ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.27 by root, Wed Nov 14 05:02:07 2007 UTC vs.
Revision 1.216 by root, Thu Nov 13 15:55:38 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 REPRESENTATION	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
51		142
52	=head1 GLOBAL FUNCTIONS	143	=head1 GLOBAL FUNCTIONS
53		144
54	These functions can be called anytime, even before initialising the	145	These functions can be called anytime, even before initialising the
55	library in any way.	146	library in any way.
…		…
60		151
61	Returns the current time as libev would use it. Please note that the	152	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	153	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	154	you actually want to know.
64		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 ()>.
		161
65	=item int ev_version_major ()	162	=item int ev_version_major ()
66		163
67	=item int ev_version_minor ()	164	=item int ev_version_minor ()
68		165
69	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
70	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
71	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
72	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
73	version of the library your program was compiled against.	170	version of the library your program was compiled against.
74		171
		172	These version numbers refer to the ABI version of the library, not the
		173	release version.
		174
75	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,
76	as this indicates an incompatible change. Minor versions are usually	176	as this indicates an incompatible change. Minor versions are usually
77	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
78	not a problem.	178	not a problem.
79		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
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	219	=item ev_set_allocator (void *(*cb)(void *ptr, long size)) [NOT REENTRANT]
81		220
82	Sets the allocation function to use (the prototype is similar to the	221	Sets the allocation function to use (the prototype is similar - the
83	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
84	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
85	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
86	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.
87		230
88	You could override this function in high-availability programs to, say,	231	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	232	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	233	or even to sleep a while and retry until some memory is available.
91		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
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	255	=item ev_set_syserr_cb (void (*cb)(const char *msg)); [NOT REENTRANT]
93		256
94	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
95	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
96	indicating the system call or subsystem causing the problem. If this	259	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	260	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	261	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	262	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	263	(such as abort).
101		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
102	=back	277	=back
103		278
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
105		280
106	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>
107	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>
108	events, and dynamically created loops which do not.	283	I<function>).
109		284
110	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
111	in your main thread (or in a separate thread) and for each thread you	286	supports signals and child events, and dynamically created loops which do
112	create, you also create another event loop. Libev itself does no locking	287	not.
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		288
117	=over 4	289	=over 4
118		290
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	291	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		292
121	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
122	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
123	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
124	flags).	296	flags. If that is troubling you, check C<ev_backend ()> afterwards).
125		297
126	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
127	function.	299	function.
128		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
129	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
130	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>).
131		314
132	It supports the following flags:	315	The following flags are supported:
133		316
134	=over 4	317	=over 4
135		318
136	=item C<EVFLAG_AUTO>	319	=item C<EVFLAG_AUTO>
137		320
138	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
139	thing, believe me).	322	thing, believe me).
140		323
141	=item C<EVFLAG_NOENV>	324	=item C<EVFLAG_NOENV>
142		325
143	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
144	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
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	328	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	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
147	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
148	around bugs.	331	around bugs.
149		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
150	=item C<EVMETHOD_SELECT> (portable select backend)	353	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		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
152	=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)
153		373
154	=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.
155		380
156	=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>.
157		383
158	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	384	=item C<EVBACKEND_EPOLL> (value 4, Linux)
159		385
160	=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).
161		390
162	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
163	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
164	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	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		424	faster than epoll for maybe up to a hundred file descriptors, depending on
		425	the usage. So sad.
		426
		427	While nominally embeddable in other event loops, this feature is broken in
		428	all kernel versions tested so far.
		429
		430	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		431	C<EVBACKEND_POLL>.
		432
		433	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		434
		435	Kqueue deserves special mention, as at the time of this writing, it
		436	was broken on all BSDs except NetBSD (usually it doesn't work reliably
		437	with anything but sockets and pipes, except on Darwin, where of course
		438	it's completely useless). Unlike epoll, however, whose brokenness
		439	is by design, these kqueue bugs can (and eventually will) be fixed
		440	without API changes to existing programs. For this reason it's not being
		441	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		442	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		443	system like NetBSD.
		444
		445	You still can embed kqueue into a normal poll or select backend and use it
		446	only for sockets (after having made sure that sockets work with kqueue on
		447	the target platform). See C<ev_embed> watchers for more info.
		448
		449	It scales in the same way as the epoll backend, but the interface to the
		450	kernel is more efficient (which says nothing about its actual speed, of
		451	course). While stopping, setting and starting an I/O watcher does never
		452	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		453	two event changes per incident. Support for C<fork ()> is very bad (but
		454	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		455	cases
		456
		457	This backend usually performs well under most conditions.
		458
		459	While nominally embeddable in other event loops, this doesn't work
		460	everywhere, so you might need to test for this. And since it is broken
		461	almost everywhere, you should only use it when you have a lot of sockets
		462	(for which it usually works), by embedding it into another event loop
		463	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,
		464	using it only for sockets.
		465
		466	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		467	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		468	C<NOTE_EOF>.
		469
		470	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		471
		472	This is not implemented yet (and might never be, unless you send me an
		473	implementation). According to reports, C</dev/poll> only supports sockets
		474	and is not embeddable, which would limit the usefulness of this backend
		475	immensely.
		476
		477	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		478
		479	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		480	it's really slow, but it still scales very well (O(active_fds)).
		481
		482	Please note that Solaris event ports can deliver a lot of spurious
		483	notifications, so you need to use non-blocking I/O or other means to avoid
		484	blocking when no data (or space) is available.
		485
		486	While this backend scales well, it requires one system call per active
		487	file descriptor per loop iteration. For small and medium numbers of file
		488	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		489	might perform better.
		490
		491	On the positive side, with the exception of the spurious readiness
		492	notifications, this backend actually performed fully to specification
		493	in all tests and is fully embeddable, which is a rare feat among the
		494	OS-specific backends (I vastly prefer correctness over speed hacks).
		495
		496	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		497	C<EVBACKEND_POLL>.
		498
		499	=item C<EVBACKEND_ALL>
		500
		501	Try all backends (even potentially broken ones that wouldn't be tried
		502	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		503	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		504
		505	It is definitely not recommended to use this flag.
165		506
166	=back	507	=back
		508
		509	If one or more of these are or'ed into the flags value, then only these
		510	backends will be tried (in the reverse order as listed here). If none are
		511	specified, all backends in C<ev_recommended_backends ()> will be tried.
		512
		513	Example: This is the most typical usage.
		514
		515	if (!ev_default_loop (0))
		516	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		517
		518	Example: Restrict libev to the select and poll backends, and do not allow
		519	environment settings to be taken into account:
		520
		521	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		522
		523	Example: Use whatever libev has to offer, but make sure that kqueue is
		524	used if available (warning, breaks stuff, best use only with your own
		525	private event loop and only if you know the OS supports your types of
		526	fds):
		527
		528	ev_default_loop (ev_recommended_backends () | EVBACKEND_KQUEUE);
167		529
168	=item struct ev_loop *ev_loop_new (unsigned int flags)	530	=item struct ev_loop *ev_loop_new (unsigned int flags)
169		531
170	Similar to C<ev_default_loop>, but always creates a new event loop that is	532	Similar to C<ev_default_loop>, but always creates a new event loop that is
171	always distinct from the default loop. Unlike the default loop, it cannot	533	always distinct from the default loop. Unlike the default loop, it cannot
172	handle signal and child watchers, and attempts to do so will be greeted by	534	handle signal and child watchers, and attempts to do so will be greeted by
173	undefined behaviour (or a failed assertion if assertions are enabled).	535	undefined behaviour (or a failed assertion if assertions are enabled).
174		536
		537	Note that this function I<is> thread-safe, and the recommended way to use
		538	libev with threads is indeed to create one loop per thread, and using the
		539	default loop in the "main" or "initial" thread.
		540
		541	Example: Try to create a event loop that uses epoll and nothing else.
		542
		543	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		544	if (!epoller)
		545	fatal ("no epoll found here, maybe it hides under your chair");
		546
175	=item ev_default_destroy ()	547	=item ev_default_destroy ()
176		548
177	Destroys the default loop again (frees all memory and kernel state	549	Destroys the default loop again (frees all memory and kernel state
178	etc.). This stops all registered event watchers (by not touching them in	550	etc.). None of the active event watchers will be stopped in the normal
179	any way whatsoever, although you cannot rely on this :).	551	sense, so e.g. C<ev_is_active> might still return true. It is your
		552	responsibility to either stop all watchers cleanly yourself I<before>
		553	calling this function, or cope with the fact afterwards (which is usually
		554	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		555	for example).
		556
		557	Note that certain global state, such as signal state (and installed signal
		558	handlers), will not be freed by this function, and related watchers (such
		559	as signal and child watchers) would need to be stopped manually.
		560
		561	In general it is not advisable to call this function except in the
		562	rare occasion where you really need to free e.g. the signal handling
		563	pipe fds. If you need dynamically allocated loops it is better to use
		564	C<ev_loop_new> and C<ev_loop_destroy>).
180		565
181	=item ev_loop_destroy (loop)	566	=item ev_loop_destroy (loop)
182		567
183	Like C<ev_default_destroy>, but destroys an event loop created by an	568	Like C<ev_default_destroy>, but destroys an event loop created by an
184	earlier call to C<ev_loop_new>.	569	earlier call to C<ev_loop_new>.
185		570
186	=item ev_default_fork ()	571	=item ev_default_fork ()
187		572
		573	This function sets a flag that causes subsequent C<ev_loop> iterations
188	This function reinitialises the kernel state for backends that have	574	to reinitialise the kernel state for backends that have one. Despite the
189	one. Despite the name, you can call it anytime, but it makes most sense	575	name, you can call it anytime, but it makes most sense after forking, in
190	after forking, in either the parent or child process (or both, but that	576	the child process (or both child and parent, but that again makes little
191	again makes little sense).	577	sense). You I<must> call it in the child before using any of the libev
		578	functions, and it will only take effect at the next C<ev_loop> iteration.
192		579
193	You I<must> call this function after forking if and only if you want to	580	On the other hand, you only need to call this function in the child
194	use the event library in both processes. If you just fork+exec, you don't	581	process if and only if you want to use the event library in the child. If
195	have to call it.	582	you just fork+exec, you don't have to call it at all.
196		583
197	The function itself is quite fast and it's usually not a problem to call	584	The function itself is quite fast and it's usually not a problem to call
198	it just in case after a fork. To make this easy, the function will fit in	585	it just in case after a fork. To make this easy, the function will fit in
199	quite nicely into a call to C<pthread_atfork>:	586	quite nicely into a call to C<pthread_atfork>:
200		587
…		…
202		589
203	=item ev_loop_fork (loop)	590	=item ev_loop_fork (loop)
204		591
205	Like C<ev_default_fork>, but acts on an event loop created by	592	Like C<ev_default_fork>, but acts on an event loop created by
206	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	593	C<ev_loop_new>. Yes, you have to call this on every allocated event loop
207	after fork, and how you do this is entirely your own problem.	594	after fork that you want to re-use in the child, and how you do this is
		595	entirely your own problem.
208		596
		597	=item int ev_is_default_loop (loop)
		598
		599	Returns true when the given loop is, in fact, the default loop, and false
		600	otherwise.
		601
		602	=item unsigned int ev_loop_count (loop)
		603
		604	Returns the count of loop iterations for the loop, which is identical to
		605	the number of times libev did poll for new events. It starts at C<0> and
		606	happily wraps around with enough iterations.
		607
		608	This value can sometimes be useful as a generation counter of sorts (it
		609	"ticks" the number of loop iterations), as it roughly corresponds with
		610	C<ev_prepare> and C<ev_check> calls.
		611
209	=item unsigned int ev_method (loop)	612	=item unsigned int ev_backend (loop)
210		613
211	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	614	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
212	use.	615	use.
213		616
214	=item ev_tstamp ev_now (loop)	617	=item ev_tstamp ev_now (loop)
215		618
216	Returns the current "event loop time", which is the time the event loop	619	Returns the current "event loop time", which is the time the event loop
217	got events and started processing them. This timestamp does not change	620	received events and started processing them. This timestamp does not
218	as long as callbacks are being processed, and this is also the base time	621	change as long as callbacks are being processed, and this is also the base
219	used for relative timers. You can treat it as the timestamp of the event	622	time used for relative timers. You can treat it as the timestamp of the
220	occuring (or more correctly, the mainloop finding out about it).	623	event occurring (or more correctly, libev finding out about it).
		624
		625	=item ev_now_update (loop)
		626
		627	Establishes the current time by querying the kernel, updating the time
		628	returned by C<ev_now ()> in the progress. This is a costly operation and
		629	is usually done automatically within C<ev_loop ()>.
		630
		631	This function is rarely useful, but when some event callback runs for a
		632	very long time without entering the event loop, updating libev's idea of
		633	the current time is a good idea.
		634
		635	See also "The special problem of time updates" in the C<ev_timer> section.
221		636
222	=item ev_loop (loop, int flags)	637	=item ev_loop (loop, int flags)
223		638
224	Finally, this is it, the event handler. This function usually is called	639	Finally, this is it, the event handler. This function usually is called
225	after you initialised all your watchers and you want to start handling	640	after you initialised all your watchers and you want to start handling
226	events.	641	events.
227		642
228	If the flags argument is specified as 0, it will not return until either	643	If the flags argument is specified as C<0>, it will not return until
229	no event watchers are active anymore or C<ev_unloop> was called.	644	either no event watchers are active anymore or C<ev_unloop> was called.
		645
		646	Please note that an explicit C<ev_unloop> is usually better than
		647	relying on all watchers to be stopped when deciding when a program has
		648	finished (especially in interactive programs), but having a program
		649	that automatically loops as long as it has to and no longer by virtue
		650	of relying on its watchers stopping correctly, that is truly a thing of
		651	beauty.
230		652
231	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	653	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle
232	those events and any outstanding ones, but will not block your process in	654	those events and any already outstanding ones, but will not block your
233	case there are no events and will return after one iteration of the loop.	655	process in case there are no events and will return after one iteration of
		656	the loop.
234		657
235	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	658	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
236	neccessary) and will handle those and any outstanding ones. It will block	659	necessary) and will handle those and any already outstanding ones. It
237	your process until at least one new event arrives, and will return after	660	will block your process until at least one new event arrives (which could
		661	be an event internal to libev itself, so there is no guarantee that a
		662	user-registered callback will be called), and will return after one
238	one iteration of the loop.	663	iteration of the loop.
239		664
240	This flags value could be used to implement alternative looping	665	This is useful if you are waiting for some external event in conjunction
241	constructs, but the C<prepare> and C<check> watchers provide a better and	666	with something not expressible using other libev watchers (i.e. "roll your
242	more generic mechanism.	667	own C<ev_loop>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		668	usually a better approach for this kind of thing.
243		669
244	Here are the gory details of what ev_loop does:	670	Here are the gory details of what C<ev_loop> does:
245		671
246	1. If there are no active watchers (reference count is zero), return.	672	- Before the first iteration, call any pending watchers.
		673	* If EVFLAG_FORKCHECK was used, check for a fork.
		674	- If a fork was detected (by any means), queue and call all fork watchers.
247	2. Queue and immediately call all prepare watchers.	675	- Queue and call all prepare watchers.
248	3. If we have been forked, recreate the kernel state.	676	- If we have been forked, detach and recreate the kernel state
		677	as to not disturb the other process.
249	4. Update the kernel state with all outstanding changes.	678	- Update the kernel state with all outstanding changes.
250	5. Update the "event loop time".	679	- Update the "event loop time" (ev_now ()).
251	6. Calculate for how long to block.	680	- Calculate for how long to sleep or block, if at all
		681	(active idle watchers, EVLOOP_NONBLOCK or not having
		682	any active watchers at all will result in not sleeping).
		683	- Sleep if the I/O and timer collect interval say so.
252	7. Block the process, waiting for events.	684	- Block the process, waiting for any events.
		685	- Queue all outstanding I/O (fd) events.
253	8. Update the "event loop time" and do time jump handling.	686	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
254	9. Queue all outstanding timers.	687	- Queue all expired timers.
255	10. Queue all outstanding periodics.	688	- Queue all expired periodics.
256	11. If no events are pending now, queue all idle watchers.	689	- Unless any events are pending now, queue all idle watchers.
257	12. Queue all check watchers.	690	- Queue all check watchers.
258	13. Call all queued watchers in reverse order (i.e. check watchers first).	691	- Call all queued watchers in reverse order (i.e. check watchers first).
		692	Signals and child watchers are implemented as I/O watchers, and will
		693	be handled here by queueing them when their watcher gets executed.
259	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	694	- If ev_unloop has been called, or EVLOOP_ONESHOT or EVLOOP_NONBLOCK
260	was used, return, otherwise continue with step #1.	695	were used, or there are no active watchers, return, otherwise
		696	continue with step *.
		697
		698	Example: Queue some jobs and then loop until no events are outstanding
		699	anymore.
		700
		701	... queue jobs here, make sure they register event watchers as long
		702	... as they still have work to do (even an idle watcher will do..)
		703	ev_loop (my_loop, 0);
		704	... jobs done or somebody called unloop. yeah!
261		705
262	=item ev_unloop (loop, how)	706	=item ev_unloop (loop, how)
263		707
264	Can be used to make a call to C<ev_loop> return early (but only after it	708	Can be used to make a call to C<ev_loop> return early (but only after it
265	has processed all outstanding events). The C<how> argument must be either	709	has processed all outstanding events). The C<how> argument must be either
266	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	710	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
267	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	711	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
268		712
		713	This "unloop state" will be cleared when entering C<ev_loop> again.
		714
		715	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
		716
269	=item ev_ref (loop)	717	=item ev_ref (loop)
270		718
271	=item ev_unref (loop)	719	=item ev_unref (loop)
272		720
273	Ref/unref can be used to add or remove a reference count on the event	721	Ref/unref can be used to add or remove a reference count on the event
274	loop: Every watcher keeps one reference, and as long as the reference	722	loop: Every watcher keeps one reference, and as long as the reference
275	count is nonzero, C<ev_loop> will not return on its own. If you have	723	count is nonzero, C<ev_loop> will not return on its own.
		724
276	a watcher you never unregister that should not keep C<ev_loop> from	725	If you have a watcher you never unregister that should not keep C<ev_loop>
277	returning, ev_unref() after starting, and ev_ref() before stopping it. For	726	from returning, call ev_unref() after starting, and ev_ref() before
		727	stopping it.
