to man).
		3657
		3658	There is no supported compilation method available on windows except
		3659	embedding it into other applications.
		3660
		3661	Not a libev limitation but worth mentioning: windows apparently doesn't
		3662	accept large writes: instead of resulting in a partial write, windows will
		3663	either accept everything or return C<ENOBUFS> if the buffer is too large,
		3664	so make sure you only write small amounts into your sockets (less than a
		3665	megabyte seems safe, but this apparently depends on the amount of memory
		3666	available).
		3667
		3668	Due to the many, low, and arbitrary limits on the win32 platform and
		3669	the abysmal performance of winsockets, using a large number of sockets
		3670	is not recommended (and not reasonable). If your program needs to use
		3671	more than a hundred or so sockets, then likely it needs to use a totally
		3672	different implementation for windows, as libev offers the POSIX readiness
		3673	notification model, which cannot be implemented efficiently on windows
		3674	(Microsoft monopoly games).
		3675
		3676	A typical way to use libev under windows is to embed it (see the embedding
		3677	section for details) and use the following F<evwrap.h> header file instead
		3678	of F<ev.h>:
		3679
		3680	#define EV_STANDALONE /* keeps ev from requiring config.h */
		3681	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		3682
		3683	#include "ev.h"
		3684
		3685	And compile the following F<evwrap.c> file into your project (make sure
		3686	you do I<not> compile the F<ev.c> or any other embedded source files!):
		3687
		3688	#include "evwrap.h"
		3689	#include "ev.c"
		3690
		3691	=over 4
		3692
		3693	=item The winsocket select function
		3694
		3695	The winsocket C<select> function doesn't follow POSIX in that it
		3696	requires socket I<handles> and not socket I<file descriptors> (it is
		3697	also extremely buggy). This makes select very inefficient, and also
		3698	requires a mapping from file descriptors to socket handles (the Microsoft
		3699	C runtime provides the function C<_open_osfhandle> for this). See the
		3700	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		3701	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		3702
		3703	The configuration for a "naked" win32 using the Microsoft runtime
		3704	libraries and raw winsocket select is:
		3705
		3706	#define EV_USE_SELECT 1
		3707	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3708
		3709	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3710	complexity in the O(n²) range when using win32.
		3711
		3712	=item Limited number of file descriptors
		3713
		3714	Windows has numerous arbitrary (and low) limits on things.
		3715
		3716	Early versions of winsocket's select only supported waiting for a maximum
		3717	of C<64> handles (probably owning to the fact that all windows kernels
		3718	can only wait for C<64> things at the same time internally; Microsoft
		3719	recommends spawning a chain of threads and wait for 63 handles and the
		3720	previous thread in each. Great).
		3721
		3722	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3723	to some high number (e.g. C<2048>) before compiling the winsocket select
		3724	call (which might be in libev or elsewhere, for example, perl does its own
		3725	select emulation on windows).
		3726
		3727	Another limit is the number of file descriptors in the Microsoft runtime
		3728	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3729	or something like this inside Microsoft). You can increase this by calling
		3730	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3731	arbitrary limit), but is broken in many versions of the Microsoft runtime
		3732	libraries.
		3733
		3734	This might get you to about C<512> or C<2048> sockets (depending on
		3735	windows version and/or the phase of the moon). To get more, you need to
		3736	wrap all I/O functions and provide your own fd management, but the cost of
		3737	calling select (O(n²)) will likely make this unworkable.
		3738
		3739	=back
		3740
		3741	=head2 PORTABILITY REQUIREMENTS
		3742
		3743	In addition to a working ISO-C implementation and of course the
		3744	backend-specific APIs, libev relies on a few additional extensions:
		3745
		3746	=over 4
		3747
		3748	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		3749	calling conventions regardless of C<ev_watcher_type *>.
		3750
		3751	Libev assumes not only that all watcher pointers have the same internal
		3752	structure (guaranteed by POSIX but not by ISO C for example), but it also
		3753	assumes that the same (machine) code can be used to call any watcher
		3754	callback: The watcher callbacks have different type signatures, but libev
		3755	calls them using an C<ev_watcher *> internally.