		728
278	example, libev itself uses this for its internal signal pipe: It is not	729	As an example, libev itself uses this for its internal signal pipe: It is
279	visible to the libev user and should not keep C<ev_loop> from exiting if	730	not visible to the libev user and should not keep C<ev_loop> from exiting
280	no event watchers registered by it are active. It is also an excellent	731	if no event watchers registered by it are active. It is also an excellent
281	way to do this for generic recurring timers or from within third-party	732	way to do this for generic recurring timers or from within third-party
282	libraries. Just remember to I<unref after start> and I<ref before stop>.	733	libraries. Just remember to I<unref after start> and I<ref before stop>
		734	(but only if the watcher wasn't active before, or was active before,
		735	respectively).
		736
		737	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
		738	running when nothing else is active.
		739
		740	ev_signal exitsig;
		741	ev_signal_init (&exitsig, sig_cb, SIGINT);
		742	ev_signal_start (loop, &exitsig);
		743	evf_unref (loop);
		744
		745	Example: For some weird reason, unregister the above signal handler again.
		746
		747	ev_ref (loop);
		748	ev_signal_stop (loop, &exitsig);
		749
		750	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		751
		752	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		753
		754	These advanced functions influence the time that libev will spend waiting
		755	for events. Both time intervals are by default C<0>, meaning that libev
		756	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		757	latency.
		758
		759	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		760	allows libev to delay invocation of I/O and timer/periodic callbacks
		761	to increase efficiency of loop iterations (or to increase power-saving
		762	opportunities).
		763
		764	The idea is that sometimes your program runs just fast enough to handle
		765	one (or very few) event(s) per loop iteration. While this makes the
		766	program responsive, it also wastes a lot of CPU time to poll for new
		767	events, especially with backends like C<select ()> which have a high
		768	overhead for the actual polling but can deliver many events at once.
		769
		770	By setting a higher I<io collect interval> you allow libev to spend more
		771	time collecting I/O events, so you can handle more events per iteration,
		772	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		773	C<ev_timer>) will be not affected. Setting this to a non-null value will
		774	introduce an additional C<ev_sleep ()> call into most loop iterations.
		775
		776	Likewise, by setting a higher I<timeout collect interval> you allow libev
		777	to spend more time collecting timeouts, at the expense of increased
		778	latency/jitter/inexactness (the watcher callback will be called
		779	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		780	value will not introduce any overhead in libev.
		781
		782	Many (busy) programs can usually benefit by setting the I/O collect
		783	interval to a value near C<0.1> or so, which is often enough for
		784	interactive servers (of course not for games), likewise for timeouts. It
		785	usually doesn't make much sense to set it to a lower value than C<0.01>,
		786	as this approaches the timing granularity of most systems.
		787
		788	Setting the I<timeout collect interval> can improve the opportunity for
		789	saving power, as the program will "bundle" timer callback invocations that
		790	are "near" in time together, by delaying some, thus reducing the number of
		791	times the process sleeps and wakes up again. Another useful technique to
		792	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		793	they fire on, say, one-second boundaries only.
		794
		795	=item ev_loop_verify (loop)
		796
		797	This function only does something when C<EV_VERIFY> support has been
		798	compiled in, which is the default for non-minimal builds. It tries to go
		799	through all internal structures and checks them for validity. If anything
		800	is found to be inconsistent, it will print an error message to standard
		801	error and call C<abort ()>.
		802
		803	This can be used to catch bugs inside libev itself: under normal
		804	circumstances, this function will never abort as of course libev keeps its
		805	data structures consistent.
283		806
284	=back	807	=back
285		808
		809
286	=head1 ANATOMY OF A WATCHER	810	=head1 ANATOMY OF A WATCHER
		811
		812	In the following description, uppercase C<TYPE> in names stands for the
		813	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		814	watchers and C<ev_io_start> for I/O watchers.
287		815
288	A watcher is a structure that you create and register to record your	816	A watcher is a structure that you create and register to record your
289	interest in some event. For instance, if you want to wait for STDIN to	817	interest in some event. For instance, if you want to wait for STDIN to
290	become readable, you would create an C<ev_io> watcher for that:	818	become readable, you would create an C<ev_io> watcher for that:
291		819
292	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	820	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
293	{	821	{
294	ev_io_stop (w);	822	ev_io_stop (w);
295	ev_unloop (loop, EVUNLOOP_ALL);	823	ev_unloop (loop, EVUNLOOP_ALL);
296	}	824	}
297		825
298	struct ev_loop *loop = ev_default_loop (0);	826	struct ev_loop *loop = ev_default_loop (0);
		827
299	struct ev_io stdin_watcher;	828	ev_io stdin_watcher;
		829
300	ev_init (&stdin_watcher, my_cb);	830	ev_init (&stdin_watcher, my_cb);
301	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	831	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
302	ev_io_start (loop, &stdin_watcher);	832	ev_io_start (loop, &stdin_watcher);
		833
303	ev_loop (loop, 0);	834	ev_loop (loop, 0);
304		835
305	As you can see, you are responsible for allocating the memory for your	836	As you can see, you are responsible for allocating the memory for your
306	watcher structures (and it is usually a bad idea to do this on the stack,	837	watcher structures (and it is I<usually> a bad idea to do this on the
307	although this can sometimes be quite valid).	838	stack).
		839
		840	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		841	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
308		842
309	Each watcher structure must be initialised by a call to C<ev_init	843	Each watcher structure must be initialised by a call to C<ev_init
310	(watcher *, callback)>, which expects a callback to be provided. This	844	(watcher *, callback)>, which expects a callback to be provided. This
311	callback gets invoked each time the event occurs (or, in the case of io	845	callback gets invoked each time the event occurs (or, in the case of I/O
312	watchers, each time the event loop detects that the file descriptor given	846	watchers, each time the event loop detects that the file descriptor given
313	is readable and/or writable).	847	is readable and/or writable).
314		848
315	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	849	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
316	with arguments specific to this watcher type. There is also a macro	850	macro to configure it, with arguments specific to the watcher type. There
317	to combine initialisation and setting in one call: C<< ev_<type>_init	851	is also a macro to combine initialisation and setting in one call: C<<
318	(watcher *, callback, ...) >>.	852	ev_TYPE_init (watcher *, callback, ...) >>.
319		853
320	To make the watcher actually watch out for events, you have to start it	854	To make the watcher actually watch out for events, you have to start it
321	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	855	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
322	*) >>), and you can stop watching for events at any time by calling the	856	*) >>), and you can stop watching for events at any time by calling the
323	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	857	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
324		858
325	As long as your watcher is active (has been started but not stopped) you	859	As long as your watcher is active (has been started but not stopped) you
326	must not touch the values stored in it. Most specifically you must never	860	must not touch the values stored in it. Most specifically you must never
327	reinitialise it or call its set method.	861	reinitialise it or call its C<ev_TYPE_set> macro.
328
329	You can check whether an event is active by calling the C<ev_is_active
330	(watcher *)> macro. To see whether an event is outstanding (but the
331	callback for it has not been called yet) you can use the C<ev_is_pending
332	(watcher *)> macro.
333		862
334	Each and every callback receives the event loop pointer as first, the	863	Each and every callback receives the event loop pointer as first, the
335	registered watcher structure as second, and a bitset of received events as	864	registered watcher structure as second, and a bitset of received events as
336	third argument.	865	third argument.
337		866
…		…
361	The signal specified in the C<ev_signal> watcher has been received by a thread.	890	The signal specified in the C<ev_signal> watcher has been received by a thread.
362		891
363	=item C<EV_CHILD>	892	=item C<EV_CHILD>
364		893
365	The pid specified in the C<ev_child> watcher has received a status change.	894	The pid specified in the C<ev_child> watcher has received a status change.
		895
		896	=item C<EV_STAT>
		897
		898	The path specified in the C<ev_stat> watcher changed its attributes somehow.
366		899
367	=item C<EV_IDLE>	900	=item C<EV_IDLE>
368		901
369	The C<ev_idle> watcher has determined that you have nothing better to do.	902	The C<ev_idle> watcher has determined that you have nothing better to do.
370		903
…		…
378	received events. Callbacks of both watcher types can start and stop as	911	received events. Callbacks of both watcher types can start and stop as
379	many watchers as they want, and all of them will be taken into account	912	many watchers as they want, and all of them will be taken into account
380	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	913	(for example, a C<ev_prepare> watcher might start an idle watcher to keep
381	C<ev_loop> from blocking).	914	C<ev_loop> from blocking).
382		915
		916	=item C<EV_EMBED>
		917
		918	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		919
		920	=item C<EV_FORK>
		921
		922	The event loop has been resumed in the child process after fork (see
		923	C<ev_fork>).
		924
		925	=item C<EV_ASYNC>
		926
		927	The given async watcher has been asynchronously notified (see C<ev_async>).
		928
383	=item C<EV_ERROR>	929	=item C<EV_ERROR>
384		930
385	An unspecified error has occured, the watcher has been stopped. This might	931	An unspecified error has occurred, the watcher has been stopped. This might
386	happen because the watcher could not be properly started because libev	932	happen because the watcher could not be properly started because libev
387	ran out of memory, a file descriptor was found to be closed or any other	933	ran out of memory, a file descriptor was found to be closed or any other
		934	problem. Libev considers these application bugs.
		935
388	problem. You best act on it by reporting the problem and somehow coping	936	You best act on it by reporting the problem and somehow coping with the
389	with the watcher being stopped.	937	watcher being stopped. Note that well-written programs should not receive
		938	an error ever, so when your watcher receives it, this usually indicates a
		939	bug in your program.
390		940
391	Libev will usually signal a few "dummy" events together with an error,	941	Libev will usually signal a few "dummy" events together with an error, for
392	for example it might indicate that a fd is readable or writable, and if	942	example it might indicate that a fd is readable or writable, and if your
393	your callbacks is well-written it can just attempt the operation and cope	943	callbacks is well-written it can just attempt the operation and cope with
394	with the error from read() or write(). This will not work in multithreaded	944	the error from read() or write(). This will not work in multi-threaded
395	programs, though, so beware.	945	programs, though, as the fd could already be closed and reused for another
		946	thing, so beware.
396		947
397	=back	948	=back
398		949
		950	=head2 GENERIC WATCHER FUNCTIONS
		951
		952	=over 4
		953
		954	=item C<ev_init> (ev_TYPE *watcher, callback)
		955
		956	This macro initialises the generic portion of a watcher. The contents
		957	of the watcher object can be arbitrary (so C<malloc> will do). Only
		958	the generic parts of the watcher are initialised, you I<need> to call
		959	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		960	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		961	which rolls both calls into one.
		962
		963	You can reinitialise a watcher at any time as long as it has been stopped
		964	(or never started) and there are no pending events outstanding.
		965
		966	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		967	int revents)>.
		968
		969	Example: Initialise an C<ev_io> watcher in two steps.
		970
		971	ev_io w;
		972	ev_init (&w, my_cb);
		973	ev_io_set (&w, STDIN_FILENO, EV_READ);
		974
		975	=item C<ev_TYPE_set> (ev_TYPE *, [args])
		976
		977	This macro initialises the type-specific parts of a watcher. You need to
		978	call C<ev_init> at least once before you call this macro, but you can
		979	call C<ev_TYPE_set> any number of times. You must not, however, call this
		980	macro on a watcher that is active (it can be pending, however, which is a
		981	difference to the C<ev_init> macro).
		982
		983	Although some watcher types do not have type-specific arguments
		984	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		985
		986	See C<ev_init>, above, for an example.
		987
		988	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		989
		990	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		991	calls into a single call. This is the most convenient method to initialise
		992	a watcher. The same limitations apply, of course.
		993
		994	Example: Initialise and set an C<ev_io> watcher in one step.
		995
		996	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		997
		998	=item C<ev_TYPE_start> (loop *, ev_TYPE *watcher)
		999
		1000	Starts (activates) the given watcher. Only active watchers will receive
		1001	events. If the watcher is already active nothing will happen.
		1002
		1003	Example: Start the C<ev_io> watcher that is being abused as example in this
		1004	whole section.
		1005
		1006	ev_io_start (EV_DEFAULT_UC, &w);
		1007
		1008	=item C<ev_TYPE_stop> (loop *, ev_TYPE *watcher)
		1009
		1010	Stops the given watcher if active, and clears the pending status (whether
		1011	the watcher was active or not).
		1012
		1013	It is possible that stopped watchers are pending - for example,
		1014	non-repeating timers are being stopped when they become pending - but
		1015	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1016	pending. If you want to free or reuse the memory used by the watcher it is
		1017	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1018
		1019	=item bool ev_is_active (ev_TYPE *watcher)
		1020
		1021	Returns a true value iff the watcher is active (i.e. it has been started
		1022	and not yet been stopped). As long as a watcher is active you must not modify
		1023	it.
		1024
		1025	=item bool ev_is_pending (ev_TYPE *watcher)
		1026
		1027	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1028	events but its callback has not yet been invoked). As long as a watcher
		1029	is pending (but not active) you must not call an init function on it (but
		1030	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1031	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1032	it).
		1033
		1034	=item callback ev_cb (ev_TYPE *watcher)
		1035
		1036	Returns the callback currently set on the watcher.
		1037
		1038	=item ev_cb_set (ev_TYPE *watcher, callback)
		1039
		1040	Change the callback. You can change the callback at virtually any time
		1041	(modulo threads).
		1042
		1043	=item ev_set_priority (ev_TYPE *watcher, priority)
		1044
		1045	=item int ev_priority (ev_TYPE *watcher)
		1046
		1047	Set and query the priority of the watcher. The priority is a small
		1048	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1049	(default: C<-2>). Pending watchers with higher priority will be invoked
		1050	before watchers with lower priority, but priority will not keep watchers
		1051	from being executed (except for C<ev_idle> watchers).
		1052
		1053	This means that priorities are I<only> used for ordering callback
		1054	invocation after new events have been received. This is useful, for
		1055	example, to reduce latency after idling, or more often, to bind two
		1056	watchers on the same event and make sure one is called first.
		1057
		1058	If you need to suppress invocation when higher priority events are pending
		1059	you need to look at C<ev_idle> watchers, which provide this functionality.
		1060
		1061	You I<must not> change the priority of a watcher as long as it is active or
		1062	pending.
		1063
		1064	The default priority used by watchers when no priority has been set is
		1065	always C<0>, which is supposed to not be too high and not be too low :).
		1066
		1067	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1068	fine, as long as you do not mind that the priority value you query might
		1069	or might not have been clamped to the valid range.
		1070
		1071	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1072
		1073	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1074	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1075	can deal with that fact, as both are simply passed through to the
		1076	callback.
		1077
		1078	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1079
		1080	If the watcher is pending, this function clears its pending status and
		1081	returns its C<revents> bitset (as if its callback was invoked). If the
		1082	watcher isn't pending it does nothing and returns C<0>.
		1083
		1084	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1085	callback to be invoked, which can be accomplished with this function.
		1086
		1087	=back
		1088
		1089
399	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1090	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
400		1091
401	Each watcher has, by default, a member C<void *data> that you can change	1092	Each watcher has, by default, a member C<void *data> that you can change
402	and read at any time, libev will completely ignore it. This can be used	1093	and read at any time: libev will completely ignore it. This can be used
403	to associate arbitrary data with your watcher. If you need more data and	1094	to associate arbitrary data with your watcher. If you need more data and
404	don't want to allocate memory and store a pointer to it in that data	1095	don't want to allocate memory and store a pointer to it in that data
405	member, you can also "subclass" the watcher type and provide your own	1096	member, you can also "subclass" the watcher type and provide your own
406	data:	1097	data:
407		1098
408	struct my_io	1099	struct my_io
409	{	1100	{
410	struct ev_io io;	1101	ev_io io;
411	int otherfd;	1102	int otherfd;
412	void *somedata;	1103	void *somedata;
413	struct whatever *mostinteresting;	1104	struct whatever *mostinteresting;
414	}	1105	};
		1106
		1107	...
		1108	struct my_io w;
		1109	ev_io_init (&w.io, my_cb, fd, EV_READ);
415		1110
416	And since your callback will be called with a pointer to the watcher, you	1111	And since your callback will be called with a pointer to the watcher, you
417	can cast it back to your own type:	1112	can cast it back to your own type:
418		1113
419	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)	1114	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
420	{	1115	{
421	struct my_io *w = (struct my_io *)w_;	1116	struct my_io *w = (struct my_io *)w_;
422	...	1117	...
423	}	1118	}
424		1119
425	More interesting and less C-conformant ways of catsing your callback type	1120	More interesting and less C-conformant ways of casting your callback type
426	have been omitted....	1121	instead have been omitted.
		1122
		1123	Another common scenario is to use some data structure with multiple
		1124	embedded watchers:
		1125
		1126	struct my_biggy
		1127	{
		1128	int some_data;
		1129	ev_timer t1;
		1130	ev_timer t2;
		1131	}
		1132
		1133	In this case getting the pointer to C<my_biggy> is a bit more
		1134	complicated: Either you store the address of your C<my_biggy> struct
		1135	in the C<data> member of the watcher (for woozies), or you need to use
		1136	some pointer arithmetic using C<offsetof> inside your watchers (for real
		1137	programmers):
		1138
		1139	#include <stddef.h>
		1140
		1141	static void
		1142	t1_cb (EV_P_ ev_timer *w, int revents)
		1143	{
		1144	struct my_biggy big = (struct my_biggy *
		1145	(((char *)w) - offsetof (struct my_biggy, t1));
		1146	}
		1147
		1148	static void
		1149	t2_cb (EV_P_ ev_timer *w, int revents)
		1150	{
		1151	struct my_biggy big = (struct my_biggy *
		1152	(((char *)w) - offsetof (struct my_biggy, t2));
		1153	}
427		1154
428		1155
429	=head1 WATCHER TYPES	1156	=head1 WATCHER TYPES
430		1157
431	This section describes each watcher in detail, but will not repeat	1158	This section describes each watcher in detail, but will not repeat
432	information given in the last section.	1159	information given in the last section. Any initialisation/set macros,
		1160	functions and members specific to the watcher type are explained.
433		1161
		1162	Members are additionally marked with either I<[read-only]>, meaning that,
		1163	while the watcher is active, you can look at the member and expect some
		1164	sensible content, but you must not modify it (you can modify it while the
		1165	watcher is stopped to your hearts content), or I<[read-write]>, which
		1166	means you can expect it to have some sensible content while the watcher
		1167	is active, but you can also modify it. Modifying it may not do something
		1168	sensible or take immediate effect (or do anything at all), but libev will
		1169	not crash or malfunction in any way.
		1170
		1171
434	=head2 C<ev_io> - is this file descriptor readable or writable	1172	=head2 C<ev_io> - is this file descriptor readable or writable?
435		1173
436	I/O watchers check whether a file descriptor is readable or writable	1174	I/O watchers check whether a file descriptor is readable or writable
437	in each iteration of the event loop (This behaviour is called	1175	in each iteration of the event loop, or, more precisely, when reading
438	level-triggering because you keep receiving events as long as the	1176	would not block the process and writing would at least be able to write
439	condition persists. Remember you can stop the watcher if you don't want to	1177	some data. This behaviour is called level-triggering because you keep
440	act on the event and neither want to receive future events).	1178	receiving events as long as the condition persists. Remember you can stop
		1179	the watcher if you don't want to act on the event and neither want to
		1180	receive future events.
441		1181
442	In general you can register as many read and/or write event watchers per	1182	In general you can register as many read and/or write event watchers per
443	fd as you want (as long as you don't confuse yourself). Setting all file	1183	fd as you want (as long as you don't confuse yourself). Setting all file
444	descriptors to non-blocking mode is also usually a good idea (but not	1184	descriptors to non-blocking mode is also usually a good idea (but not
445	required if you know what you are doing).	1185	required if you know what you are doing).
446		1186
447	You have to be careful with dup'ed file descriptors, though. Some backends	1187	If you cannot use non-blocking mode, then force the use of a
448	(the linux epoll backend is a notable example) cannot handle dup'ed file	1188	known-to-be-good backend (at the time of this writing, this includes only
449	descriptors correctly if you register interest in two or more fds pointing	1189	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).
450	to the same underlying file/socket etc. description (that is, they share
451	the same underlying "file open").
452		1190
453	If you must do this, then force the use of a known-to-be-good backend	1191	Another thing you have to watch out for is that it is quite easy to
454	(at the time of this writing, this includes only EVMETHOD_SELECT and	1192	receive "spurious" readiness notifications, that is your callback might
455	EVMETHOD_POLL).	1193	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
		1194	because there is no data. Not only are some backends known to create a
		1195	lot of those (for example Solaris ports), it is very easy to get into
		1196	this situation even with a relatively standard program structure. Thus
		1197	it is best to always use non-blocking I/O: An extra C<read>(2) returning
		1198	C<EAGAIN> is far preferable to a program hanging until some data arrives.
		1199
		1200	If you cannot run the fd in non-blocking mode (for example you should
		1201	not play around with an Xlib connection), then you have to separately
		1202	re-test whether a file descriptor is really ready with a known-to-be good
		1203	interface such as poll (fortunately in our Xlib example, Xlib already
		1204	does this on its own, so its quite safe to use). Some people additionally
		1205	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1206	indefinitely.
		1207
		1208	But really, best use non-blocking mode.
		1209
		1210	=head3 The special problem of disappearing file descriptors
		1211
		1212	Some backends (e.g. kqueue, epoll) need to be told about closing a file
		1213	descriptor (either due to calling C<close> explicitly or any other means,
		1214	such as C<dup2>). The reason is that you register interest in some file
		1215	descriptor, but when it goes away, the operating system will silently drop
		1216	this interest. If another file descriptor with the same number then is
		1217	registered with libev, there is no efficient way to see that this is, in
		1218	fact, a different file descriptor.
		1219
		1220	To avoid having to explicitly tell libev about such cases, libev follows
		1221	the following policy: Each time C<ev_io_set> is being called, libev
		1222	will assume that this is potentially a new file descriptor, otherwise
		1223	it is assumed that the file descriptor stays the same. That means that
		1224	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1225	descriptor even if the file descriptor number itself did not change.
		1226
		1227	This is how one would do it normally anyway, the important point is that
		1228	the libev application should not optimise around libev but should leave
		1229	optimisations to libev.
		1230
		1231	=head3 The special problem of dup'ed file descriptors
		1232
		1233	Some backends (e.g. epoll), cannot register events for file descriptors,
		1234	but only events for the underlying file descriptions. That means when you
		1235	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1236	events for them, only one file descriptor might actually receive events.
		1237
		1238	There is no workaround possible except not registering events
		1239	for potentially C<dup ()>'ed file descriptors, or to resort to
		1240	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1241
		1242	=head3 The special problem of fork
		1243
		1244	Some backends (epoll, kqueue) do not support C<fork ()> at all or exhibit
		1245	useless behaviour. Libev fully supports fork, but needs to be told about
		1246	it in the child.
		1247
		1248	To support fork in your programs, you either have to call
		1249	C<ev_default_fork ()> or C<ev_loop_fork ()> after a fork in the child,
		1250	enable C<EVFLAG_FORKCHECK>, or resort to C<EVBACKEND_SELECT> or
		1251	C<EVBACKEND_POLL>.
		1252
		1253	=head3 The special problem of SIGPIPE
		1254
		1255	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1256	when writing to a pipe whose other end has been closed, your program gets
		1257	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1258	this is sensible behaviour, for daemons, this is usually undesirable.
		1259
		1260	So when you encounter spurious, unexplained daemon exits, make sure you
		1261	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1262	somewhere, as that would have given you a big clue).
		1263
		1264
		1265	=head3 Watcher-Specific Functions
456		1266
457	=over 4	1267	=over 4
458		1268
459	=item ev_io_init (ev_io *, callback, int fd, int events)	1269	=item ev_io_init (ev_io *, callback, int fd, int events)
460		1270
461	=item ev_io_set (ev_io *, int fd, int events)	1271	=item ev_io_set (ev_io *, int fd, int events)
462		1272
463	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1273	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
464	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1274	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
465	EV_WRITE> to receive the given events.	1275	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1276
		1277	=item int fd [read-only]
		1278
		1279	The file descriptor being watched.
		1280
		1281	=item int events [read-only]
		1282
		1283	The events being watched.
466		1284
467	=back	1285	=back
468		1286
		1287	=head3 Examples
		1288
		1289	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1290	readable, but only once. Since it is likely line-buffered, you could
		1291	attempt to read a whole line in the callback.
		1292
		1293	static void
		1294	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1295	{
		1296	ev_io_stop (loop, w);
		1297	.. read from stdin here (or from w->fd) and handle any I/O errors
		1298	}
		1299
		1300	...
		1301	struct ev_loop *loop = ev_default_init (0);
		1302	ev_io stdin_readable;
		1303	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1304	ev_io_start (loop, &stdin_readable);
		1305	ev_loop (loop, 0);
		1306
		1307
469	=head2 C<ev_timer> - relative and optionally recurring timeouts	1308	=head2 C<ev_timer> - relative and optionally repeating timeouts
470		1309
471	Timer watchers are simple relative timers that generate an event after a	1310	Timer watchers are simple relative timers that generate an event after a
472	given time, and optionally repeating in regular intervals after that.	1311	given time, and optionally repeating in regular intervals after that.
473		1312
474	The timers are based on real time, that is, if you register an event that	1313	The timers are based on real time, that is, if you register an event that
475	times out after an hour and you reset your system clock to last years	1314	times out after an hour and you reset your system clock to January last
476	time, it will still time out after (roughly) and hour. "Roughly" because	1315	year, it will still time out after (roughly) one hour. "Roughly" because
477	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1316	detecting time jumps is hard, and some inaccuracies are unavoidable (the
478	monotonic clock option helps a lot here).	1317	monotonic clock option helps a lot here).
		1318
		1319	The callback is guaranteed to be invoked only I<after> its timeout has
		1320	passed, but if multiple timers become ready during the same loop iteration
		1321	then order of execution is undefined.
		1322
		1323	=head3 Be smart about timeouts
		1324
		1325	Many real-world problems involve some kind of timeout, usually for error
		1326	recovery. A typical example is an HTTP request - if the other side hangs,
		1327	you want to raise some error after a while.
		1328
		1329	What follows are some ways to handle this problem, from obvious and
		1330	inefficient to smart and efficient.
		1331
		1332	In the following, a 60 second activity timeout is assumed - a timeout that
		1333	gets reset to 60 seconds each time there is activity (e.g. each time some
		1334	data or other life sign was received).
		1335
		1336	=over 4
		1337
		1338	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1339
		1340	This is the most obvious, but not the most simple way: In the beginning,
		1341	start the watcher:
		1342
		1343	ev_timer_init (timer, callback, 60., 0.);
		1344	ev_timer_start (loop, timer);
		1345
		1346	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1347	and start it again:
		1348
		1349	ev_timer_stop (loop, timer);
		1350	ev_timer_set (timer, 60., 0.);
		1351	ev_timer_start (loop, timer);
		1352
		1353	This is relatively simple to implement, but means that each time there is
		1354	some activity, libev will first have to remove the timer from its internal
		1355	data structure and then add it again. Libev tries to be fast, but it's
		1356	still not a constant-time operation.
		1357
		1358	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1359
		1360	This is the easiest way, and involves using C<ev_timer_again> instead of
		1361	C<ev_timer_start>.
		1362
		1363	To implement this, configure an C<ev_timer> with a C<repeat> value
		1364	of C<60> and then call C<ev_timer_again> at start and each time you
		1365	successfully read or write some data. If you go into an idle state where
		1366	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1367	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1368
		1369	That means you can ignore both the C<ev_timer_start> function and the
		1370	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1371	member and C<ev_timer_again>.
		1372
		1373	At start:
		1374
		1375	ev_timer_init (timer, callback);
		1376	timer->repeat = 60.;
		1377	ev_timer_again (loop, timer);
		1378
		1379	Each time there is some activity:
		1380
		1381	ev_timer_again (loop, timer);
		1382
		1383	It is even possible to change the time-out on the fly, regardless of
		1384	whether the watcher is active or not:
		1385
		1386	timer->repeat = 30.;
		1387	ev_timer_again (loop, timer);
		1388
		1389	This is slightly more efficient then stopping/starting the timer each time
		1390	you want to modify its timeout value, as libev does not have to completely
		1391	remove and re-insert the timer from/into its internal data structure.
		1392
		1393	It is, however, even simpler than the "obvious" way to do it.
		1394
		1395	=item 3. Let the timer time out, but then re-arm it as required.
		1396
		1397	This method is more tricky, but usually most efficient: Most timeouts are
		1398	relatively long compared to the intervals between other activity - in
		1399	our example, within 60 seconds, there are usually many I/O events with
		1400	associated activity resets.
		1401
		1402	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1403	but remember the time of last activity, and check for a real timeout only
		1404	within the callback:
		1405
		1406	ev_tstamp last_activity; // time of last activity
		1407
		1408	static void
		1409	callback (EV_P_ ev_timer *w, int revents)
		1410	{
		1411	ev_tstamp now = ev_now (EV_A);
		1412	ev_tstamp timeout = last_activity + 60.;
		1413
		1414	// if last_activity + 60. is older than now, we did time out
		1415	if (timeout < now)
		1416	{
		1417	// timeout occured, take action
		1418	}
		1419	else
		1420	{
		1421	// callback was invoked, but there was some activity, re-arm
		1422	// the watcher to fire in last_activity + 60, which is
		1423	// guaranteed to be in the future, so "again" is positive:
		1424	w->repeat = timeout - now;
		1425	ev_timer_again (EV_A_ w);
		1426	}
		1427	}
		1428
		1429	To summarise the callback: first calculate the real timeout (defined
		1430	as "60 seconds after the last activity"), then check if that time has
		1431	been reached, which means something I<did>, in fact, time out. Otherwise
		1432	the callback was invoked too early (C<timeout> is in the future), so
		1433	re-schedule the timer to fire at that future time, to see if maybe we have
		1434	a timeout then.
		1435
		1436	Note how C<ev_timer_again> is used, taking advantage of the
		1437	C<ev_timer_again> optimisation when the timer is already running.
		1438
		1439	This scheme causes more callback invocations (about one every 60 seconds
		1440	minus half the average time between activity), but virtually no calls to
		1441	libev to change the timeout.
		1442
		1443	To start the timer, simply initialise the watcher and set C<last_activity>
		1444	to the current time (meaning we just have some activity :), then call the
		1445	callback, which will "do the right thing" and start the timer:
		1446
		1447	ev_timer_init (timer, callback);
		1448	last_activity = ev_now (loop);
		1449	callback (loop, timer, EV_TIMEOUT);
		1450
		1451	And when there is some activity, simply store the current time in
		1452	C<last_activity>, no libev calls at all:
		1453
		1454	last_actiivty = ev_now (loop);
		1455
		1456	This technique is slightly more complex, but in most cases where the
		1457	time-out is unlikely to be triggered, much more efficient.
		1458
		1459	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1460	callback :) - just change the timeout and invoke the callback, which will
		1461	fix things for you.
		1462
		1463	=item 4. Wee, just use a double-linked list for your timeouts.
		1464
		1465	If there is not one request, but many thousands (millions...), all
		1466	employing some kind of timeout with the same timeout value, then one can
		1467	do even better:
		1468
		1469	When starting the timeout, calculate the timeout value and put the timeout
		1470	at the I<end> of the list.
		1471
		1472	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1473	the list is expected to fire (for example, using the technique #3).
		1474
		1475	When there is some activity, remove the timer from the list, recalculate
		1476	the timeout, append it to the end of the list again, and make sure to
		1477	update the C<ev_timer> if it was taken from the beginning of the list.
		1478
		1479	This way, one can manage an unlimited number of timeouts in O(1) time for
		1480	starting, stopping and updating the timers, at the expense of a major
		1481	complication, and having to use a constant timeout. The constant timeout
		1482	ensures that the list stays sorted.
		1483
		1484	=back
		1485
		1486	So which method the best?
		1487
		1488	Method #2 is a simple no-brain-required solution that is adequate in most
		1489	situations. Method #3 requires a bit more thinking, but handles many cases
		1490	better, and isn't very complicated either. In most case, choosing either
		1491	one is fine, with #3 being better in typical situations.
		1492
		1493	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1494	rather complicated, but extremely efficient, something that really pays
		1495	off after the first million or so of active timers, i.e. it's usually
		1496	overkill :)
		1497
		1498	=head3 The special problem of time updates
		1499
		1500	Establishing the current time is a costly operation (it usually takes at
		1501	least two system calls): EV therefore updates its idea of the current
		1502	time only before and after C<ev_loop> collects new events, which causes a
		1503	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		1504	lots of events in one iteration.
479		1505
480	The relative timeouts are calculated relative to the C<ev_now ()>	1506	The relative timeouts are calculated relative to the C<ev_now ()>
481	time. This is usually the right thing as this timestamp refers to the time	1507	time. This is usually the right thing as this timestamp refers to the time
482	of the event triggering whatever timeout you are modifying/starting. If	1508	of the event triggering whatever timeout you are modifying/starting. If
483	you suspect event processing to be delayed and you *need* to base the timeout	1509	you suspect event processing to be delayed and you I<need> to base the
484	on the current time, use something like this to adjust for this:	1510	timeout on the current time, use something like this to adjust for this:
485		1511
486	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	1512	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);
487		1513
		1514	If the event loop is suspended for a long time, you can also force an
		1515	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		1516	()>.
		1517
		1518	=head3 Watcher-Specific Functions and Data Members
		1519
488	=over 4	1520	=over 4
489		1521
490	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	1522	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
491		1523
492	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	1524	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
493		1525
494	Configure the timer to trigger after C<after> seconds. If C<repeat> is	1526	Configure the timer to trigger after C<after> seconds. If C<repeat>
495	C<0.>, then it will automatically be stopped. If it is positive, then the	1527	is C<0.>, then it will automatically be stopped once the timeout is
496	timer will automatically be configured to trigger again C<repeat> seconds	1528	reached. If it is positive, then the timer will automatically be
497	later, again, and again, until stopped manually.	1529	configured to trigger again C<repeat> seconds later, again, and again,
		1530	until stopped manually.
498		1531
499	The timer itself will do a best-effort at avoiding drift, that is, if you	1532	The timer itself will do a best-effort at avoiding drift, that is, if
500	configure a timer to trigger every 10 seconds, then it will trigger at	1533	you configure a timer to trigger every 10 seconds, then it will normally
501	exactly 10 second intervals. If, however, your program cannot keep up with	1534	trigger at exactly 10 second intervals. If, however, your program cannot
502	the timer (because it takes longer than those 10 seconds to do stuff) the	1535	keep up with the timer (because it takes longer than those 10 seconds to
503	timer will not fire more than once per event loop iteration.	1536	do stuff) the timer will not fire more than once per event loop iteration.
504		1537
505	=item ev_timer_again (loop)	1538	=item ev_timer_again (loop, ev_timer *)
506		1539
507	This will act as if the timer timed out and restart it again if it is	1540	This will act as if the timer timed out and restart it again if it is
508	repeating. The exact semantics are:	1541	repeating. The exact semantics are:
509		1542
		1543	If the timer is pending, its pending status is cleared.
		1544
510	If the timer is started but nonrepeating, stop it.	1545	If the timer is started but non-repeating, stop it (as if it timed out).
511		1546
512	If the timer is repeating, either start it if necessary (with the repeat	1547	If the timer is repeating, either start it if necessary (with the
513	value), or reset the running timer to the repeat value.	1548	C<repeat> value), or reset the running timer to the C<repeat> value.
514		1549
515	This sounds a bit complicated, but here is a useful and typical	1550	This sounds a bit complicated, see "Be smart about timeouts", above, for a
516	example: Imagine you have a tcp connection and you want a so-called idle	1551	usage example.
517	timeout, that is, you want to be called when there have been, say, 60	1552
518	seconds of inactivity on the socket. The easiest way to do this is to	1553	=item ev_tstamp repeat [read-write]
519	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each	1554
520	time you successfully read or write some data. If you go into an idle	1555	The current C<repeat> value. Will be used each time the watcher times out
521	state where you do not expect data to travel on the socket, you can stop	1556	or C<ev_timer_again> is called, and determines the next timeout (if any),
522	the timer, and again will automatically restart it if need be.	1557	which is also when any modifications are taken into account.
523		1558
524	=back	1559	=back
525		1560
		1561	=head3 Examples
		1562
		1563	Example: Create a timer that fires after 60 seconds.
		1564
		1565	static void
		1566	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1567	{
		1568	.. one minute over, w is actually stopped right here
		1569	}
		1570
		1571	ev_timer mytimer;
		1572	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		1573	ev_timer_start (loop, &mytimer);
		1574
		1575	Example: Create a timeout timer that times out after 10 seconds of
		1576	inactivity.
		1577
		1578	static void
		1579	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		1580	{
		1581	.. ten seconds without any activity
		1582	}
		1583
		1584	ev_timer mytimer;
		1585	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		1586	ev_timer_again (&mytimer); /* start timer */
		1587	ev_loop (loop, 0);
		1588
		1589	// and in some piece of code that gets executed on any "activity":
		1590	// reset the timeout to start ticking again at 10 seconds
		1591	ev_timer_again (&mytimer);
		1592
		1593
526	=head2 C<ev_periodic> - to cron or not to cron	1594	=head2 C<ev_periodic> - to cron or not to cron?