		3756
		3757	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3758
		3759	The type C<sig_atomic_t volatile> (or whatever is defined as
		3760	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		3761	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3762	believed to be sufficiently portable.
		3763
		3764	=item C<sigprocmask> must work in a threaded environment
		3765
		3766	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3767	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3768	pthread implementations will either allow C<sigprocmask> in the "main
		3769	thread" or will block signals process-wide, both behaviours would
		3770	be compatible with libev. Interaction between C<sigprocmask> and
		3771	C<pthread_sigmask> could complicate things, however.
		3772
		3773	The most portable way to handle signals is to block signals in all threads
		3774	except the initial one, and run the default loop in the initial thread as
		3775	well.
		3776
		3777	=item C<long> must be large enough for common memory allocation sizes
		3778
		3779	To improve portability and simplify its API, libev uses C<long> internally
		3780	instead of C<size_t> when allocating its data structures. On non-POSIX
		3781	systems (Microsoft...) this might be unexpectedly low, but is still at
		3782	least 31 bits everywhere, which is enough for hundreds of millions of
		3783	watchers.
		3784
		3785	=item C<double> must hold a time value in seconds with enough accuracy
		3786
		3787	The type C<double> is used to represent timestamps. It is required to
		3788	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3789	enough for at least into the year 4000. This requirement is fulfilled by
		3790	implementations implementing IEEE 754 (basically all existing ones).
		3791
		3792	=back
		3793
		3794	If you know of other additional requirements drop me a note.
		3795
		3796
		3797	=head1 ALGORITHMIC COMPLEXITIES
		3798
		3799	In this section the complexities of (many of) the algorithms used inside
		3800	libev will be documented. For complexity discussions about backends see
		3801	the documentation for C<ev_default_init>.
		3802
		3803	All of the following are about amortised time: If an array needs to be
		3804	extended, libev needs to realloc and move the whole array, but this
		3805	happens asymptotically rarer with higher number of elements, so O(1) might
		3806	mean that libev does a lengthy realloc operation in rare cases, but on
		3807	average it is much faster and asymptotically approaches constant time.
		3808
		3809	=over 4
		3810
		3811	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		3812
		3813	This means that, when you have a watcher that triggers in one hour and
		3814	there are 100 watchers that would trigger before that, then inserting will
		3815	have to skip roughly seven (C<ld 100>) of these watchers.
		3816
		3817	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		3818
		3819	That means that changing a timer costs less than removing/adding them,
		3820	as only the relative motion in the event queue has to be paid for.
		3821
		3822	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		3823
		3824	These just add the watcher into an array or at the head of a list.
		3825
		3826	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		3827
		3828	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		3829
		3830	These watchers are stored in lists, so they need to be walked to find the
		3831	correct watcher to remove. The lists are usually short (you don't usually
		3832	have many watchers waiting for the same fd or signal: one is typical, two
		3833	is rare).
		3834
		3835	=item Finding the next timer in each loop iteration: O(1)
		3836
		3837	By virtue of using a binary or 4-heap, the next timer is always found at a
		3838	fixed position in the storage array.
		3839
		3840	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3841
		3842	A change means an I/O watcher gets started or stopped, which requires
		3843	libev to recalculate its status (and possibly tell the kernel, depending
		3844	on backend and whether C<ev_io_set> was used).
		3845
		3846	=item Activating one watcher (putting it into the pending state): O(1)
		3847
		3848	=item Priority handling: O(number_of_priorities)
		3849
		3850	Priorities are implemented by allocating some space for each
		3851	priority. When doing priority-based operations, libev usually has to
		3852	linearly search all the priorities, but starting/stopping and activating
		3853	watchers becomes O(1) with respect to priority handling.
		3854
		3855	=item Sending an ev_async: O(1)
		3856
		3857	=item Processing ev_async_send: O(number_of_async_watchers)
		3858
		3859	=item Processing signals: O(max_signal_number)
		3860
		3861	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3862	calls in the current loop iteration. Checking for async and signal events
		3863	involves iterating over all running async watchers or all signal numbers.
		3864
		3865	=back
		3866
		3867
767	=head1 AUTHOR	3868	=head1 AUTHOR
768		3869
769	Marc Lehmann <libev@schmorp.de>.	3870	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
770		3871

Diff Legend

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