527		1595
528	Periodic watchers are also timers of a kind, but they are very versatile	1596	Periodic watchers are also timers of a kind, but they are very versatile
529	(and unfortunately a bit complex).	1597	(and unfortunately a bit complex).
530		1598
531	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1599	Unlike C<ev_timer>'s, they are not based on real time (or relative time)
532	but on wallclock time (absolute time). You can tell a periodic watcher	1600	but on wall clock time (absolute time). You can tell a periodic watcher
533	to trigger "at" some specific point in time. For example, if you tell a	1601	to trigger after some specific point in time. For example, if you tell a
534	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	1602	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()
535	+ 10.>) and then reset your system clock to the last year, then it will	1603	+ 10.>, that is, an absolute time not a delay) and then reset your system
		1604	clock to January of the previous year, then it will take more than year
536	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	1605	to trigger the event (unlike an C<ev_timer>, which would still trigger
537	roughly 10 seconds later and of course not if you reset your system time	1606	roughly 10 seconds later as it uses a relative timeout).
538	again).
539		1607
540	They can also be used to implement vastly more complex timers, such as	1608	C<ev_periodic>s can also be used to implement vastly more complex timers,
541	triggering an event on eahc midnight, local time.	1609	such as triggering an event on each "midnight, local time", or other
		1610	complicated rules.
		1611
		1612	As with timers, the callback is guaranteed to be invoked only when the
		1613	time (C<at>) has passed, but if multiple periodic timers become ready
		1614	during the same loop iteration, then order of execution is undefined.
		1615
		1616	=head3 Watcher-Specific Functions and Data Members
542		1617
543	=over 4	1618	=over 4
544		1619
545	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1620	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)
546		1621
547	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1622	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)
548		1623
549	Lots of arguments, lets sort it out... There are basically three modes of	1624	Lots of arguments, lets sort it out... There are basically three modes of
550	operation, and we will explain them from simplest to complex:	1625	operation, and we will explain them from simplest to most complex:
551
552		1626
553	=over 4	1627	=over 4
554		1628
555	=item * absolute timer (interval = reschedule_cb = 0)	1629	=item * absolute timer (at = time, interval = reschedule_cb = 0)
556		1630
557	In this configuration the watcher triggers an event at the wallclock time	1631	In this configuration the watcher triggers an event after the wall clock
558	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	1632	time C<at> has passed. It will not repeat and will not adjust when a time
559	that is, if it is to be run at January 1st 2011 then it will run when the	1633	jump occurs, that is, if it is to be run at January 1st 2011 then it will
560	system time reaches or surpasses this time.	1634	only run when the system clock reaches or surpasses this time.
561		1635
562	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	1636	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)
563		1637
564	In this mode the watcher will always be scheduled to time out at the next	1638	In this mode the watcher will always be scheduled to time out at the next
565	C<at + N * interval> time (for some integer N) and then repeat, regardless	1639	C<at + N * interval> time (for some integer N, which can also be negative)
566	of any time jumps.	1640	and then repeat, regardless of any time jumps.
567		1641
568	This can be used to create timers that do not drift with respect to system	1642	This can be used to create timers that do not drift with respect to the
569	time:	1643	system clock, for example, here is a C<ev_periodic> that triggers each
		1644	hour, on the hour:
570		1645
571	ev_periodic_set (&periodic, 0., 3600., 0);	1646	ev_periodic_set (&periodic, 0., 3600., 0);
572		1647
573	This doesn't mean there will always be 3600 seconds in between triggers,	1648	This doesn't mean there will always be 3600 seconds in between triggers,
574	but only that the the callback will be called when the system time shows a	1649	but only that the callback will be called when the system time shows a
575	full hour (UTC), or more correctly, when the system time is evenly divisible	1650	full hour (UTC), or more correctly, when the system time is evenly divisible
576	by 3600.	1651	by 3600.
577		1652
578	Another way to think about it (for the mathematically inclined) is that	1653	Another way to think about it (for the mathematically inclined) is that
579	C<ev_periodic> will try to run the callback in this mode at the next possible	1654	C<ev_periodic> will try to run the callback in this mode at the next possible
580	time where C<time = at (mod interval)>, regardless of any time jumps.	1655	time where C<time = at (mod interval)>, regardless of any time jumps.
581		1656
		1657	For numerical stability it is preferable that the C<at> value is near
		1658	C<ev_now ()> (the current time), but there is no range requirement for
		1659	this value, and in fact is often specified as zero.
		1660
		1661	Note also that there is an upper limit to how often a timer can fire (CPU
		1662	speed for example), so if C<interval> is very small then timing stability
		1663	will of course deteriorate. Libev itself tries to be exact to be about one
		1664	millisecond (if the OS supports it and the machine is fast enough).
		1665
582	=item * manual reschedule mode (reschedule_cb = callback)	1666	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)
583		1667
584	In this mode the values for C<interval> and C<at> are both being	1668	In this mode the values for C<interval> and C<at> are both being
585	ignored. Instead, each time the periodic watcher gets scheduled, the	1669	ignored. Instead, each time the periodic watcher gets scheduled, the
586	reschedule callback will be called with the watcher as first, and the	1670	reschedule callback will be called with the watcher as first, and the
587	current time as second argument.	1671	current time as second argument.
588		1672
589	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	1673	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,
590	ever, or make any event loop modifications>. If you need to stop it,	1674	ever, or make ANY event loop modifications whatsoever>.
591	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by
592	starting a prepare watcher).
593		1675
		1676	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		1677	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		1678	only event loop modification you are allowed to do).
		1679
594	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	1680	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
595	ev_tstamp now)>, e.g.:	1681	*w, ev_tstamp now)>, e.g.:
596		1682
		1683	static ev_tstamp
597	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	1684	my_rescheduler (ev_periodic *w, ev_tstamp now)
598	{	1685	{
599	return now + 60.;	1686	return now + 60.;
600	}	1687	}
601		1688
602	It must return the next time to trigger, based on the passed time value	1689	It must return the next time to trigger, based on the passed time value
603	(that is, the lowest time value larger than to the second argument). It	1690	(that is, the lowest time value larger than to the second argument). It
604	will usually be called just before the callback will be triggered, but	1691	will usually be called just before the callback will be triggered, but
605	might be called at other times, too.	1692	might be called at other times, too.
606		1693
607	NOTE: I<< This callback must always return a time that is later than the	1694	NOTE: I<< This callback must always return a time that is higher than or
608	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	1695	equal to the passed C<now> value >>.
609		1696
610	This can be used to create very complex timers, such as a timer that	1697	This can be used to create very complex timers, such as a timer that
611	triggers on each midnight, local time. To do this, you would calculate the	1698	triggers on "next midnight, local time". To do this, you would calculate the
612	next midnight after C<now> and return the timestamp value for this. How	1699	next midnight after C<now> and return the timestamp value for this. How
613	you do this is, again, up to you (but it is not trivial, which is the main	1700	you do this is, again, up to you (but it is not trivial, which is the main
614	reason I omitted it as an example).	1701	reason I omitted it as an example).
615		1702
616	=back	1703	=back
…		…
620	Simply stops and restarts the periodic watcher again. This is only useful	1707	Simply stops and restarts the periodic watcher again. This is only useful
621	when you changed some parameters or the reschedule callback would return	1708	when you changed some parameters or the reschedule callback would return
622	a different time than the last time it was called (e.g. in a crond like	1709	a different time than the last time it was called (e.g. in a crond like
623	program when the crontabs have changed).	1710	program when the crontabs have changed).
624		1711
		1712	=item ev_tstamp ev_periodic_at (ev_periodic *)
		1713
		1714	When active, returns the absolute time that the watcher is supposed to
		1715	trigger next.
		1716
		1717	=item ev_tstamp offset [read-write]
		1718
		1719	When repeating, this contains the offset value, otherwise this is the
		1720	absolute point in time (the C<at> value passed to C<ev_periodic_set>).
		1721
		1722	Can be modified any time, but changes only take effect when the periodic
		1723	timer fires or C<ev_periodic_again> is being called.
		1724
		1725	=item ev_tstamp interval [read-write]
		1726
		1727	The current interval value. Can be modified any time, but changes only
		1728	take effect when the periodic timer fires or C<ev_periodic_again> is being
		1729	called.
		1730
		1731	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		1732
		1733	The current reschedule callback, or C<0>, if this functionality is
		1734	switched off. Can be changed any time, but changes only take effect when
		1735	the periodic timer fires or C<ev_periodic_again> is being called.
		1736
625	=back	1737	=back
626		1738
		1739	=head3 Examples
		1740
		1741	Example: Call a callback every hour, or, more precisely, whenever the
		1742	system time is divisible by 3600. The callback invocation times have
		1743	potentially a lot of jitter, but good long-term stability.
		1744
		1745	static void
		1746	clock_cb (struct ev_loop *loop, ev_io *w, int revents)
		1747	{
		1748	... its now a full hour (UTC, or TAI or whatever your clock follows)
		1749	}
		1750
		1751	ev_periodic hourly_tick;
		1752	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		1753	ev_periodic_start (loop, &hourly_tick);
		1754
		1755	Example: The same as above, but use a reschedule callback to do it:
		1756
		1757	#include <math.h>
		1758
		1759	static ev_tstamp
		1760	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		1761	{
		1762	return now + (3600. - fmod (now, 3600.));
		1763	}
		1764
		1765	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		1766
		1767	Example: Call a callback every hour, starting now:
		1768
		1769	ev_periodic hourly_tick;
		1770	ev_periodic_init (&hourly_tick, clock_cb,
		1771	fmod (ev_now (loop), 3600.), 3600., 0);
		1772	ev_periodic_start (loop, &hourly_tick);
		1773
		1774
627	=head2 C<ev_signal> - signal me when a signal gets signalled	1775	=head2 C<ev_signal> - signal me when a signal gets signalled!
628		1776
629	Signal watchers will trigger an event when the process receives a specific	1777	Signal watchers will trigger an event when the process receives a specific
630	signal one or more times. Even though signals are very asynchronous, libev	1778	signal one or more times. Even though signals are very asynchronous, libev
631	will try it's best to deliver signals synchronously, i.e. as part of the	1779	will try it's best to deliver signals synchronously, i.e. as part of the
632	normal event processing, like any other event.	1780	normal event processing, like any other event.
633		1781
		1782	If you want signals asynchronously, just use C<sigaction> as you would
		1783	do without libev and forget about sharing the signal. You can even use
		1784	C<ev_async> from a signal handler to synchronously wake up an event loop.
		1785
634	You can configure as many watchers as you like per signal. Only when the	1786	You can configure as many watchers as you like per signal. Only when the
635	first watcher gets started will libev actually register a signal watcher	1787	first watcher gets started will libev actually register a signal handler
636	with the kernel (thus it coexists with your own signal handlers as long	1788	with the kernel (thus it coexists with your own signal handlers as long as
637	as you don't register any with libev). Similarly, when the last signal	1789	you don't register any with libev for the same signal). Similarly, when
638	watcher for a signal is stopped libev will reset the signal handler to	1790	the last signal watcher for a signal is stopped, libev will reset the
639	SIG_DFL (regardless of what it was set to before).	1791	signal handler to SIG_DFL (regardless of what it was set to before).
		1792
		1793	If possible and supported, libev will install its handlers with
		1794	C<SA_RESTART> behaviour enabled, so system calls should not be unduly
		1795	interrupted. If you have a problem with system calls getting interrupted by
		1796	signals you can block all signals in an C<ev_check> watcher and unblock
		1797	them in an C<ev_prepare> watcher.
		1798
		1799	=head3 Watcher-Specific Functions and Data Members
640		1800
641	=over 4	1801	=over 4
642		1802
643	=item ev_signal_init (ev_signal *, callback, int signum)	1803	=item ev_signal_init (ev_signal *, callback, int signum)
644		1804
645	=item ev_signal_set (ev_signal *, int signum)	1805	=item ev_signal_set (ev_signal *, int signum)
646		1806
647	Configures the watcher to trigger on the given signal number (usually one	1807	Configures the watcher to trigger on the given signal number (usually one
648	of the C<SIGxxx> constants).	1808	of the C<SIGxxx> constants).
649		1809
		1810	=item int signum [read-only]
		1811
		1812	The signal the watcher watches out for.
		1813
650	=back	1814	=back
651		1815
		1816	=head3 Examples
		1817
		1818	Example: Try to exit cleanly on SIGINT.
		1819
		1820	static void
		1821	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		1822	{
		1823	ev_unloop (loop, EVUNLOOP_ALL);
		1824	}
		1825
		1826	ev_signal signal_watcher;
		1827	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		1828	ev_signal_start (loop, &signal_watcher);
		1829
		1830
652	=head2 C<ev_child> - wait for pid status changes	1831	=head2 C<ev_child> - watch out for process status changes
653		1832
654	Child watchers trigger when your process receives a SIGCHLD in response to	1833	Child watchers trigger when your process receives a SIGCHLD in response to
655	some child status changes (most typically when a child of yours dies).	1834	some child status changes (most typically when a child of yours dies or
		1835	exits). It is permissible to install a child watcher I<after> the child
		1836	has been forked (which implies it might have already exited), as long
		1837	as the event loop isn't entered (or is continued from a watcher), i.e.,
		1838	forking and then immediately registering a watcher for the child is fine,
		1839	but forking and registering a watcher a few event loop iterations later is
		1840	not.
		1841
		1842	Only the default event loop is capable of handling signals, and therefore
		1843	you can only register child watchers in the default event loop.
		1844
		1845	=head3 Process Interaction
		1846
		1847	Libev grabs C<SIGCHLD> as soon as the default event loop is
		1848	initialised. This is necessary to guarantee proper behaviour even if
		1849	the first child watcher is started after the child exits. The occurrence
		1850	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		1851	synchronously as part of the event loop processing. Libev always reaps all
		1852	children, even ones not watched.
		1853
		1854	=head3 Overriding the Built-In Processing
		1855
		1856	Libev offers no special support for overriding the built-in child
		1857	processing, but if your application collides with libev's default child
		1858	handler, you can override it easily by installing your own handler for
		1859	C<SIGCHLD> after initialising the default loop, and making sure the
		1860	default loop never gets destroyed. You are encouraged, however, to use an
		1861	event-based approach to child reaping and thus use libev's support for
		1862	that, so other libev users can use C<ev_child> watchers freely.
		1863
		1864	=head3 Stopping the Child Watcher
		1865
		1866	Currently, the child watcher never gets stopped, even when the
		1867	child terminates, so normally one needs to stop the watcher in the
		1868	callback. Future versions of libev might stop the watcher automatically
		1869	when a child exit is detected.
		1870
		1871	=head3 Watcher-Specific Functions and Data Members
656		1872
657	=over 4	1873	=over 4
658		1874
659	=item ev_child_init (ev_child *, callback, int pid)	1875	=item ev_child_init (ev_child *, callback, int pid, int trace)
660		1876
661	=item ev_child_set (ev_child *, int pid)	1877	=item ev_child_set (ev_child *, int pid, int trace)
662		1878
663	Configures the watcher to wait for status changes of process C<pid> (or	1879	Configures the watcher to wait for status changes of process C<pid> (or
664	I<any> process if C<pid> is specified as C<0>). The callback can look	1880	I<any> process if C<pid> is specified as C<0>). The callback can look
665	at the C<rstatus> member of the C<ev_child> watcher structure to see	1881	at the C<rstatus> member of the C<ev_child> watcher structure to see
666	the status word (use the macros from C<sys/wait.h> and see your systems	1882	the status word (use the macros from C<sys/wait.h> and see your systems
667	C<waitpid> documentation). The C<rpid> member contains the pid of the	1883	C<waitpid> documentation). The C<rpid> member contains the pid of the
668	process causing the status change.	1884	process causing the status change. C<trace> must be either C<0> (only
		1885	activate the watcher when the process terminates) or C<1> (additionally
		1886	activate the watcher when the process is stopped or continued).
		1887
		1888	=item int pid [read-only]
		1889
		1890	The process id this watcher watches out for, or C<0>, meaning any process id.
		1891
		1892	=item int rpid [read-write]
		1893
		1894	The process id that detected a status change.
		1895
		1896	=item int rstatus [read-write]
		1897
		1898	The process exit/trace status caused by C<rpid> (see your systems
		1899	C<waitpid> and C<sys/wait.h> documentation for details).
669		1900
670	=back	1901	=back
671		1902
		1903	=head3 Examples
		1904
		1905	Example: C<fork()> a new process and install a child handler to wait for
		1906	its completion.
		1907
		1908	ev_child cw;
		1909
		1910	static void
		1911	child_cb (EV_P_ ev_child *w, int revents)
		1912	{
		1913	ev_child_stop (EV_A_ w);
		1914	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		1915	}
		1916
		1917	pid_t pid = fork ();
		1918
		1919	if (pid < 0)
		1920	// error
		1921	else if (pid == 0)
		1922	{
		1923	// the forked child executes here
		1924	exit (1);
		1925	}
		1926	else
		1927	{
		1928	ev_child_init (&cw, child_cb, pid, 0);
		1929	ev_child_start (EV_DEFAULT_ &cw);
		1930	}
		1931
		1932
		1933	=head2 C<ev_stat> - did the file attributes just change?
		1934
		1935	This watches a file system path for attribute changes. That is, it calls
		1936	C<stat> on that path in regular intervals (or when the OS says it changed)
		1937	and sees if it changed compared to the last time, invoking the callback if
		1938	it did.
		1939
		1940	The path does not need to exist: changing from "path exists" to "path does
		1941	not exist" is a status change like any other. The condition "path does not
		1942	exist" (or more correctly "path cannot be stat'ed") is signified by the
		1943	C<st_nlink> field being zero (which is otherwise always forced to be at
		1944	least one) and all the other fields of the stat buffer having unspecified
		1945	contents.
		1946
		1947	The path I<must not> end in a slash or contain special components such as
		1948	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		1949	your working directory changes, then the behaviour is undefined.
		1950
		1951	Since there is no portable change notification interface available, the
		1952	portable implementation simply calls C<stat(2)> regularly on the path
		1953	to see if it changed somehow. You can specify a recommended polling
		1954	interval for this case. If you specify a polling interval of C<0> (highly
		1955	recommended!) then a I<suitable, unspecified default> value will be used
		1956	(which you can expect to be around five seconds, although this might
		1957	change dynamically). Libev will also impose a minimum interval which is
		1958	currently around C<0.1>, but that's usually overkill.
		1959
		1960	This watcher type is not meant for massive numbers of stat watchers,
		1961	as even with OS-supported change notifications, this can be
		1962	resource-intensive.
		1963
		1964	At the time of this writing, the only OS-specific interface implemented
		1965	is the Linux inotify interface (implementing kqueue support is left as an
		1966	exercise for the reader. Note, however, that the author sees no way of
		1967	implementing C<ev_stat> semantics with kqueue, except as a hint).
		1968
		1969	=head3 ABI Issues (Largefile Support)
		1970
		1971	Libev by default (unless the user overrides this) uses the default
		1972	compilation environment, which means that on systems with large file
		1973	support disabled by default, you get the 32 bit version of the stat
		1974	structure. When using the library from programs that change the ABI to
		1975	use 64 bit file offsets the programs will fail. In that case you have to
		1976	compile libev with the same flags to get binary compatibility. This is
		1977	obviously the case with any flags that change the ABI, but the problem is
		1978	most noticeably displayed with ev_stat and large file support.
		1979
		1980	The solution for this is to lobby your distribution maker to make large
		1981	file interfaces available by default (as e.g. FreeBSD does) and not
		1982	optional. Libev cannot simply switch on large file support because it has
		1983	to exchange stat structures with application programs compiled using the
		1984	default compilation environment.
		1985
		1986	=head3 Inotify and Kqueue
		1987
		1988	When C<inotify (7)> support has been compiled into libev and present at
		1989	runtime, it will be used to speed up change detection where possible. The
		1990	inotify descriptor will be created lazily when the first C<ev_stat>
		1991	watcher is being started.
		1992
		1993	Inotify presence does not change the semantics of C<ev_stat> watchers
		1994	except that changes might be detected earlier, and in some cases, to avoid
		1995	making regular C<stat> calls. Even in the presence of inotify support
		1996	there are many cases where libev has to resort to regular C<stat> polling,
		1997	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		1998	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		1999	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2000	xfs are fully working) libev usually gets away without polling.
		2001
		2002	There is no support for kqueue, as apparently it cannot be used to
		2003	implement this functionality, due to the requirement of having a file
		2004	descriptor open on the object at all times, and detecting renames, unlinks
		2005	etc. is difficult.
		2006
		2007	=head3 C<stat ()> is a synchronous operation
		2008
		2009	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2010	the process. The exception are C<ev_stat> watchers - those call C<stat
		2011	()>, which is a synchronous operation.
		2012
		2013	For local paths, this usually doesn't matter: unless the system is very
		2014	busy or the intervals between stat's are large, a stat call will be fast,
		2015	as the path data is suually in memory already (except when starting the
		2016	watcher).
		2017
		2018	For networked file systems, calling C<stat ()> can block an indefinite
		2019	time due to network issues, and even under good conditions, a stat call
		2020	often takes multiple milliseconds.
		2021
		2022	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2023	paths, although this is fully supported by libev.
		2024
		2025	=head3 The special problem of stat time resolution
		2026
		2027	The C<stat ()> system call only supports full-second resolution portably,
		2028	and even on systems where the resolution is higher, most file systems
		2029	still only support whole seconds.
		2030
		2031	That means that, if the time is the only thing that changes, you can
		2032	easily miss updates: on the first update, C<ev_stat> detects a change and
		2033	calls your callback, which does something. When there is another update
		2034	within the same second, C<ev_stat> will be unable to detect unless the
		2035	stat data does change in other ways (e.g. file size).
		2036
		2037	The solution to this is to delay acting on a change for slightly more
		2038	than a second (or till slightly after the next full second boundary), using
		2039	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2040	ev_timer_again (loop, w)>).
		2041
		2042	The C<.02> offset is added to work around small timing inconsistencies
		2043	of some operating systems (where the second counter of the current time
		2044	might be be delayed. One such system is the Linux kernel, where a call to
		2045	C<gettimeofday> might return a timestamp with a full second later than
		2046	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2047	update file times then there will be a small window where the kernel uses
		2048	the previous second to update file times but libev might already execute
		2049	the timer callback).
		2050
		2051	=head3 Watcher-Specific Functions and Data Members
		2052
		2053	=over 4
		2054
		2055	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2056
		2057	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2058
		2059	Configures the watcher to wait for status changes of the given
		2060	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2061	be detected and should normally be specified as C<0> to let libev choose
		2062	a suitable value. The memory pointed to by C<path> must point to the same
		2063	path for as long as the watcher is active.
		2064
		2065	The callback will receive an C<EV_STAT> event when a change was detected,
		2066	relative to the attributes at the time the watcher was started (or the
		2067	last change was detected).
		2068
		2069	=item ev_stat_stat (loop, ev_stat *)
		2070
		2071	Updates the stat buffer immediately with new values. If you change the
		2072	watched path in your callback, you could call this function to avoid
		2073	detecting this change (while introducing a race condition if you are not
		2074	the only one changing the path). Can also be useful simply to find out the
		2075	new values.
		2076
		2077	=item ev_statdata attr [read-only]
		2078
		2079	The most-recently detected attributes of the file. Although the type is
		2080	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2081	suitable for your system, but you can only rely on the POSIX-standardised
		2082	members to be present. If the C<st_nlink> member is C<0>, then there was
		2083	some error while C<stat>ing the file.
		2084
		2085	=item ev_statdata prev [read-only]
		2086
		2087	The previous attributes of the file. The callback gets invoked whenever
		2088	C<prev> != C<attr>, or, more precisely, one or more of these members
		2089	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2090	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2091
		2092	=item ev_tstamp interval [read-only]
		2093
		2094	The specified interval.
		2095
		2096	=item const char *path [read-only]
		2097
		2098	The file system path that is being watched.
		2099
		2100	=back
		2101
		2102	=head3 Examples
		2103
		2104	Example: Watch C</etc/passwd> for attribute changes.
		2105
		2106	static void
		2107	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2108	{
		2109	/* /etc/passwd changed in some way */
		2110	if (w->attr.st_nlink)
		2111	{
		2112	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2113	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2114	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2115	}
		2116	else
		2117	/* you shalt not abuse printf for puts */
		2118	puts ("wow, /etc/passwd is not there, expect problems. "
		2119	"if this is windows, they already arrived\n");
		2120	}
		2121
		2122	...
		2123	ev_stat passwd;
		2124
		2125	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2126	ev_stat_start (loop, &passwd);
		2127
		2128	Example: Like above, but additionally use a one-second delay so we do not
		2129	miss updates (however, frequent updates will delay processing, too, so
		2130	one might do the work both on C<ev_stat> callback invocation I<and> on
		2131	C<ev_timer> callback invocation).
		2132
		2133	static ev_stat passwd;
		2134	static ev_timer timer;
		2135
		2136	static void
		2137	timer_cb (EV_P_ ev_timer *w, int revents)
		2138	{
		2139	ev_timer_stop (EV_A_ w);
		2140
		2141	/* now it's one second after the most recent passwd change */
		2142	}
		2143
		2144	static void
		2145	stat_cb (EV_P_ ev_stat *w, int revents)
		2146	{
		2147	/* reset the one-second timer */
		2148	ev_timer_again (EV_A_ &timer);
		2149	}
		2150
		2151	...
		2152	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2153	ev_stat_start (loop, &passwd);
		2154	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2155
		2156
672	=head2 C<ev_idle> - when you've got nothing better to do	2157	=head2 C<ev_idle> - when you've got nothing better to do...
673		2158
674	Idle watchers trigger events when there are no other events are pending	2159	Idle watchers trigger events when no other events of the same or higher
675	(prepare, check and other idle watchers do not count). That is, as long	2160	priority are pending (prepare, check and other idle watchers do not count
676	as your process is busy handling sockets or timeouts (or even signals,	2161	as receiving "events").
677	imagine) it will not be triggered. But when your process is idle all idle	2162
678	watchers are being called again and again, once per event loop iteration -	2163	That is, as long as your process is busy handling sockets or timeouts
		2164	(or even signals, imagine) of the same or higher priority it will not be
		2165	triggered. But when your process is idle (or only lower-priority watchers
		2166	are pending), the idle watchers are being called once per event loop
679	until stopped, that is, or your process receives more events and becomes	2167	iteration - until stopped, that is, or your process receives more events
680	busy.	2168	and becomes busy again with higher priority stuff.
681		2169
682	The most noteworthy effect is that as long as any idle watchers are	2170	The most noteworthy effect is that as long as any idle watchers are
683	active, the process will not block when waiting for new events.	2171	active, the process will not block when waiting for new events.
684		2172
685	Apart from keeping your process non-blocking (which is a useful	2173	Apart from keeping your process non-blocking (which is a useful
686	effect on its own sometimes), idle watchers are a good place to do	2174	effect on its own sometimes), idle watchers are a good place to do
687	"pseudo-background processing", or delay processing stuff to after the	2175	"pseudo-background processing", or delay processing stuff to after the
688	event loop has handled all outstanding events.	2176	event loop has handled all outstanding events.
689		2177
		2178	=head3 Watcher-Specific Functions and Data Members
		2179
690	=over 4	2180	=over 4
691		2181
692	=item ev_idle_init (ev_signal *, callback)	2182	=item ev_idle_init (ev_signal *, callback)
693		2183
694	Initialises and configures the idle watcher - it has no parameters of any	2184	Initialises and configures the idle watcher - it has no parameters of any
695	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2185	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
696	believe me.	2186	believe me.
697		2187
698	=back	2188	=back
699		2189
		2190	=head3 Examples
		2191
		2192	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2193	callback, free it. Also, use no error checking, as usual.
		2194
		2195	static void
		2196	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2197	{
		2198	free (w);
		2199	// now do something you wanted to do when the program has
		2200	// no longer anything immediate to do.
		2201	}
		2202
		2203	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2204	ev_idle_init (idle_watcher, idle_cb);
		2205	ev_idle_start (loop, idle_cb);
		2206
		2207
700	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	2208	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
701		2209
702	Prepare and check watchers are usually (but not always) used in tandem:	2210	Prepare and check watchers are usually (but not always) used in pairs:
703	prepare watchers get invoked before the process blocks and check watchers	2211	prepare watchers get invoked before the process blocks and check watchers
704	afterwards.	2212	afterwards.
705		2213
		2214	You I<must not> call C<ev_loop> or similar functions that enter
		2215	the current event loop from either C<ev_prepare> or C<ev_check>
		2216	watchers. Other loops than the current one are fine, however. The
		2217	rationale behind this is that you do not need to check for recursion in
		2218	those watchers, i.e. the sequence will always be C<ev_prepare>, blocking,
		2219	C<ev_check> so if you have one watcher of each kind they will always be
		2220	called in pairs bracketing the blocking call.
		2221
706	Their main purpose is to integrate other event mechanisms into libev. This	2222	Their main purpose is to integrate other event mechanisms into libev and
707	could be used, for example, to track variable changes, implement your own	2223	their use is somewhat advanced. They could be used, for example, to track
708	watchers, integrate net-snmp or a coroutine library and lots more.	2224	variable changes, implement your own watchers, integrate net-snmp or a
		2225	coroutine library and lots more. They are also occasionally useful if
		2226	you cache some data and want to flush it before blocking (for example,
		2227	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		2228	watcher).
709		2229
710	This is done by examining in each prepare call which file descriptors need	2230	This is done by examining in each prepare call which file descriptors
711	to be watched by the other library, registering C<ev_io> watchers for	2231	need to be watched by the other library, registering C<ev_io> watchers
712	them and starting an C<ev_timer> watcher for any timeouts (many libraries	2232	for them and starting an C<ev_timer> watcher for any timeouts (many
713	provide just this functionality). Then, in the check watcher you check for	2233	libraries provide exactly this functionality). Then, in the check watcher,
714	any events that occured (by checking the pending status of all watchers	2234	you check for any events that occurred (by checking the pending status
715	and stopping them) and call back into the library. The I/O and timer	2235	of all watchers and stopping them) and call back into the library. The
716	callbacks will never actually be called (but must be valid nevertheless,	2236	I/O and timer callbacks will never actually be called (but must be valid
717	because you never know, you know?).	2237	nevertheless, because you never know, you know?).
718		2238
719	As another example, the Perl Coro module uses these hooks to integrate	2239	As another example, the Perl Coro module uses these hooks to integrate
720	coroutines into libev programs, by yielding to other active coroutines	2240	coroutines into libev programs, by yielding to other active coroutines
721	during each prepare and only letting the process block if no coroutines	2241	during each prepare and only letting the process block if no coroutines
722	are ready to run (it's actually more complicated: it only runs coroutines	2242	are ready to run (it's actually more complicated: it only runs coroutines
723	with priority higher than or equal to the event loop and one coroutine	2243	with priority higher than or equal to the event loop and one coroutine
724	of lower priority, but only once, using idle watchers to keep the event	2244	of lower priority, but only once, using idle watchers to keep the event
725	loop from blocking if lower-priority coroutines are active, thus mapping	2245	loop from blocking if lower-priority coroutines are active, thus mapping
726	low-priority coroutines to idle/background tasks).	2246	low-priority coroutines to idle/background tasks).
727		2247
		2248	It is recommended to give C<ev_check> watchers highest (C<EV_MAXPRI>)
		2249	priority, to ensure that they are being run before any other watchers
		2250	after the poll (this doesn't matter for C<ev_prepare> watchers).
		2251
		2252	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		2253	activate ("feed") events into libev. While libev fully supports this, they
		2254	might get executed before other C<ev_check> watchers did their job. As
		2255	C<ev_check> watchers are often used to embed other (non-libev) event
		2256	loops those other event loops might be in an unusable state until their
		2257	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		2258	others).
		2259
		2260	=head3 Watcher-Specific Functions and Data Members
		2261
728	=over 4	2262	=over 4
729		2263
730	=item ev_prepare_init (ev_prepare *, callback)	2264	=item ev_prepare_init (ev_prepare *, callback)
731		2265
732	=item ev_check_init (ev_check *, callback)	2266	=item ev_check_init (ev_check *, callback)
733		2267
734	Initialises and configures the prepare or check watcher - they have no	2268	Initialises and configures the prepare or check watcher - they have no
735	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	2269	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
736	macros, but using them is utterly, utterly and completely pointless.	2270	macros, but using them is utterly, utterly, utterly and completely
		2271	pointless.
737		2272
738	=back	2273	=back
739		2274
		2275	=head3 Examples
		2276
		2277	There are a number of principal ways to embed other event loops or modules
		2278	into libev. Here are some ideas on how to include libadns into libev
		2279	(there is a Perl module named C<EV::ADNS> that does this, which you could
		2280	use as a working example. Another Perl module named C<EV::Glib> embeds a
		2281	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		2282	Glib event loop).
		2283
		2284	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		2285	and in a check watcher, destroy them and call into libadns. What follows
		2286	is pseudo-code only of course. This requires you to either use a low
		2287	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		2288	the callbacks for the IO/timeout watchers might not have been called yet.
		2289
		2290	static ev_io iow [nfd];
		2291	static ev_timer tw;
		2292
		2293	static void
		2294	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		2295	{
		2296	}
		2297
		2298	// create io watchers for each fd and a timer before blocking
		2299	static void
		2300	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		2301	{
		2302	int timeout = 3600000;
		2303	struct pollfd fds [nfd];
		2304	// actual code will need to loop here and realloc etc.
		2305	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		2306
		2307	/* the callback is illegal, but won't be called as we stop during check */
		2308	ev_timer_init (&tw, 0, timeout * 1e-3);
		2309	ev_timer_start (loop, &tw);
		2310
		2311	// create one ev_io per pollfd
		2312	for (int i = 0; i < nfd; ++i)
		2313	{
		2314	ev_io_init (iow + i, io_cb, fds [i].fd,
		2315	((fds [i].events & POLLIN ? EV_READ : 0)
		2316	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		2317
		2318	fds [i].revents = 0;
		2319	ev_io_start (loop, iow + i);
		2320	}
		2321	}
		2322
		2323	// stop all watchers after blocking
		2324	static void
		2325	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		2326	{
		2327	ev_timer_stop (loop, &tw);
		2328
		2329	for (int i = 0; i < nfd; ++i)
		2330	{
		2331	// set the relevant poll flags
		2332	// could also call adns_processreadable etc. here
		2333	struct pollfd *fd = fds + i;
		2334	int revents = ev_clear_pending (iow + i);
		2335	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		2336	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		2337
		2338	// now stop the watcher
		2339	ev_io_stop (loop, iow + i);
		2340	}
		2341
		2342	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		2343	}
		2344
		2345	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		2346	in the prepare watcher and would dispose of the check watcher.
		2347
		2348	Method 3: If the module to be embedded supports explicit event
		2349	notification (libadns does), you can also make use of the actual watcher
		2350	callbacks, and only destroy/create the watchers in the prepare watcher.
		2351
		2352	static void
		2353	timer_cb (EV_P_ ev_timer *w, int revents)
		2354	{
		2355	adns_state ads = (adns_state)w->data;
		2356	update_now (EV_A);
		2357
		2358	adns_processtimeouts (ads, &tv_now);
		2359	}
		2360
		2361	static void
		2362	io_cb (EV_P_ ev_io *w, int revents)
		2363	{
		2364	adns_state ads = (adns_state)w->data;
		2365	update_now (EV_A);
		2366
		2367	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		2368	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		2369	}
		2370
		2371	// do not ever call adns_afterpoll
		2372
		2373	Method 4: Do not use a prepare or check watcher because the module you
		2374	want to embed is not flexible enough to support it. Instead, you can
		2375	override their poll function. The drawback with this solution is that the
		2376	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		2377	this approach, effectively embedding EV as a client into the horrible
		2378	libglib event loop.
		2379
		2380	static gint
		2381	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		2382	{
		2383	int got_events = 0;
		2384
		2385	for (n = 0; n < nfds; ++n)
		2386	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		2387
		2388	if (timeout >= 0)
		2389	// create/start timer
		2390
		2391	// poll
		2392	ev_loop (EV_A_ 0);
		2393
		2394	// stop timer again
		2395	if (timeout >= 0)
		2396	ev_timer_stop (EV_A_ &to);
		2397
		2398	// stop io watchers again - their callbacks should have set
		2399	for (n = 0; n < nfds; ++n)
		2400	ev_io_stop (EV_A_ iow [n]);
		2401
		2402	return got_events;
		2403	}
		2404
		2405
		2406	=head2 C<ev_embed> - when one backend isn't enough...
		2407
		2408	This is a rather advanced watcher type that lets you embed one event loop
		2409	into another (currently only C<ev_io> events are supported in the embedded
		2410	loop, other types of watchers might be handled in a delayed or incorrect
		2411	fashion and must not be used).
		2412
		2413	There are primarily two reasons you would want that: work around bugs and
		2414	prioritise I/O.
		2415
		2416	As an example for a bug workaround, the kqueue backend might only support
		2417	sockets on some platform, so it is unusable as generic backend, but you
		2418	still want to make use of it because you have many sockets and it scales
		2419	so nicely. In this case, you would create a kqueue-based loop and embed
		2420	it into your default loop (which might use e.g. poll). Overall operation
		2421	will be a bit slower because first libev has to call C<poll> and then
		2422	C<kevent>, but at least you can use both mechanisms for what they are
		2423	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		2424
		2425	As for prioritising I/O: under rare circumstances you have the case where
		2426	some fds have to be watched and handled very quickly (with low latency),
		2427	and even priorities and idle watchers might have too much overhead. In
		2428	this case you would put all the high priority stuff in one loop and all
		2429	the rest in a second one, and embed the second one in the first.
		2430
		2431	As long as the watcher is active, the callback will be invoked every time
		2432	there might be events pending in the embedded loop. The callback must then
		2433	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke
		2434	their callbacks (you could also start an idle watcher to give the embedded
		2435	loop strictly lower priority for example). You can also set the callback
		2436	to C<0>, in which case the embed watcher will automatically execute the
		2437	embedded loop sweep.
		2438
		2439	As long as the watcher is started it will automatically handle events. The
		2440	callback will be invoked whenever some events have been handled. You can
		2441	set the callback to C<0> to avoid having to specify one if you are not
		2442	interested in that.
		2443
		2444	Also, there have not currently been made special provisions for forking:
		2445	when you fork, you not only have to call C<ev_loop_fork> on both loops,
		2446	but you will also have to stop and restart any C<ev_embed> watchers
		2447	yourself - but you can use a fork watcher to handle this automatically,
		2448	and future versions of libev might do just that.
		2449
		2450	Unfortunately, not all backends are embeddable: only the ones returned by
		2451	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		2452	portable one.
		2453
		2454	So when you want to use this feature you will always have to be prepared
		2455	that you cannot get an embeddable loop. The recommended way to get around
		2456	this is to have a separate variables for your embeddable loop, try to
		2457	create it, and if that fails, use the normal loop for everything.
		2458
		2459	=head3 C<ev_embed> and fork
		2460
		2461	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2462	automatically be applied to the embedded loop as well, so no special
		2463	fork handling is required in that case. When the watcher is not running,
		2464	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2465	as applicable.
		2466
		2467	=head3 Watcher-Specific Functions and Data Members
		2468
		2469	=over 4
		2470
		2471	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		2472
		2473	=item ev_embed_set (ev_embed *, callback, struct ev_loop *embedded_loop)
		2474
		2475	Configures the watcher to embed the given loop, which must be
		2476	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		2477	invoked automatically, otherwise it is the responsibility of the callback
		2478	to invoke it (it will continue to be called until the sweep has been done,
		2479	if you do not want that, you need to temporarily stop the embed watcher).
		2480
		2481	=item ev_embed_sweep (loop, ev_embed *)
		2482
		2483	Make a single, non-blocking sweep over the embedded loop. This works
		2484	similarly to C<ev_loop (embedded_loop, EVLOOP_NONBLOCK)>, but in the most
		2485	appropriate way for embedded loops.
		2486
		2487	=item struct ev_loop *other [read-only]
		2488
		2489	The embedded event loop.
		2490
		2491	=back
		2492
		2493	=head3 Examples
		2494
		2495	Example: Try to get an embeddable event loop and embed it into the default
		2496	event loop. If that is not possible, use the default loop. The default
		2497	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		2498	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		2499	used).
		2500
		2501	struct ev_loop *loop_hi = ev_default_init (0);
		2502	struct ev_loop *loop_lo = 0;
		2503	ev_embed embed;
		2504
		2505	// see if there is a chance of getting one that works
		2506	// (remember that a flags value of 0 means autodetection)
		2507	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		2508	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		2509	: 0;
		2510
		2511	// if we got one, then embed it, otherwise default to loop_hi
		2512	if (loop_lo)
		2513	{
		2514	ev_embed_init (&embed, 0, loop_lo);
		2515	ev_embed_start (loop_hi, &embed);
		2516	}
		2517	else
		2518	loop_lo = loop_hi;
		2519
		2520	Example: Check if kqueue is available but not recommended and create
		2521	a kqueue backend for use with sockets (which usually work with any
		2522	kqueue implementation). Store the kqueue/socket-only event loop in
		2523	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		2524
		2525	struct ev_loop *loop = ev_default_init (0);
		2526	struct ev_loop *loop_socket = 0;
		2527	ev_embed embed;
		2528
		2529	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		2530	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		2531	{
		2532	ev_embed_init (&embed, 0, loop_socket);
		2533	ev_embed_start (loop, &embed);
		2534	}
		2535
		2536	if (!loop_socket)
		2537	loop_socket = loop;
		2538
		2539	// now use loop_socket for all sockets, and loop for everything else
		2540
		2541
		2542	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		2543
		2544	Fork watchers are called when a C<fork ()> was detected (usually because
		2545	whoever is a good citizen cared to tell libev about it by calling
		2546	C<ev_default_fork> or C<ev_loop_fork>). The invocation is done before the
		2547	event loop blocks next and before C<ev_check> watchers are being called,
		2548	and only in the child after the fork. If whoever good citizen calling
		2549	C<ev_default_fork> cheats and calls it in the wrong process, the fork
		2550	handlers will be invoked, too, of course.
		2551
		2552	=head3 Watcher-Specific Functions and Data Members
		2553
		2554	=over 4
		2555
		2556	=item ev_fork_init (ev_signal *, callback)
		2557
		2558	Initialises and configures the fork watcher - it has no parameters of any
		2559	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		2560	believe me.
		2561
		2562	=back
		2563
		2564
		2565	=head2 C<ev_async> - how to wake up another event loop
		2566
		2567	In general, you cannot use an C<ev_loop> from multiple threads or other
		2568	asynchronous sources such as signal handlers (as opposed to multiple event
		2569	loops - those are of course safe to use in different threads).
		2570
		2571	Sometimes, however, you need to wake up another event loop you do not
		2572	control, for example because it belongs to another thread. This is what
		2573	C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
		2574	can signal it by calling C<ev_async_send>, which is thread- and signal
		2575	safe.
		2576
		2577	This functionality is very similar to C<ev_signal> watchers, as signals,
		2578	too, are asynchronous in nature, and signals, too, will be compressed
		2579	(i.e. the number of callback invocations may be less than the number of
		2580	C<ev_async_sent> calls).
		2581
		2582	Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
		2583	just the default loop.
		2584
		2585	=head3 Queueing
		2586
		2587	C<ev_async> does not support queueing of data in any way. The reason
		2588	is that the author does not know of a simple (or any) algorithm for a
		2589	multiple-writer-single-reader queue that works in all cases and doesn't
		2590	need elaborate support such as pthreads.
		2591
		2592	That means that if you want to queue data, you have to provide your own
		2593	queue. But at least I can tell you how to implement locking around your
		2594	queue:
		2595
		2596	=over 4
		2597
		2598	=item queueing from a signal handler context
		2599
		2600	To implement race-free queueing, you simply add to the queue in the signal
		2601	handler but you block the signal handler in the watcher callback. Here is
		2602	an example that does that for some fictitious SIGUSR1 handler:
		2603
		2604	static ev_async mysig;
		2605
		2606	static void
		2607	sigusr1_handler (void)
		2608	{
		2609	sometype data;
		2610
		2611	// no locking etc.
		2612	queue_put (data);
		2613	ev_async_send (EV_DEFAULT_ &mysig);
		2614	}
		2615
		2616	static void
		2617	mysig_cb (EV_P_ ev_async *w, int revents)
		2618	{
		2619	sometype data;
		2620	sigset_t block, prev;
		2621
		2622	sigemptyset (&block);
		2623	sigaddset (&block, SIGUSR1);
		2624	sigprocmask (SIG_BLOCK, &block, &prev);
		2625
		2626	while (queue_get (&data))
		2627	process (data);
		2628
		2629	if (sigismember (&prev, SIGUSR1)
		2630	sigprocmask (SIG_UNBLOCK, &block, 0);
		2631	}
		2632
		2633	(Note: pthreads in theory requires you to use C<pthread_setmask>
		2634	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		2635	either...).
		2636
		2637	=item queueing from a thread context
		2638
		2639	The strategy for threads is different, as you cannot (easily) block
		2640	threads but you can easily preempt them, so to queue safely you need to
		2641	employ a traditional mutex lock, such as in this pthread example:
		2642
		2643	static ev_async mysig;
		2644	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		2645
		2646	static void
		2647	otherthread (void)
		2648	{
		2649	// only need to lock the actual queueing operation
		2650	pthread_mutex_lock (&mymutex);
		2651	queue_put (data);
		2652	pthread_mutex_unlock (&mymutex);
		2653
		2654	ev_async_send (EV_DEFAULT_ &mysig);
		2655	}
		2656
		2657	static void
		2658	mysig_cb (EV_P_ ev_async *w, int revents)
		2659	{
		2660	pthread_mutex_lock (&mymutex);
		2661
		2662	while (queue_get (&data))
		2663	process (data);
		2664
		2665	pthread_mutex_unlock (&mymutex);
		2666	}
		2667
		2668	=back
		2669
		2670
		2671	=head3 Watcher-Specific Functions and Data Members
		2672
		2673	=over 4
		2674
		2675	=item ev_async_init (ev_async *, callback)
		2676
		2677	Initialises and configures the async watcher - it has no parameters of any
		2678	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		2679	trust me.
		2680
		2681	=item ev_async_send (loop, ev_async *)
		2682
		2683	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		2684	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
		2685	C<ev_feed_event>, this call is safe to do from other threads, signal or
		2686	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
		2687	section below on what exactly this means).
		2688
		2689	This call incurs the overhead of a system call only once per loop iteration,
		2690	so while the overhead might be noticeable, it doesn't apply to repeated
		2691	calls to C<ev_async_send>.
		2692
		2693	=item bool = ev_async_pending (ev_async *)
		2694
		2695	Returns a non-zero value when C<ev_async_send> has been called on the
		2696	watcher but the event has not yet been processed (or even noted) by the
		2697	event loop.
		2698
		2699	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		2700	the loop iterates next and checks for the watcher to have become active,
		2701	it will reset the flag again. C<ev_async_pending> can be used to very
		2702	quickly check whether invoking the loop might be a good idea.
		2703
		2704	Not that this does I<not> check whether the watcher itself is pending, only
		2705	whether it has been requested to make this watcher pending.
		2706
		2707	=back
		2708
		2709
740	=head1 OTHER FUNCTIONS	2710	=head1 OTHER FUNCTIONS
741		2711
742	There are some other functions of possible interest. Described. Here. Now.	2712	There are some other functions of possible interest. Described. Here. Now.
743		2713
744	=over 4	2714	=over 4
745		2715
746	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2716	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
747		2717
748	This function combines a simple timer and an I/O watcher, calls your	2718	This function combines a simple timer and an I/O watcher, calls your
749	callback on whichever event happens first and automatically stop both	2719	callback on whichever event happens first and automatically stops both
750	watchers. This is useful if you want to wait for a single event on an fd	2720	watchers. This is useful if you want to wait for a single event on an fd
751	or timeout without having to allocate/configure/start/stop/free one or	2721	or timeout without having to allocate/configure/start/stop/free one or
752	more watchers yourself.	2722	more watchers yourself.
753		2723
754	If C<fd> is less than 0, then no I/O watcher will be started and events	2724	If C<fd> is less than 0, then no I/O watcher will be started and the
755	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2725	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
756	C<events> set will be craeted and started.	2726	the given C<fd> and C<events> set will be created and started.
757		2727
758	If C<timeout> is less than 0, then no timeout watcher will be	2728	If C<timeout> is less than 0, then no timeout watcher will be
759	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2729	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
760	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2730	repeat = 0) will be started. C<0> is a valid timeout.
761	dubious value.
762		2731
763	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	2732	The callback has the type C<void (*cb)(int revents, void *arg)> and gets
764	passed an C<revents> set like normal event callbacks (a combination of	2733	passed an C<revents> set like normal event callbacks (a combination of
765	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2734	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
766	value passed to C<ev_once>:	2735	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2736	a timeout and an io event at the same time - you probably should give io
		2737	events precedence.
767		2738
		2739	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		2740
768	static void stdin_ready (int revents, void *arg)	2741	static void stdin_ready (int revents, void *arg)
769	{	2742	{
770	if (revents & EV_TIMEOUT)
771	/* doh, nothing entered */;
772	else if (revents & EV_READ)	2743	if (revents & EV_READ)
773	/* stdin might have data for us, joy! */;	2744	/* stdin might have data for us, joy! */;
		2745	else if (revents & EV_TIMEOUT)
		2746	/* doh, nothing entered */;
774	}	2747	}
775		2748
776	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2749	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
777		2750
778	=item ev_feed_event (loop, watcher, int events)	2751	=item ev_feed_event (struct ev_loop *, watcher *, int revents)
779		2752
780	Feeds the given event set into the event loop, as if the specified event	2753	Feeds the given event set into the event loop, as if the specified event
781	had happened for the specified watcher (which must be a pointer to an	2754	had happened for the specified watcher (which must be a pointer to an
782	initialised but not necessarily started event watcher).	2755	initialised but not necessarily started event watcher).
783		2756
784	=item ev_feed_fd_event (loop, int fd, int revents)	2757	=item ev_feed_fd_event (struct ev_loop *, int fd, int revents)
785		2758
786	Feed an event on the given fd, as if a file descriptor backend detected	2759	Feed an event on the given fd, as if a file descriptor backend detected
787	the given events it.	2760	the given events it.
788		2761
789	=item ev_feed_signal_event (loop, int signum)	2762	=item ev_feed_signal_event (struct ev_loop *loop, int signum)
790		2763
791	Feed an event as if the given signal occured (loop must be the default loop!).	2764	Feed an event as if the given signal occurred (C<loop> must be the default
		2765	loop!).
792		2766
793	=back	2767	=back
		2768
794		2769
795	=head1 LIBEVENT EMULATION	2770	=head1 LIBEVENT EMULATION
796		2771
797	Libev offers a compatibility emulation layer for libevent. It cannot	2772	Libev offers a compatibility emulation layer for libevent. It cannot
798	emulate the internals of libevent, so here are some usage hints:	2773	emulate the internals of libevent, so here are some usage hints:
…		…
810		2785
811	=item * Priorities are not currently supported. Initialising priorities	2786	=item * Priorities are not currently supported. Initialising priorities
812	will fail and all watchers will have the same priority, even though there	2787	will fail and all watchers will have the same priority, even though there
813	is an ev_pri field.	2788	is an ev_pri field.
814		2789
		2790	=item * In libevent, the last base created gets the signals, in libev, the
		2791	first base created (== the default loop) gets the signals.
		2792
815	=item * Other members are not supported.	2793	=item * Other members are not supported.
816		2794
817	=item * The libev emulation is I<not> ABI compatible to libevent, you need	2795	=item * The libev emulation is I<not> ABI compatible to libevent, you need
818	to use the libev header file and library.	2796	to use the libev header file and library.
819		2797
820	=back	2798	=back
821		2799
822	=head1 C++ SUPPORT	2800	=head1 C++ SUPPORT
823		2801
824	TBD.	2802	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		2803	you to use some convenience methods to start/stop watchers and also change
		2804	the callback model to a model using method callbacks on objects.
		2805
		2806	To use it,
		2807
		2808	#include <ev++.h>
		2809
		2810	This automatically includes F<ev.h> and puts all of its definitions (many
		2811	of them macros) into the global namespace. All C++ specific things are
		2812	put into the C<ev> namespace. It should support all the same embedding
		2813	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		2814
		2815	Care has been taken to keep the overhead low. The only data member the C++
		2816	classes add (compared to plain C-style watchers) is the event loop pointer
		2817	that the watcher is associated with (or no additional members at all if
		2818	you disable C<EV_MULTIPLICITY> when embedding libev).
		2819
		2820	Currently, functions, and static and non-static member functions can be
		2821	used as callbacks. Other types should be easy to add as long as they only
		2822	need one additional pointer for context. If you need support for other
		2823	types of functors please contact the author (preferably after implementing
		2824	it).
		2825
		2826	Here is a list of things available in the C<ev> namespace:
		2827
		2828	=over 4
		2829
		2830	=item C<ev::READ>, C<ev::WRITE> etc.
		2831
		2832	These are just enum values with the same values as the C<EV_READ> etc.
		2833	macros from F<ev.h>.
		2834
		2835	=item C<ev::tstamp>, C<ev::now>
		2836
		2837	Aliases to the same types/functions as with the C<ev_> prefix.
		2838
		2839	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		2840
		2841	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		2842	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		2843	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		2844	defines by many implementations.
		2845
		2846	All of those classes have these methods:
		2847
		2848	=over 4
		2849
		2850	=item ev::TYPE::TYPE ()
		2851
		2852	=item ev::TYPE::TYPE (struct ev_loop *)
		2853
		2854	=item ev::TYPE::~TYPE
		2855
		2856	The constructor (optionally) takes an event loop to associate the watcher
		2857	with. If it is omitted, it will use C<EV_DEFAULT>.
		2858
		2859	The constructor calls C<ev_init> for you, which means you have to call the
		2860	C<set> method before starting it.
		2861
		2862	It will not set a callback, however: You have to call the templated C<set>
		2863	method to set a callback before you can start the watcher.
		2864
		2865	(The reason why you have to use a method is a limitation in C++ which does
		2866	not allow explicit template arguments for constructors).
		2867
		2868	The destructor automatically stops the watcher if it is active.
		2869
		2870	=item w->set<class, &class::method> (object *)
		2871
		2872	This method sets the callback method to call. The method has to have a
		2873	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		2874	first argument and the C<revents> as second. The object must be given as
		2875	parameter and is stored in the C<data> member of the watcher.
		2876
		2877	This method synthesizes efficient thunking code to call your method from
		2878	the C callback that libev requires. If your compiler can inline your
		2879	callback (i.e. it is visible to it at the place of the C<set> call and
		2880	your compiler is good :), then the method will be fully inlined into the
		2881	thunking function, making it as fast as a direct C callback.
		2882
		2883	Example: simple class declaration and watcher initialisation
		2884
		2885	struct myclass
		2886	{
		2887	void io_cb (ev::io &w, int revents) { }
		2888	}
		2889
		2890	myclass obj;
		2891	ev::io iow;
		2892	iow.set <myclass, &myclass::io_cb> (&obj);
		2893
		2894	=item w->set<function> (void *data = 0)
		2895
		2896	Also sets a callback, but uses a static method or plain function as
		2897	callback. The optional C<data> argument will be stored in the watcher's
		2898	C<data> member and is free for you to use.
		2899
		2900	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		2901
		2902	See the method-C<set> above for more details.
		2903
		2904	Example: Use a plain function as callback.
		2905
		2906	static void io_cb (ev::io &w, int revents) { }
		2907	iow.set <io_cb> ();
		2908
		2909	=item w->set (struct ev_loop *)
		2910
		2911	Associates a different C<struct ev_loop> with this watcher. You can only
		2912	do this when the watcher is inactive (and not pending either).
		2913
		2914	=item w->set ([arguments])
		2915
		2916	Basically the same as C<ev_TYPE_set>, with the same arguments. Must be
		2917	called at least once. Unlike the C counterpart, an active watcher gets
		2918	automatically stopped and restarted when reconfiguring it with this
		2919	method.
		2920
		2921	=item w->start ()
		2922
		2923	Starts the watcher. Note that there is no C<loop> argument, as the
		2924	constructor already stores the event loop.
		2925
		2926	=item w->stop ()
		2927
		2928	Stops the watcher if it is active. Again, no C<loop> argument.
		2929
		2930	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		2931
		2932	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		2933	C<ev_TYPE_again> function.
		2934
		2935	=item w->sweep () (C<ev::embed> only)
		2936
		2937	Invokes C<ev_embed_sweep>.
		2938
		2939	=item w->update () (C<ev::stat> only)
		2940
		2941	Invokes C<ev_stat_stat>.
		2942
		2943	=back
		2944
		2945	=back
		2946
		2947	Example: Define a class with an IO and idle watcher, start one of them in
		2948	the constructor.
		2949
		2950	class myclass
		2951	{
		2952	ev::io io ; void io_cb (ev::io &w, int revents);
		2953	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		2954
		2955	myclass (int fd)
		2956	{
		2957	io .set <myclass, &myclass::io_cb > (this);
		2958	idle.set <myclass, &myclass::idle_cb> (this);
		2959
		2960	io.start (fd, ev::READ);
		2961	}
		2962	};
		2963
		2964
		2965	=head1 OTHER LANGUAGE BINDINGS
		2966
		2967	Libev does not offer other language bindings itself, but bindings for a
		2968	number of languages exist in the form of third-party packages. If you know
		2969	any interesting language binding in addition to the ones listed here, drop
		2970	me a note.
		2971
		2972	=over 4
		2973
		2974	=item Perl
		2975
		2976	The EV module implements the full libev API and is actually used to test
		2977	libev. EV is developed together with libev. Apart from the EV core module,
		2978	there are additional modules that implement libev-compatible interfaces
		2979	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		2980	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		2981	and C<EV::Glib>).
		2982
		2983	It can be found and installed via CPAN, its homepage is at
		2984	L<http://software.schmorp.de/pkg/EV>.
		2985
		2986	=item Python
		2987
		2988	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		2989	seems to be quite complete and well-documented. Note, however, that the
		2990	patch they require for libev is outright dangerous as it breaks the ABI
		2991	for everybody else, and therefore, should never be applied in an installed
		2992	libev (if python requires an incompatible ABI then it needs to embed
		2993	libev).
		2994
		2995	=item Ruby
		2996
		2997	Tony Arcieri has written a ruby extension that offers access to a subset
		2998	of the libev API and adds file handle abstractions, asynchronous DNS and
		2999	more on top of it. It can be found via gem servers. Its homepage is at
		3000	L<http://rev.rubyforge.org/>.
		3001
		3002	=item D
		3003
		3004	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		3005	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3006
		3007	=item Ocaml
		3008
		3009	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3010	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		3011
		3012	=back
		3013
		3014
		3015	=head1 MACRO MAGIC
		3016
		3017	Libev can be compiled with a variety of options, the most fundamental
		3018	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		3019	functions and callbacks have an initial C<struct ev_loop *> argument.
		3020
		3021	To make it easier to write programs that cope with either variant, the
		3022	following macros are defined:
		3023
		3024	=over 4
		3025
		3026	=item C<EV_A>, C<EV_A_>
		3027
		3028	This provides the loop I<argument> for functions, if one is required ("ev
		3029	loop argument"). The C<EV_A> form is used when this is the sole argument,
		3030	C<EV_A_> is used when other arguments are following. Example:
		3031
		3032	ev_unref (EV_A);
		3033	ev_timer_add (EV_A_ watcher);
		3034	ev_loop (EV_A_ 0);
		3035
		3036	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		3037	which is often provided by the following macro.
		3038
		3039	=item C<EV_P>, C<EV_P_>
		3040
		3041	This provides the loop I<parameter> for functions, if one is required ("ev
		3042	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		3043	C<EV_P_> is used when other parameters are following. Example:
		3044
		3045	// this is how ev_unref is being declared
		3046	static void ev_unref (EV_P);
		3047
		3048	// this is how you can declare your typical callback
		3049	static void cb (EV_P_ ev_timer *w, int revents)
		3050
		3051	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		3052	suitable for use with C<EV_A>.
		3053
		3054	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		3055
		3056	Similar to the other two macros, this gives you the value of the default
		3057	loop, if multiple loops are supported ("ev loop default").
		3058
		3059	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		3060
		3061	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		3062	default loop has been initialised (C<UC> == unchecked). Their behaviour
		3063	is undefined when the default loop has not been initialised by a previous
		3064	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		3065
		3066	It is often prudent to use C<EV_DEFAULT> when initialising the first
		3067	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		3068
		3069	=back
		3070
		3071	Example: Declare and initialise a check watcher, utilising the above
		3072	macros so it will work regardless of whether multiple loops are supported
		3073	or not.
		3074
		3075	static void
		3076	check_cb (EV_P_ ev_timer *w, int revents)
		3077	{
		3078	ev_check_stop (EV_A_ w);
		3079	}
		3080
		3081	ev_check check;
		3082	ev_check_init (&check, check_cb);
		3083	ev_check_start (EV_DEFAULT_ &check);
		3084	ev_loop (EV_DEFAULT_ 0);
		3085
		3086	=head1 EMBEDDING
		3087
		3088	Libev can (and often is) directly embedded into host
		3089	applications. Examples of applications that embed it include the Deliantra
		3090	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		3091	and rxvt-unicode.
		3092
		3093	The goal is to enable you to just copy the necessary files into your
		3094	source directory without having to change even a single line in them, so
		3095	you can easily upgrade by simply copying (or having a checked-out copy of
		3096	libev somewhere in your source tree).
		3097
		3098	=head2 FILESETS
		3099
		3100	Depending on what features you need you need to include one or more sets of files
		3101	in your application.
		3102
		3103	=head3 CORE EVENT LOOP
		3104
		3105	To include only the libev core (all the C<ev_*> functions), with manual
		3106	configuration (no autoconf):
		3107
		3108	#define EV_STANDALONE 1
		3109	#include "ev.c"
		3110
		3111	This will automatically include F<ev.h>, too, and should be done in a
		3112	single C source file only to provide the function implementations. To use
		3113	it, do the same for F<ev.h> in all files wishing to use this API (best
		3114	done by writing a wrapper around F<ev.h> that you can include instead and
		3115	where you can put other configuration options):
		3116
		3117	#define EV_STANDALONE 1
		3118	#include "ev.h"
		3119
		3120	Both header files and implementation files can be compiled with a C++
		3121	compiler (at least, that's a stated goal, and breakage will be treated
		3122	as a bug).
		3123
		3124	You need the following files in your source tree, or in a directory
		3125	in your include path (e.g. in libev/ when using -Ilibev):
		3126
		3127	ev.h
		3128	ev.c
		3129	ev_vars.h
		3130	ev_wrap.h
		3131
		3132	ev_win32.c required on win32 platforms only
		3133
		3134	ev_select.c only when select backend is enabled (which is enabled by default)
		3135	ev_poll.c only when poll backend is enabled (disabled by default)
		3136	ev_epoll.c only when the epoll backend is enabled (disabled by default)
		3137	ev_kqueue.c only when the kqueue backend is enabled (disabled by default)
		3138	ev_port.c only when the solaris port backend is enabled (disabled by default)
		3139
		3140	F<ev.c> includes the backend files directly when enabled, so you only need
		3141	to compile this single file.
		3142
		3143	=head3 LIBEVENT COMPATIBILITY API
		3144
		3145	To include the libevent compatibility API, also include:
		3146
		3147	#include "event.c"
		3148
		3149	in the file including F<ev.c>, and:
		3150
		3151	#include "event.h"
		3152
		3153	in the files that want to use the libevent API. This also includes F<ev.h>.
		3154
		3155	You need the following additional files for this:
		3156
		3157	event.h
		3158	event.c
		3159
		3160	=head3 AUTOCONF SUPPORT
		3161
		3162	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		3163	whatever way you want, you can also C<m4_include([libev.m4])> in your
		3164	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		3165	include F<config.h> and configure itself accordingly.
		3166
		3167	For this of course you need the m4 file:
		3168
		3169	libev.m4
		3170
		3171	=head2 PREPROCESSOR SYMBOLS/MACROS
		3172
		3173	Libev can be configured via a variety of preprocessor symbols you have to
		3174	define before including any of its files. The default in the absence of
		3175	autoconf is documented for every option.
		3176
		3177	=over 4
		3178
		3179	=item EV_STANDALONE
		3180
		3181	Must always be C<1> if you do not use autoconf configuration, which
		3182	keeps libev from including F<config.h>, and it also defines dummy
		3183	implementations for some libevent functions (such as logging, which is not
		3184	supported). It will also not define any of the structs usually found in
		3185	F<event.h> that are not directly supported by the libev core alone.
		3186
		3187	=item EV_USE_MONOTONIC
		3188
		3189	If defined to be C<1>, libev will try to detect the availability of the
		3190	monotonic clock option at both compile time and runtime. Otherwise no use
		3191	of the monotonic clock option will be attempted. If you enable this, you
		3192	usually have to link against librt or something similar. Enabling it when
		3193	the functionality isn't available is safe, though, although you have
		3194	to make sure you link against any libraries where the C<clock_gettime>
		3195	function is hiding in (often F<-lrt>).
		3196
		3197	=item EV_USE_REALTIME
		3198
		3199	If defined to be C<1>, libev will try to detect the availability of the
		3200	real-time clock option at compile time (and assume its availability at
		3201	runtime if successful). Otherwise no use of the real-time clock option will
		3202	be attempted. This effectively replaces C<gettimeofday> by C<clock_get
		3203	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the
		3204	note about libraries in the description of C<EV_USE_MONOTONIC>, though.
		3205
		3206	=item EV_USE_NANOSLEEP
		3207
		3208	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		3209	and will use it for delays. Otherwise it will use C<select ()>.
		3210
		3211	=item EV_USE_EVENTFD
		3212
		3213	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		3214	available and will probe for kernel support at runtime. This will improve
		3215	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		3216	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		3217	2.7 or newer, otherwise disabled.
		3218
		3219	=item EV_USE_SELECT
		3220
		3221	If undefined or defined to be C<1>, libev will compile in support for the
		3222	C<select>(2) backend. No attempt at auto-detection will be done: if no
		3223	other method takes over, select will be it. Otherwise the select backend
		3224	will not be compiled in.
		3225
		3226	=item EV_SELECT_USE_FD_SET
		3227
		3228	If defined to C<1>, then the select backend will use the system C<fd_set>
		3229	structure. This is useful if libev doesn't compile due to a missing
		3230	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on
		3231	exotic systems. This usually limits the range of file descriptors to some
		3232	low limit such as 1024 or might have other limitations (winsocket only
		3233	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might
		3234	influence the size of the C<fd_set> used.
		3235
		3236	=item EV_SELECT_IS_WINSOCKET
		3237
		3238	When defined to C<1>, the select backend will assume that
		3239	select/socket/connect etc. don't understand file descriptors but
		3240	wants osf handles on win32 (this is the case when the select to
		3241	be used is the winsock select). This means that it will call
		3242	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		3243	it is assumed that all these functions actually work on fds, even
		3244	on win32. Should not be defined on non-win32 platforms.
		3245
		3246	=item EV_FD_TO_WIN32_HANDLE
		3247
		3248	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		3249	file descriptors to socket handles. When not defining this symbol (the
		3250	default), then libev will call C<_get_osfhandle>, which is usually
		3251	correct. In some cases, programs use their own file descriptor management,
		3252	in which case they can provide this function to map fds to socket handles.
		3253
		3254	=item EV_USE_POLL
		3255
		3256	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		3257	backend. Otherwise it will be enabled on non-win32 platforms. It
		3258	takes precedence over select.
		3259
		3260	=item EV_USE_EPOLL
		3261
		3262	If defined to be C<1>, libev will compile in support for the Linux
		3263	C<epoll>(7) backend. Its availability will be detected at runtime,
		3264	otherwise another method will be used as fallback. This is the preferred
		3265	backend for GNU/Linux systems. If undefined, it will be enabled if the
		3266	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3267
		3268	=item EV_USE_KQUEUE
		3269
		3270	If defined to be C<1>, libev will compile in support for the BSD style
		3271	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		3272	otherwise another method will be used as fallback. This is the preferred
		3273	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		3274	supports some types of fds correctly (the only platform we found that
		3275	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		3276	not be used unless explicitly requested. The best way to use it is to find
		3277	out whether kqueue supports your type of fd properly and use an embedded
		3278	kqueue loop.
		3279
		3280	=item EV_USE_PORT
		3281
		3282	If defined to be C<1>, libev will compile in support for the Solaris
		3283	10 port style backend. Its availability will be detected at runtime,
		3284	otherwise another method will be used as fallback. This is the preferred
		3285	backend for Solaris 10 systems.
		3286
		3287	=item EV_USE_DEVPOLL
		3288
		3289	Reserved for future expansion, works like the USE symbols above.
		3290
		3291	=item EV_USE_INOTIFY
		3292
		3293	If defined to be C<1>, libev will compile in support for the Linux inotify
		3294	interface to speed up C<ev_stat> watchers. Its actual availability will
		3295	be detected at runtime. If undefined, it will be enabled if the headers
		3296	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		3297
		3298	=item EV_ATOMIC_T
		3299
		3300	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		3301	access is atomic with respect to other threads or signal contexts. No such
		3302	type is easily found in the C language, so you can provide your own type
		3303	that you know is safe for your purposes. It is used both for signal handler "locking"
		3304	as well as for signal and thread safety in C<ev_async> watchers.
		3305
		3306	In the absence of this define, libev will use C<sig_atomic_t volatile>
		3307	(from F<signal.h>), which is usually good enough on most platforms.
		3308
		3309	=item EV_H
		3310
		3311	The name of the F<ev.h> header file used to include it. The default if
		3312	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		3313	used to virtually rename the F<ev.h> header file in case of conflicts.
		3314
		3315	=item EV_CONFIG_H
		3316
		3317	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		3318	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		3319	C<EV_H>, above.
		3320
		3321	=item EV_EVENT_H
		3322
		3323	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		3324	of how the F<event.h> header can be found, the default is C<"event.h">.
		3325
		3326	=item EV_PROTOTYPES
		3327
		3328	If defined to be C<0>, then F<ev.h> will not define any function
		3329	prototypes, but still define all the structs and other symbols. This is
		3330	occasionally useful if you want to provide your own wrapper functions
		3331	around libev functions.
		3332
		3333	=item EV_MULTIPLICITY
		3334
		3335	If undefined or defined to C<1>, then all event-loop-specific functions
		3336	will have the C<struct ev_loop *> as first argument, and you can create
		3337	additional independent event loops. Otherwise there will be no support
		3338	for multiple event loops and there is no first event loop pointer
		3339	argument. Instead, all functions act on the single default loop.
		3340
		3341	=item EV_MINPRI
		3342
		3343	=item EV_MAXPRI
		3344
		3345	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		3346	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		3347	provide for more priorities by overriding those symbols (usually defined
		3348	to be C<-2> and C<2>, respectively).
		3349
		3350	When doing priority-based operations, libev usually has to linearly search
		3351	all the priorities, so having many of them (hundreds) uses a lot of space
		3352	and time, so using the defaults of five priorities (-2 .. +2) is usually
		3353	fine.
		3354
		3355	If your embedding application does not need any priorities, defining these
		3356	both to C<0> will save some memory and CPU.
		3357
		3358	=item EV_PERIODIC_ENABLE
		3359
		3360	If undefined or defined to be C<1>, then periodic timers are supported. If
		3361	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3362	code.
		3363
		3364	=item EV_IDLE_ENABLE
		3365
		3366	If undefined or defined to be C<1>, then idle watchers are supported. If
		3367	defined to be C<0>, then they are not. Disabling them saves a few kB of
		3368	code.
		3369
		3370	=item EV_EMBED_ENABLE
		3371
		3372	If undefined or defined to be C<1>, then embed watchers are supported. If
		3373	defined to be C<0>, then they are not. Embed watchers rely on most other
		3374	watcher types, which therefore must not be disabled.
		3375
		3376	=item EV_STAT_ENABLE
		3377
		3378	If undefined or defined to be C<1>, then stat watchers are supported. If
		3379	defined to be C<0>, then they are not.
		3380
		3381	=item EV_FORK_ENABLE
		3382
		3383	If undefined or defined to be C<1>, then fork watchers are supported. If
		3384	defined to be C<0>, then they are not.
		3385
		3386	=item EV_ASYNC_ENABLE
		3387
		3388	If undefined or defined to be C<1>, then async watchers are supported. If
		3389	defined to be C<0>, then they are not.
		3390
		3391	=item EV_MINIMAL
		3392
		3393	If you need to shave off some kilobytes of code at the expense of some
		3394	speed, define this symbol to C<1>. Currently this is used to override some
		3395	inlining decisions, saves roughly 30% code size on amd64. It also selects a
		3396	much smaller 2-heap for timer management over the default 4-heap.
		3397
		3398	=item EV_PID_HASHSIZE
		3399
		3400	C<ev_child> watchers use a small hash table to distribute workload by
		3401	pid. The default size is C<16> (or C<1> with C<EV_MINIMAL>), usually more
		3402	than enough. If you need to manage thousands of children you might want to
		3403	increase this value (I<must> be a power of two).
		3404
		3405	=item EV_INOTIFY_HASHSIZE
		3406
		3407	C<ev_stat> watchers use a small hash table to distribute workload by
		3408	inotify watch id. The default size is C<16> (or C<1> with C<EV_MINIMAL>),
		3409	usually more than enough. If you need to manage thousands of C<ev_stat>
		3410	watchers you might want to increase this value (I<must> be a power of
		3411	two).
		3412
		3413	=item EV_USE_4HEAP
		3414
		3415	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3416	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		3417	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		3418	faster performance with many (thousands) of watchers.
		3419
		3420	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3421	(disabled).
		3422
		3423	=item EV_HEAP_CACHE_AT
		3424
		3425	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		3426	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		3427	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		3428	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		3429	but avoids random read accesses on heap changes. This improves performance
		3430	noticeably with many (hundreds) of watchers.
		3431
		3432	The default is C<1> unless C<EV_MINIMAL> is set in which case it is C<0>
		3433	(disabled).
		3434
		3435	=item EV_VERIFY
		3436
		3437	Controls how much internal verification (see C<ev_loop_verify ()>) will
		3438	be done: If set to C<0>, no internal verification code will be compiled
		3439	in. If set to C<1>, then verification code will be compiled in, but not
		3440	called. If set to C<2>, then the internal verification code will be
		3441	called once per loop, which can slow down libev. If set to C<3>, then the
		3442	verification code will be called very frequently, which will slow down
		3443	libev considerably.
		3444
		3445	The default is C<1>, unless C<EV_MINIMAL> is set, in which case it will be
		3446	C<0>.
		3447
		3448	=item EV_COMMON
		3449
		3450	By default, all watchers have a C<void *data> member. By redefining
		3451	this macro to a something else you can include more and other types of
		3452	members. You have to define it each time you include one of the files,
		3453	though, and it must be identical each time.
		3454
		3455	For example, the perl EV module uses something like this:
		3456
		3457	#define EV_COMMON \
		3458	SV *self; /* contains this struct */ \
		3459	SV *cb_sv, *fh /* note no trailing ";" */
		3460
		3461	=item EV_CB_DECLARE (type)
		3462
		3463	=item EV_CB_INVOKE (watcher, revents)
		3464
		3465	=item ev_set_cb (ev, cb)
		3466
		3467	Can be used to change the callback member declaration in each watcher,
		3468	and the way callbacks are invoked and set. Must expand to a struct member
		3469	definition and a statement, respectively. See the F<ev.h> header file for
		3470	their default definitions. One possible use for overriding these is to
		3471	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		3472	method calls instead of plain function calls in C++.
		3473
		3474	=back
		3475
		3476	=head2 EXPORTED API SYMBOLS
		3477
		3478	If you need to re-export the API (e.g. via a DLL) and you need a list of
		3479	exported symbols, you can use the provided F<Symbol.*> files which list
		3480	all public symbols, one per line:
		3481
		3482	Symbols.ev for libev proper
		3483	Symbols.event for the libevent emulation
		3484
		3485	This can also be used to rename all public symbols to avoid clashes with
		3486	multiple versions of libev linked together (which is obviously bad in
		3487	itself, but sometimes it is inconvenient to avoid this).
		3488
		3489	A sed command like this will create wrapper C<#define>'s that you need to
		3490	include before including F<ev.h>:
		3491
		3492	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		3493
		3494	This would create a file F<wrap.h> which essentially looks like this:
		3495
		3496	#define ev_backend myprefix_ev_backend
		3497	#define ev_check_start myprefix_ev_check_start
		3498	#define ev_check_stop myprefix_ev_check_stop
		3499	...
		3500
		3501	=head2 EXAMPLES
		3502
		3503	For a real-world example of a program the includes libev
		3504	verbatim, you can have a look at the EV perl module
		3505	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		3506	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		3507	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		3508	will be compiled. It is pretty complex because it provides its own header
		3509	file.
		3510
		3511	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		3512	that everybody includes and which overrides some configure choices:
		3513
		3514	#define EV_MINIMAL 1
		3515	#define EV_USE_POLL 0
		3516	#define EV_MULTIPLICITY 0
		3517	#define EV_PERIODIC_ENABLE 0
		3518	#define EV_STAT_ENABLE 0
		3519	#define EV_FORK_ENABLE 0
		3520	#define EV_CONFIG_H <config.h>
		3521	#define EV_MINPRI 0
		3522	#define EV_MAXPRI 0
		3523
		3524	#include "ev++.h"
		3525
		3526	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		3527
		3528	#include "ev_cpp.h"
		3529	#include "ev.c"
		3530
		3531	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
		3532
		3533	=head2 THREADS AND COROUTINES
		3534
		3535	=head3 THREADS
		3536
		3537	All libev functions are reentrant and thread-safe unless explicitly
		3538	documented otherwise, but libev implements no locking itself. This means
		3539	that you can use as many loops as you want in parallel, as long as there
		3540	are no concurrent calls into any libev function with the same loop
		3541	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		3542	of course): libev guarantees that different event loops share no data
		3543	structures that need any locking.
		3544
		3545	Or to put it differently: calls with different loop parameters can be done
		3546	concurrently from multiple threads, calls with the same loop parameter
		3547	must be done serially (but can be done from different threads, as long as
		3548	only one thread ever is inside a call at any point in time, e.g. by using
		3549	a mutex per loop).
		3550
		3551	Specifically to support threads (and signal handlers), libev implements
		3552	so-called C<ev_async> watchers, which allow some limited form of
		3553	concurrency on the same event loop, namely waking it up "from the
		3554	outside".
		3555
		3556	If you want to know which design (one loop, locking, or multiple loops
		3557	without or something else still) is best for your problem, then I cannot
		3558	help you, but here is some generic advice:
		3559
		3560	=over 4
		3561
		3562	=item * most applications have a main thread: use the default libev loop
		3563	in that thread, or create a separate thread running only the default loop.
		3564
		3565	This helps integrating other libraries or software modules that use libev
		3566	themselves and don't care/know about threading.
		3567
		3568	=item * one loop per thread is usually a good model.
		3569
		3570	Doing this is almost never wrong, sometimes a better-performance model
		3571	exists, but it is always a good start.
		3572
		3573	=item * other models exist, such as the leader/follower pattern, where one
		3574	loop is handed through multiple threads in a kind of round-robin fashion.
		3575
		3576	Choosing a model is hard - look around, learn, know that usually you can do
		3577	better than you currently do :-)
		3578
		3579	=item * often you need to talk to some other thread which blocks in the
		3580	event loop.
		3581
		3582	C<ev_async> watchers can be used to wake them up from other threads safely
		3583	(or from signal contexts...).
		3584
		3585	An example use would be to communicate signals or other events that only
		3586	work in the default loop by registering the signal watcher with the
		3587	default loop and triggering an C<ev_async> watcher from the default loop
		3588	watcher callback into the event loop interested in the signal.
		3589
		3590	=back
		3591
		3592	=head3 COROUTINES
		3593
		3594	Libev is very accommodating to coroutines ("cooperative threads"):
		3595	libev fully supports nesting calls to its functions from different
		3596	coroutines (e.g. you can call C<ev_loop> on the same loop from two
		3597	different coroutines, and switch freely between both coroutines running the
		3598	loop, as long as you don't confuse yourself). The only exception is that
		3599	you must not do this from C<ev_periodic> reschedule callbacks.
		3600
		3601	Care has been taken to ensure that libev does not keep local state inside
		3602	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3603	they do not call any callbacks.
		3604
		3605	=head2 COMPILER WARNINGS
		3606
		3607	Depending on your compiler and compiler settings, you might get no or a
		3608	lot of warnings when compiling libev code. Some people are apparently
		3609	scared by this.
		3610
		3611	However, these are unavoidable for many reasons. For one, each compiler
		3612	has different warnings, and each user has different tastes regarding
		3613	warning options. "Warn-free" code therefore cannot be a goal except when
		3614	targeting a specific compiler and compiler-version.
		3615
		3616	Another reason is that some compiler warnings require elaborate
		3617	workarounds, or other changes to the code that make it less clear and less
		3618	maintainable.
		3619
		3620	And of course, some compiler warnings are just plain stupid, or simply
		3621	wrong (because they don't actually warn about the condition their message
		3622	seems to warn about). For example, certain older gcc versions had some
		3623	warnings that resulted an extreme number of false positives. These have
		3624	been fixed, but some people still insist on making code warn-free with
		3625	such buggy versions.
		3626
		3627	While libev is written to generate as few warnings as possible,
		3628	"warn-free" code is not a goal, and it is recommended not to build libev
		3629	with any compiler warnings enabled unless you are prepared to cope with
		3630	them (e.g. by ignoring them). Remember that warnings are just that:
		3631	warnings, not errors, or proof of bugs.
		3632
		3633
		3634	=head2 VALGRIND
		3635
		3636	Valgrind has a special section here because it is a popular tool that is
		3637	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		3638
		3639	If you think you found a bug (memory leak, uninitialised data access etc.)
		3640	in libev, then check twice: If valgrind reports something like:
		3641
		3642	==2274== definitely lost: 0 bytes in 0 blocks.
		3643	==2274== possibly lost: 0 bytes in 0 blocks.
		3644	==2274== still reachable: 256 bytes in 1 blocks.
		3645
		3646	Then there is no memory leak, just as memory accounted to global variables
		3647	is not a memleak - the memory is still being referenced, and didn't leak.
		3648
		3649	Similarly, under some circumstances, valgrind might report kernel bugs
		3650	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3651	although an acceptable workaround has been found here), or it might be
		3652	confused.
		3653
		3654	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		3655	make it into some kind of religion.
		3656
		3657	If you are unsure about something, feel free to contact the mailing list
		3658	with the full valgrind report and an explanation on why you think this
		3659	is a bug in libev (best check the archives, too :). However, don't be
		3660	annoyed when you get a brisk "this is no bug" answer and take the chance
		3661	of learning how to interpret valgrind properly.
		3662
		3663	If you need, for some reason, empty reports from valgrind for your project
		3664	I suggest using suppression lists.
		3665
		3666
		3667	=head1 PORTABILITY NOTES
		3668
		3669	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		3670
		3671	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		3672	requires, and its I/O model is fundamentally incompatible with the POSIX
		3673	model. Libev still offers limited functionality on this platform in
		3674	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		3675	descriptors. This only applies when using Win32 natively, not when using
		3676	e.g. cygwin.
		3677
		3678	Lifting these limitations would basically require the full
		3679	re-implementation of the I/O system. If you are into these kinds of
		3680	things, then note that glib does exactly that for you in a very portable
		3681	way (note also that glib is the slowest event library known to man).
		3682
		3683	There is no supported compilation method available on windows except
		3684	embedding it into other applications.
		3685
		3686	Not a libev limitation but worth mentioning: windows apparently doesn't
		3687	accept large writes: instead of resulting in a partial write, windows will
		3688	either accept everything or return C<ENOBUFS> if the buffer is too large,
		3689	so make sure you only write small amounts into your sockets (less than a
		3690	megabyte seems safe, but this apparently depends on the amount of memory
		3691	available).
		3692
		3693	Due to the many, low, and arbitrary limits on the win32 platform and
		3694	the abysmal performance of winsockets, using a large number of sockets
		3695	is not recommended (and not reasonable). If your program needs to use
		3696	more than a hundred or so sockets, then likely it needs to use a totally
		3697	different implementation for windows, as libev offers the POSIX readiness
		3698	notification model, which cannot be implemented efficiently on windows
		3699	(Microsoft monopoly games).
		3700
		3701	A typical way to use libev under windows is to embed it (see the embedding
		3702	section for details) and use the following F<evwrap.h> header file instead
		3703	of F<ev.h>:
		3704
		3705	#define EV_STANDALONE /* keeps ev from requiring config.h */
		3706	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		3707
		3708	#include "ev.h"
		3709
		3710	And compile the following F<evwrap.c> file into your project (make sure
		3711	you do I<not> compile the F<ev.c> or any other embedded source files!):
		3712
		3713	#include "evwrap.h"
		3714	#include "ev.c"
		3715
		3716	=over 4
		3717
		3718	=item The winsocket select function
		3719
		3720	The winsocket C<select> function doesn't follow POSIX in that it
		3721	requires socket I<handles> and not socket I<file descriptors> (it is
		3722	also extremely buggy). This makes select very inefficient, and also
		3723	requires a mapping from file descriptors to socket handles (the Microsoft
		3724	C runtime provides the function C<_open_osfhandle> for this). See the
		3725	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		3726	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		3727
		3728	The configuration for a "naked" win32 using the Microsoft runtime
		3729	libraries and raw winsocket select is:
		3730
		3731	#define EV_USE_SELECT 1
		3732	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		3733
		3734	Note that winsockets handling of fd sets is O(n), so you can easily get a
		3735	complexity in the O(n²) range when using win32.
		3736
		3737	=item Limited number of file descriptors
		3738
		3739	Windows has numerous arbitrary (and low) limits on things.
		3740
		3741	Early versions of winsocket's select only supported waiting for a maximum
		3742	of C<64> handles (probably owning to the fact that all windows kernels
		3743	can only wait for C<64> things at the same time internally; Microsoft
		3744	recommends spawning a chain of threads and wait for 63 handles and the
		3745	previous thread in each. Great).
		3746
		3747	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		3748	to some high number (e.g. C<2048>) before compiling the winsocket select
		3749	call (which might be in libev or elsewhere, for example, perl does its own
		3750	select emulation on windows).
		3751
		3752	Another limit is the number of file descriptors in the Microsoft runtime
		3753	libraries, which by default is C<64> (there must be a hidden I<64> fetish
		3754	or something like this inside Microsoft). You can increase this by calling
		3755	C<_setmaxstdio>, which can increase this limit to C<2048> (another
		3756	arbitrary limit), but is broken in many versions of the Microsoft runtime
		3757	libraries.
		3758
		3759	This might get you to about C<512> or C<2048> sockets (depending on
		3760	windows version and/or the phase of the moon). To get more, you need to
		3761	wrap all I/O functions and provide your own fd management, but the cost of
		3762	calling select (O(n²)) will likely make this unworkable.
		3763
		3764	=back
		3765
		3766	=head2 PORTABILITY REQUIREMENTS
		3767
		3768	In addition to a working ISO-C implementation and of course the
		3769	backend-specific APIs, libev relies on a few additional extensions:
		3770
		3771	=over 4
		3772
		3773	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		3774	calling conventions regardless of C<ev_watcher_type *>.
		3775
		3776	Libev assumes not only that all watcher pointers have the same internal
		3777	structure (guaranteed by POSIX but not by ISO C for example), but it also
		3778	assumes that the same (machine) code can be used to call any watcher
		3779	callback: The watcher callbacks have different type signatures, but libev
		3780	calls them using an C<ev_watcher *> internally.
		3781
		3782	=item C<sig_atomic_t volatile> must be thread-atomic as well
		3783
		3784	The type C<sig_atomic_t volatile> (or whatever is defined as
		3785	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		3786	threads. This is not part of the specification for C<sig_atomic_t>, but is
		3787	believed to be sufficiently portable.
		3788
		3789	=item C<sigprocmask> must work in a threaded environment
		3790
		3791	Libev uses C<sigprocmask> to temporarily block signals. This is not
		3792	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		3793	pthread implementations will either allow C<sigprocmask> in the "main
		3794	thread" or will block signals process-wide, both behaviours would
		3795	be compatible with libev. Interaction between C<sigprocmask> and
		3796	C<pthread_sigmask> could complicate things, however.
		3797
		3798	The most portable way to handle signals is to block signals in all threads
		3799	except the initial one, and run the default loop in the initial thread as
		3800	well.
		3801
		3802	=item C<long> must be large enough for common memory allocation sizes
		3803
		3804	To improve portability and simplify its API, libev uses C<long> internally
		3805	instead of C<size_t> when allocating its data structures. On non-POSIX
		3806	systems (Microsoft...) this might be unexpectedly low, but is still at
		3807	least 31 bits everywhere, which is enough for hundreds of millions of
		3808	watchers.
		3809
		3810	=item C<double> must hold a time value in seconds with enough accuracy
		3811
		3812	The type C<double> is used to represent timestamps. It is required to
		3813	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
		3814	enough for at least into the year 4000. This requirement is fulfilled by
		3815	implementations implementing IEEE 754 (basically all existing ones).
		3816
		3817	=back
		3818
		3819	If you know of other additional requirements drop me a note.
		3820
		3821
		3822	=head1 ALGORITHMIC COMPLEXITIES
		3823
		3824	In this section the complexities of (many of) the algorithms used inside
		3825	libev will be documented. For complexity discussions about backends see
		3826	the documentation for C<ev_default_init>.
		3827
		3828	All of the following are about amortised time: If an array needs to be
		3829	extended, libev needs to realloc and move the whole array, but this
		3830	happens asymptotically rarer with higher number of elements, so O(1) might
		3831	mean that libev does a lengthy realloc operation in rare cases, but on
		3832	average it is much faster and asymptotically approaches constant time.
		3833
		3834	=over 4
		3835
		3836	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		3837
		3838	This means that, when you have a watcher that triggers in one hour and
		3839	there are 100 watchers that would trigger before that, then inserting will
		3840	have to skip roughly seven (C<ld 100>) of these watchers.
		3841
		3842	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		3843
		3844	That means that changing a timer costs less than removing/adding them,
		3845	as only the relative motion in the event queue has to be paid for.
		3846
		3847	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		3848
		3849	These just add the watcher into an array or at the head of a list.
		3850
		3851	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		3852
		3853	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		3854
		3855	These watchers are stored in lists, so they need to be walked to find the
		3856	correct watcher to remove. The lists are usually short (you don't usually
		3857	have many watchers waiting for the same fd or signal: one is typical, two
		3858	is rare).
		3859
		3860	=item Finding the next timer in each loop iteration: O(1)
		3861
		3862	By virtue of using a binary or 4-heap, the next timer is always found at a
		3863	fixed position in the storage array.
		3864
		3865	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3866
		3867	A change means an I/O watcher gets started or stopped, which requires
		3868	libev to recalculate its status (and possibly tell the kernel, depending
		3869	on backend and whether C<ev_io_set> was used).
		3870
		3871	=item Activating one watcher (putting it into the pending state): O(1)
		3872
		3873	=item Priority handling: O(number_of_priorities)
		3874
		3875	Priorities are implemented by allocating some space for each
		3876	priority. When doing priority-based operations, libev usually has to
		3877	linearly search all the priorities, but starting/stopping and activating
		3878	watchers becomes O(1) with respect to priority handling.
		3879
		3880	=item Sending an ev_async: O(1)
		3881
		3882	=item Processing ev_async_send: O(number_of_async_watchers)
		3883
		3884	=item Processing signals: O(max_signal_number)
		3885
		3886	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3887	calls in the current loop iteration. Checking for async and signal events
		3888	involves iterating over all running async watchers or all signal numbers.
		3889
		3890	=back
		3891
825		3892
826	=head1 AUTHOR	3893	=head1 AUTHOR
827		3894
828	Marc Lehmann <libev@schmorp.de>.	3895	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
829		3896

Diff Legend

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